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);
37 struct dnode_of_data dn;
42 sb_start_pagefault(inode->i_sb);
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);
49 mutex_unlock_op(sbi, ilock);
53 old_blk_addr = dn.data_blkaddr;
55 if (old_blk_addr == NULL_ADDR) {
56 err = reserve_new_block(&dn);
59 mutex_unlock_op(sbi, ilock);
64 mutex_unlock_op(sbi, ilock);
67 if (page->mapping != inode->i_mapping ||
68 page_offset(page) >= i_size_read(inode) ||
69 !PageUptodate(page)) {
76 * check to see if the page is mapped already (no holes)
78 if (PageMappedToDisk(page))
82 wait_on_page_writeback(page);
84 /* page is wholly or partially inside EOF */
85 if (((page->index + 1) << PAGE_CACHE_SHIFT) > i_size_read(inode)) {
87 offset = i_size_read(inode) & ~PAGE_CACHE_MASK;
88 zero_user_segment(page, offset, PAGE_CACHE_SIZE);
91 SetPageUptodate(page);
93 file_update_time(vma->vm_file);
95 sb_end_pagefault(inode->i_sb);
96 return block_page_mkwrite_return(err);
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,
105 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
107 struct inode *inode = file->f_mapping->host;
108 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
110 bool need_cp = false;
111 struct writeback_control wbc = {
112 .sync_mode = WB_SYNC_ALL,
113 .nr_to_write = LONG_MAX,
117 if (inode->i_sb->s_flags & MS_RDONLY)
120 trace_f2fs_sync_file_enter(inode);
121 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
123 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
127 /* guarantee free sections for fsync */
128 f2fs_balance_fs(sbi);
130 mutex_lock(&inode->i_mutex);
132 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
135 if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
137 else if (is_cp_file(inode))
139 else if (!space_for_roll_forward(sbi))
141 else if (!is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
145 /* all the dirty node pages should be flushed for POR */
146 ret = f2fs_sync_fs(inode->i_sb, 1);
148 /* if there is no written node page, write its inode page */
149 while (!sync_node_pages(sbi, inode->i_ino, &wbc)) {
150 ret = f2fs_write_inode(inode, NULL);
154 filemap_fdatawait_range(sbi->node_inode->i_mapping,
156 ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
159 mutex_unlock(&inode->i_mutex);
160 trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
164 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
167 vma->vm_ops = &f2fs_file_vm_ops;
171 static int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
173 int nr_free = 0, ofs = dn->ofs_in_node;
174 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
175 struct f2fs_node *raw_node;
178 raw_node = page_address(dn->node_page);
179 addr = blkaddr_in_node(raw_node) + ofs;
181 for ( ; count > 0; count--, addr++, dn->ofs_in_node++) {
182 block_t blkaddr = le32_to_cpu(*addr);
183 if (blkaddr == NULL_ADDR)
186 update_extent_cache(NULL_ADDR, dn);
187 invalidate_blocks(sbi, blkaddr);
188 dec_valid_block_count(sbi, dn->inode, 1);
192 set_page_dirty(dn->node_page);
195 dn->ofs_in_node = ofs;
197 trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
198 dn->ofs_in_node, nr_free);
202 void truncate_data_blocks(struct dnode_of_data *dn)
204 truncate_data_blocks_range(dn, ADDRS_PER_BLOCK);
207 static void truncate_partial_data_page(struct inode *inode, u64 from)
209 unsigned offset = from & (PAGE_CACHE_SIZE - 1);
215 page = find_data_page(inode, from >> PAGE_CACHE_SHIFT);
220 wait_on_page_writeback(page);
221 zero_user(page, offset, PAGE_CACHE_SIZE - offset);
222 set_page_dirty(page);
223 f2fs_put_page(page, 1);
226 static int truncate_blocks(struct inode *inode, u64 from)
228 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
229 unsigned int blocksize = inode->i_sb->s_blocksize;
230 struct dnode_of_data dn;
232 int count = 0, ilock = -1;
235 trace_f2fs_truncate_blocks_enter(inode, from);
237 free_from = (pgoff_t)
238 ((from + blocksize - 1) >> (sbi->log_blocksize));
240 ilock = mutex_lock_op(sbi);
241 set_new_dnode(&dn, inode, NULL, NULL, 0);
242 err = get_dnode_of_data(&dn, free_from, LOOKUP_NODE);
246 mutex_unlock_op(sbi, ilock);
247 trace_f2fs_truncate_blocks_exit(inode, err);
251 if (IS_INODE(dn.node_page))
252 count = ADDRS_PER_INODE;
