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>
29 static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
32 struct page *page = vmf->page;
33 struct inode *inode = file_inode(vma->vm_file);
34 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 */
44 ilock = mutex_lock_op(sbi);
45 set_new_dnode(&dn, inode, NULL, NULL, 0);
46 err = get_dnode_of_data(&dn, page->index, ALLOC_NODE);
48 mutex_unlock_op(sbi, ilock);
52 old_blk_addr = dn.data_blkaddr;
54 if (old_blk_addr == NULL_ADDR) {
55 err = reserve_new_block(&dn);
58 mutex_unlock_op(sbi, ilock);
63 mutex_unlock_op(sbi, ilock);
66 if (page->mapping != inode->i_mapping ||
67 page_offset(page) >= i_size_read(inode) ||
68 !PageUptodate(page)) {
75 * check to see if the page is mapped already (no holes)
77 if (PageMappedToDisk(page))
81 wait_on_page_writeback(page);
83 /* page is wholly or partially inside EOF */
84 if (((page->index + 1) << PAGE_CACHE_SHIFT) > i_size_read(inode)) {
86 offset = i_size_read(inode) & ~PAGE_CACHE_MASK;
87 zero_user_segment(page, offset, PAGE_CACHE_SIZE);
90 SetPageUptodate(page);
92 file_update_time(vma->vm_file);
94 sb_end_pagefault(inode->i_sb);
95 return block_page_mkwrite_return(err);
98 static const struct vm_operations_struct f2fs_file_vm_ops = {
99 .fault = filemap_fault,
100 .page_mkwrite = f2fs_vm_page_mkwrite,
101 .remap_pages = generic_file_remap_pages,
104 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
106 struct inode *inode = file->f_mapping->host;
107 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
109 bool need_cp = false;
110 struct writeback_control wbc = {
111 .sync_mode = WB_SYNC_ALL,
112 .nr_to_write = LONG_MAX,
116 if (inode->i_sb->s_flags & MS_RDONLY)
119 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
123 /* guarantee free sections for fsync */
124 f2fs_balance_fs(sbi);
126 mutex_lock(&inode->i_mutex);
128 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
131 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
133 else if (is_cp_file(inode))
135 else if (!space_for_roll_forward(sbi))
137 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
141 /* all the dirty node pages should be flushed for POR */
142 ret = f2fs_sync_fs(inode->i_sb, 1);
144 /* if there is no written node page, write its inode page */
145 while (!sync_node_pages(sbi, inode->i_ino, &wbc)) {
146 ret = f2fs_write_inode(inode, NULL);
150 filemap_fdatawait_range(sbi->node_inode->i_mapping,
152 ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
155 mutex_unlock(&inode->i_mutex);
159 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
162 vma->vm_ops = &f2fs_file_vm_ops;
166 static int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
168 int nr_free = 0, ofs = dn->ofs_in_node;
169 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
170 struct f2fs_node *raw_node;
173 raw_node = page_address(dn->node_page);
174 addr = blkaddr_in_node(raw_node) + ofs;
176 for ( ; count > 0; count--, addr++, dn->ofs_in_node++) {
177 block_t blkaddr = le32_to_cpu(*addr);
178 if (blkaddr == NULL_ADDR)
181 update_extent_cache(NULL_ADDR, dn);
182 invalidate_blocks(sbi, blkaddr);
183 dec_valid_block_count(sbi, dn->inode, 1);
187 set_page_dirty(dn->node_page);
190 dn->ofs_in_node = ofs;
194 void truncate_data_blocks(struct dnode_of_data *dn)
196 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
199 static void truncate_partial_data_page(struct inode *inode, u64 from)
201 unsigned offset = from & (PAGE_CACHE_SIZE - 1);
207 page = find_data_page(inode, from >> PAGE_CACHE_SHIFT);
212 wait_on_page_writeback(page);
213 zero_user(page, offset, PAGE_CACHE_SIZE - offset);
214 set_page_dirty(page);
215 f2fs_put_page(page, 1);
218 static int truncate_blocks(struct inode *inode, u64 from)
220 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
221 unsigned int blocksize = inode->i_sb->s_blocksize;
222 struct dnode_of_data dn;
224 int count = 0, ilock = -1;
227 free_from = (pgoff_t)
228 ((from + blocksize - 1) >> (sbi->log_blocksize));
230 ilock = mutex_lock_op(sbi);
231 set_new_dnode(&dn, inode, NULL, NULL, 0);
232 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
236 mutex_unlock_op(sbi, ilock);
240 if (IS_INODE(dn.node_page))
