1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
6 * Code to deal with the mess that is clustered mmap.
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation; either
13 * version 2 of the License, or (at your option) any later version.
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 021110-1307, USA.
27 #include <linux/types.h>
28 #include <linux/highmem.h>
29 #include <linux/pagemap.h>
30 #include <linux/uio.h>
31 #include <linux/signal.h>
32 #include <linux/rbtree.h>
34 #define MLOG_MASK_PREFIX ML_FILE_IO
35 #include <cluster/masklog.h>
47 static int ocfs2_fault(struct vm_area_struct *area, struct vm_fault *vmf)
52 mlog_entry("(area=%p, page offset=%lu)\n", area, vmf->pgoff);
54 ocfs2_block_signals(&oldset);
55 ret = filemap_fault(area, vmf);
56 ocfs2_unblock_signals(&oldset);
58 mlog_exit_ptr(vmf->page);
62 static int __ocfs2_page_mkwrite(struct inode *inode, struct buffer_head *di_bh,
66 struct address_space *mapping = inode->i_mapping;
67 loff_t pos = page_offset(page);
68 unsigned int len = PAGE_CACHE_SIZE;
70 struct page *locked_page = NULL;
72 loff_t size = i_size_read(inode);
75 * Another node might have truncated while we were waiting on
78 last_index = size >> PAGE_CACHE_SHIFT;
79 if (page->index > last_index) {
85 * The i_size check above doesn't catch the case where nodes
86 * truncated and then re-extended the file. We'll re-check the
87 * page mapping after taking the page lock inside of
88 * ocfs2_write_begin_nolock().
90 if (!PageUptodate(page) || page->mapping != inode->i_mapping) {
92 * the page has been umapped in ocfs2_data_downconvert_worker.
93 * So return 0 here and let VFS retry.
100 * Call ocfs2_write_begin() and ocfs2_write_end() to take
101 * advantage of the allocation code there. We pass a write
102 * length of the whole page (chopped to i_size) to make sure
103 * the whole thing is allocated.
105 * Since we know the page is up to date, we don't have to
106 * worry about ocfs2_write_begin() skipping some buffer reads
107 * because the "write" would invalidate their data.
109 if (page->index == last_index)
110 len = size & ~PAGE_CACHE_MASK;
112 ret = ocfs2_write_begin_nolock(mapping, pos, len, 0, &locked_page,
113 &fsdata, di_bh, page);
120 ret = ocfs2_write_end_nolock(mapping, pos, len, len, locked_page,
132 static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
134 struct page *page = vmf->page;
135 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
136 struct buffer_head *di_bh = NULL;
140 ocfs2_block_signals(&oldset);
143 * The cluster locks taken will block a truncate from another
144 * node. Taking the data lock will also ensure that we don't
145 * attempt page truncation as part of a downconvert.
147 ret = ocfs2_inode_lock(inode, &di_bh, 1);
154 * The alloc sem should be enough to serialize with
155 * ocfs2_truncate_file() changing i_size as well as any thread
156 * modifying the inode btree.
158 down_write(&OCFS2_I(inode)->ip_alloc_sem);
160 ret = __ocfs2_page_mkwrite(inode, di_bh, page);
162 up_write(&OCFS2_I(inode)->ip_alloc_sem);
165 ocfs2_inode_unlock(inode, 1);
168 ocfs2_unblock_signals(&oldset);
170 ret = VM_FAULT_SIGBUS;
174 static const struct vm_operations_struct ocfs2_file_vm_ops = {
175 .fault = ocfs2_fault,
176 .page_mkwrite = ocfs2_page_mkwrite,
179 int ocfs2_mmap(struct file *file, struct vm_area_struct *vma)
181 int ret = 0, lock_level = 0;
183 ret = ocfs2_inode_lock_atime(file->f_dentry->d_inode,
184 file->f_vfsmnt, &lock_level);
189 ocfs2_inode_unlock(file->f_dentry->d_inode, lock_level);
191 vma->vm_ops = &ocfs2_file_vm_ops;
192 vma->vm_flags |= VM_CAN_NONLINEAR;