254 count = ADDRS_PER_BLOCK;
256 count -= dn.ofs_in_node;
259 if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
260 truncate_data_blocks_range(&dn, count);
266 err = truncate_inode_blocks(inode, free_from);
267 mutex_unlock_op(sbi, ilock);
269 /* lastly zero out the first data page */
270 truncate_partial_data_page(inode, from);
272 trace_f2fs_truncate_blocks_exit(inode, err);
276 void f2fs_truncate(struct inode *inode)
278 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
279 S_ISLNK(inode->i_mode)))
282 trace_f2fs_truncate(inode);
284 if (!truncate_blocks(inode, i_size_read(inode))) {
285 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
286 mark_inode_dirty(inode);
290 static int f2fs_getattr(struct vfsmount *mnt,
291 struct dentry *dentry, struct kstat *stat)
293 struct inode *inode = dentry->d_inode;
294 generic_fillattr(inode, stat);
299 #ifdef CONFIG_F2FS_FS_POSIX_ACL
300 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
302 struct f2fs_inode_info *fi = F2FS_I(inode);
303 unsigned int ia_valid = attr->ia_valid;
305 if (ia_valid & ATTR_UID)
306 inode->i_uid = attr->ia_uid;
307 if (ia_valid & ATTR_GID)
308 inode->i_gid = attr->ia_gid;
309 if (ia_valid & ATTR_ATIME)
310 inode->i_atime = timespec_trunc(attr->ia_atime,
311 inode->i_sb->s_time_gran);
312 if (ia_valid & ATTR_MTIME)
313 inode->i_mtime = timespec_trunc(attr->ia_mtime,
314 inode->i_sb->s_time_gran);
315 if (ia_valid & ATTR_CTIME)
316 inode->i_ctime = timespec_trunc(attr->ia_ctime,
317 inode->i_sb->s_time_gran);
318 if (ia_valid & ATTR_MODE) {
319 umode_t mode = attr->ia_mode;
321 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
323 set_acl_inode(fi, mode);
327 #define __setattr_copy setattr_copy
330 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
332 struct inode *inode = dentry->d_inode;
333 struct f2fs_inode_info *fi = F2FS_I(inode);
336 err = inode_change_ok(inode, attr);
340 if ((attr->ia_valid & ATTR_SIZE) &&
341 attr->ia_size != i_size_read(inode)) {
342 truncate_setsize(inode, attr->ia_size);
343 f2fs_truncate(inode);
344 f2fs_balance_fs(F2FS_SB(inode->i_sb));
347 __setattr_copy(inode, attr);
349 if (attr->ia_valid & ATTR_MODE) {
350 err = f2fs_acl_chmod(inode);
351 if (err || is_inode_flag_set(fi, FI_ACL_MODE)) {
352 inode->i_mode = fi->i_acl_mode;
353 clear_inode_flag(fi, FI_ACL_MODE);
357 mark_inode_dirty(inode);
361 const struct inode_operations f2fs_file_inode_operations = {
362 .getattr = f2fs_getattr,
363 .setattr = f2fs_setattr,
364 .get_acl = f2fs_get_acl,
365 #ifdef CONFIG_F2FS_FS_XATTR
366 .setxattr = generic_setxattr,
367 .getxattr = generic_getxattr,
368 .listxattr = f2fs_listxattr,
369 .removexattr = generic_removexattr,
373 static void fill_zero(struct inode *inode, pgoff_t index,
374 loff_t start, loff_t len)
376 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
383 f2fs_balance_fs(sbi);
385 ilock = mutex_lock_op(sbi);
386 page = get_new_data_page(inode, index, false);
387 mutex_unlock_op(sbi, ilock);
390 wait_on_page_writeback(page);
391 zero_user(page, start, len);
392 set_page_dirty(page);
393 f2fs_put_page(page, 1);
397 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
402 for (index = pg_start; index < pg_end; index++) {
403 struct dnode_of_data dn;
405 set_new_dnode(&dn, inode, NULL, NULL, 0);
406 err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
413 if (dn.data_blkaddr != NULL_ADDR)
414 truncate_data_blocks_range(&dn, 1);
420 static int punch_hole(struct inode *inode, loff_t offset, loff_t len, int mode)
422 pgoff_t pg_start, pg_end;
423 loff_t off_start, off_end;
426 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
427 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
429 off_start = offset & (PAGE_CACHE_SIZE - 1);
430 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
432 if (pg_start == pg_end) {
433 fill_zero(inode, pg_start, off_start,
434 off_end - off_start);
437 fill_zero(inode, pg_start++, off_start,
438 PAGE_CACHE_SIZE - off_start);
440 fill_zero(inode, pg_end, 0, off_end);
442 if (pg_start < pg_end) {
443 struct address_space *mapping = inode->i_mapping;
444 loff_t blk_start, blk_end;
445 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
448 f2fs_balance_fs(sbi);
450 blk_start = pg_start << PAGE_CACHE_SHIFT;