241 count = ADDRS_PER_INODE;
243 count = ADDRS_PER_BLOCK;
245 count -= dn.ofs_in_node;
248 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
249 truncate_data_blocks_range(&dn, count);
255 err = truncate_inode_blocks(inode, free_from);
256 mutex_unlock_op(sbi, ilock);
258 /* lastly zero out the first data page */
259 truncate_partial_data_page(inode, from);
264 void f2fs_truncate(struct inode *inode)
266 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
267 S_ISLNK(inode->i_mode)))
270 if (!truncate_blocks(inode, i_size_read(inode))) {
271 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
272 mark_inode_dirty(inode);
276 static int f2fs_getattr(struct vfsmount *mnt,
277 struct dentry *dentry, struct kstat *stat)
279 struct inode *inode = dentry->d_inode;
280 generic_fillattr(inode, stat);
285 #ifdef CONFIG_F2FS_FS_POSIX_ACL
286 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
288 struct f2fs_inode_info *fi = F2FS_I(inode);
289 unsigned int ia_valid = attr->ia_valid;
291 if (ia_valid & ATTR_UID)
292 inode->i_uid = attr->ia_uid;
293 if (ia_valid & ATTR_GID)
294 inode->i_gid = attr->ia_gid;
295 if (ia_valid & ATTR_ATIME)
296 inode->i_atime = timespec_trunc(attr->ia_atime,
297 inode->i_sb->s_time_gran);
298 if (ia_valid & ATTR_MTIME)
299 inode->i_mtime = timespec_trunc(attr->ia_mtime,
300 inode->i_sb->s_time_gran);
301 if (ia_valid & ATTR_CTIME)
302 inode->i_ctime = timespec_trunc(attr->ia_ctime,
303 inode->i_sb->s_time_gran);
304 if (ia_valid & ATTR_MODE) {
305 umode_t mode = attr->ia_mode;
307 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
309 set_acl_inode(fi, mode);
313 #define __setattr_copy setattr_copy
316 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
318 struct inode *inode = dentry->d_inode;
319 struct f2fs_inode_info *fi = F2FS_I(inode);
322 err = inode_change_ok(inode, attr);
326 if ((attr->ia_valid & ATTR_SIZE) &&
327 attr->ia_size != i_size_read(inode)) {
328 truncate_setsize(inode, attr->ia_size);
329 f2fs_truncate(inode);
330 f2fs_balance_fs(F2FS_SB(inode->i_sb));
333 __setattr_copy(inode, attr);
335 if (attr->ia_valid & ATTR_MODE) {
336 err = f2fs_acl_chmod(inode);
337 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
338 inode->i_mode = fi->i_acl_mode;
339 clear_inode_flag(fi, FI_ACL_MODE);
343 mark_inode_dirty(inode);
347 const struct inode_operations f2fs_file_inode_operations = {
348 .getattr = f2fs_getattr,
349 .setattr = f2fs_setattr,
350 .get_acl = f2fs_get_acl,
351 #ifdef CONFIG_F2FS_FS_XATTR
352 .setxattr = generic_setxattr,
353 .getxattr = generic_getxattr,
354 .listxattr = f2fs_listxattr,
355 .removexattr = generic_removexattr,
359 static void fill_zero(struct inode *inode, pgoff_t index,
360 loff_t start, loff_t len)
362 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
369 f2fs_balance_fs(sbi);
371 ilock = mutex_lock_op(sbi);
372 page = get_new_data_page(inode, index, false);
373 mutex_unlock_op(sbi, ilock);
376 wait_on_page_writeback(page);
377 zero_user(page, start, len);
378 set_page_dirty(page);
379 f2fs_put_page(page, 1);
383 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
388 for (index = pg_start; index < pg_end; index++) {
389 struct dnode_of_data dn;
391 set_new_dnode(&dn, inode, NULL, NULL, 0);
392 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
399 if (dn.data_blkaddr != NULL_ADDR)
400 truncate_data_blocks_range(&dn, 1);
406 static int punch_hole(struct inode *inode, loff_t offset, loff_t len, int mode)
408 pgoff_t pg_start, pg_end;
409 loff_t off_start, off_end;
412 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
413 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
415 off_start = offset & (PAGE_CACHE_SIZE - 1);
416 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
418 if (pg_start == pg_end) {
419 fill_zero(inode, pg_start, off_start,
420 off_end - off_start);
423 fill_zero(inode, pg_start++, off_start,
424 PAGE_CACHE_SIZE - off_start);
426 fill_zero(inode, pg_end, 0, off_end);
428 if (pg_start < pg_end) {
429 struct address_space *mapping = inode->i_mapping;
430 loff_t blk_start, blk_end;
431 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
434 f2fs_balance_fs(sbi);