451 blk_end = pg_end << PAGE_CACHE_SHIFT;
452 truncate_inode_pages_range(mapping, blk_start,
455 ilock = mutex_lock_op(sbi);
456 ret = truncate_hole(inode, pg_start, pg_end);
457 mutex_unlock_op(sbi, ilock);
461 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
462 i_size_read(inode) <= (offset + len)) {
463 i_size_write(inode, offset);
464 mark_inode_dirty(inode);
470 static int expand_inode_data(struct inode *inode, loff_t offset,
471 loff_t len, int mode)
473 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
474 pgoff_t index, pg_start, pg_end;
475 loff_t new_size = i_size_read(inode);
476 loff_t off_start, off_end;
479 ret = inode_newsize_ok(inode, (len + offset));
483 pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
484 pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
486 off_start = offset & (PAGE_CACHE_SIZE - 1);
487 off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
489 for (index = pg_start; index <= pg_end; index++) {
490 struct dnode_of_data dn;
493 ilock = mutex_lock_op(sbi);
494 set_new_dnode(&dn, inode, NULL, NULL, 0);
495 ret = get_dnode_of_data(&dn, index, ALLOC_NODE);
497 mutex_unlock_op(sbi, ilock);
501 if (dn.data_blkaddr == NULL_ADDR) {
502 ret = reserve_new_block(&dn);
505 mutex_unlock_op(sbi, ilock);
510 mutex_unlock_op(sbi, ilock);
512 if (pg_start == pg_end)
513 new_size = offset + len;
514 else if (index == pg_start && off_start)
515 new_size = (index + 1) << PAGE_CACHE_SHIFT;
516 else if (index == pg_end)
517 new_size = (index << PAGE_CACHE_SHIFT) + off_end;
519 new_size += PAGE_CACHE_SIZE;
522 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
523 i_size_read(inode) < new_size) {
524 i_size_write(inode, new_size);
525 mark_inode_dirty(inode);
531 static long f2fs_fallocate(struct file *file, int mode,
532 loff_t offset, loff_t len)
534 struct inode *inode = file_inode(file);
537 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
540 if (mode & FALLOC_FL_PUNCH_HOLE)
541 ret = punch_hole(inode, offset, len, mode);
543 ret = expand_inode_data(inode, offset, len, mode);
546 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
547 mark_inode_dirty(inode);
549 trace_f2fs_fallocate(inode, mode, offset, len, ret);
553 #define F2FS_REG_FLMASK (~(FS_DIRSYNC_FL | FS_TOPDIR_FL))
554 #define F2FS_OTHER_FLMASK (FS_NODUMP_FL | FS_NOATIME_FL)
556 static inline __u32 f2fs_mask_flags(umode_t mode, __u32 flags)
560 else if (S_ISREG(mode))
561 return flags & F2FS_REG_FLMASK;
563 return flags & F2FS_OTHER_FLMASK;
566 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
568 struct inode *inode = file_inode(filp);
569 struct f2fs_inode_info *fi = F2FS_I(inode);
574 case FS_IOC_GETFLAGS:
575 flags = fi->i_flags & FS_FL_USER_VISIBLE;
576 return put_user(flags, (int __user *) arg);
577 case FS_IOC_SETFLAGS:
579 unsigned int oldflags;
581 ret = mnt_want_write(filp->f_path.mnt);
585 if (!inode_owner_or_capable(inode)) {
590 if (get_user(flags, (int __user *) arg)) {
595 flags = f2fs_mask_flags(inode->i_mode, flags);
597 mutex_lock(&inode->i_mutex);
599 oldflags = fi->i_flags;
601 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
602 if (!capable(CAP_LINUX_IMMUTABLE)) {
603 mutex_unlock(&inode->i_mutex);
609 flags = flags & FS_FL_USER_MODIFIABLE;
610 flags |= oldflags & ~FS_FL_USER_MODIFIABLE;
612 mutex_unlock(&inode->i_mutex);
614 f2fs_set_inode_flags(inode);
615 inode->i_ctime = CURRENT_TIME;
616 mark_inode_dirty(inode);
618 mnt_drop_write(filp->f_path.mnt);
627 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
630 case F2FS_IOC32_GETFLAGS:
631 cmd = F2FS_IOC_GETFLAGS;
633 case F2FS_IOC32_SETFLAGS:
634 cmd = F2FS_IOC_SETFLAGS;
639 return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
643 const struct file_operations f2fs_file_operations = {
644 .llseek = generic_file_llseek,
645 .read = do_sync_read,
646 .write = do_sync_write,
647 .aio_read = generic_file_aio_read,
648 .aio_write = generic_file_aio_write,
649 .open = generic_file_open,
650 .mmap = f2fs_file_mmap,
651 .fsync = f2fs_sync_file,
652 .fallocate = f2fs_fallocate,
653 .unlocked_ioctl = f2fs_ioctl,
655 .compat_ioctl = f2fs_compat_ioctl,
657 .splice_read = generic_file_splice_read,
658 .splice_write = generic_file_splice_write,