436 blk_start = pg_start << PAGE_CACHE_SHIFT;
437 blk_end = pg_end << PAGE_CACHE_SHIFT;
438 truncate_inode_pages_range(mapping, blk_start,
441 ilock = mutex_lock_op(sbi);
442 ret = truncate_hole(inode, pg_start, pg_end);
443 mutex_unlock_op(sbi, ilock);
447 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
448 i_size_read(inode) <= (offset + len)) {
449 i_size_write(inode, offset);
450 mark_inode_dirty(inode);
456 static int expand_inode_data(struct inode *inode, loff_t offset,
457 loff_t len, int mode)
459 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
460 pgoff_t index, pg_start, pg_end;
461 loff_t new_size = i_size_read(inode);
462 loff_t off_start, off_end;
465 ret = inode_newsize_ok(inode, (len + offset));
469 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
470 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
472 off_start = offset & (PAGE_CACHE_SIZE - 1);
473 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
475 for (index = pg_start; index <= pg_end; index++) {
476 struct dnode_of_data dn;
479 ilock = mutex_lock_op(sbi);
480 set_new_dnode(&dn, inode, NULL, NULL, 0);
481 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
483 mutex_unlock_op(sbi, ilock);
487 if (dn.data_blkaddr == NULL_ADDR) {
488 ret = reserve_new_block(&dn);
491 mutex_unlock_op(sbi, ilock);
496 mutex_unlock_op(sbi, ilock);
498 if (pg_start == pg_end)
499 new_size = offset + len;
500 else if (index == pg_start && off_start)
501 new_size = (index + 1) << PAGE_CACHE_SHIFT;
502 else if (index == pg_end)
503 new_size = (index << PAGE_CACHE_SHIFT) + off_end;
505 new_size += PAGE_CACHE_SIZE;
508 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
509 i_size_read(inode) < new_size) {
510 i_size_write(inode, new_size);
511 mark_inode_dirty(inode);
517 static long f2fs_fallocate(struct file *file, int mode,
518 loff_t offset, loff_t len)
520 struct inode *inode = file_inode(file);
523 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
526 if (mode & FALLOC_FL_PUNCH_HOLE)
527 ret = punch_hole(inode, offset, len, mode);
529 ret = expand_inode_data(inode, offset, len, mode);
532 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
533 mark_inode_dirty(inode);
538 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
539 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
541 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
545 else if (S_ISREG(mode))
546 return flags & F2FS_REG_FLMASK;
548 return flags & F2FS_OTHER_FLMASK;
551 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
553 struct inode *inode = file_inode(filp);
554 struct f2fs_inode_info *fi = F2FS_I(inode);
559 case FS_IOC_GETFLAGS:
560 flags = fi->i_flags & FS_FL_USER_VISIBLE;
561 return put_user(flags, (int __user *) arg);
562 case FS_IOC_SETFLAGS:
564 unsigned int oldflags;
566 ret = mnt_want_write(filp->f_path.mnt);
570 if (!inode_owner_or_capable(inode)) {
575 if (get_user(flags, (int __user *) arg)) {
580 flags = f2fs_mask_flags(inode->i_mode, flags);
582 mutex_lock(&inode->i_mutex);
584 oldflags = fi->i_flags;
586 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
587 if (!capable(CAP_LINUX_IMMUTABLE)) {
588 mutex_unlock(&inode->i_mutex);
594 flags = flags & FS_FL_USER_MODIFIABLE;
595 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
597 mutex_unlock(&inode->i_mutex);
599 f2fs_set_inode_flags(inode);
600 inode->i_ctime = CURRENT_TIME;
601 mark_inode_dirty(inode);
603 mnt_drop_write(filp->f_path.mnt);
612 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
615 case F2FS_IOC32_GETFLAGS:
616 cmd = F2FS_IOC_GETFLAGS;
618 case F2FS_IOC32_SETFLAGS:
619 cmd = F2FS_IOC_SETFLAGS;
624 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
628 const struct file_operations f2fs_file_operations = {
629 .llseek = generic_file_llseek,
630 .read = do_sync_read,
631 .write = do_sync_write,
632 .aio_read = generic_file_aio_read,
633 .aio_write = generic_file_aio_write,
634 .open = generic_file_open,
635 .mmap = f2fs_file_mmap,
636 .fsync = f2fs_sync_file,
637 .fallocate = f2fs_fallocate,
638 .unlocked_ioctl = f2fs_ioctl,
640 .compat_ioctl = f2fs_compat_ioctl,
642 .splice_read = generic_file_splice_read,
643 .splice_write = generic_file_splice_write,