Commit | Line | Data |
---|---|---|
ac27a0ec | 1 | /* |
617ba13b | 2 | * linux/fs/ext4/file.c |
ac27a0ec DK |
3 | * |
4 | * Copyright (C) 1992, 1993, 1994, 1995 | |
5 | * Remy Card (card@masi.ibp.fr) | |
6 | * Laboratoire MASI - Institut Blaise Pascal | |
7 | * Universite Pierre et Marie Curie (Paris VI) | |
8 | * | |
9 | * from | |
10 | * | |
11 | * linux/fs/minix/file.c | |
12 | * | |
13 | * Copyright (C) 1991, 1992 Linus Torvalds | |
14 | * | |
617ba13b | 15 | * ext4 fs regular file handling primitives |
ac27a0ec DK |
16 | * |
17 | * 64-bit file support on 64-bit platforms by Jakub Jelinek | |
18 | * (jj@sunsite.ms.mff.cuni.cz) | |
19 | */ | |
20 | ||
21 | #include <linux/time.h> | |
22 | #include <linux/fs.h> | |
dab291af | 23 | #include <linux/jbd2.h> |
bc0b0d6d TT |
24 | #include <linux/mount.h> |
25 | #include <linux/path.h> | |
3dcf5451 CH |
26 | #include "ext4.h" |
27 | #include "ext4_jbd2.h" | |
ac27a0ec DK |
28 | #include "xattr.h" |
29 | #include "acl.h" | |
30 | ||
31 | /* | |
32 | * Called when an inode is released. Note that this is different | |
617ba13b | 33 | * from ext4_file_open: open gets called at every open, but release |
ac27a0ec DK |
34 | * gets called only when /all/ the files are closed. |
35 | */ | |
af5bc92d | 36 | static int ext4_release_file(struct inode *inode, struct file *filp) |
ac27a0ec | 37 | { |
7d8f9f7d TT |
38 | if (EXT4_I(inode)->i_state & EXT4_STATE_DA_ALLOC_CLOSE) { |
39 | ext4_alloc_da_blocks(inode); | |
40 | EXT4_I(inode)->i_state &= ~EXT4_STATE_DA_ALLOC_CLOSE; | |
41 | } | |
ac27a0ec DK |
42 | /* if we are the last writer on the inode, drop the block reservation */ |
43 | if ((filp->f_mode & FMODE_WRITE) && | |
d6014301 AK |
44 | (atomic_read(&inode->i_writecount) == 1) && |
45 | !EXT4_I(inode)->i_reserved_data_blocks) | |
ac27a0ec | 46 | { |
0e855ac8 | 47 | down_write(&EXT4_I(inode)->i_data_sem); |
c2ea3fde | 48 | ext4_discard_preallocations(inode); |
0e855ac8 | 49 | up_write(&EXT4_I(inode)->i_data_sem); |
ac27a0ec DK |
50 | } |
51 | if (is_dx(inode) && filp->private_data) | |
617ba13b | 52 | ext4_htree_free_dir_info(filp->private_data); |
ac27a0ec DK |
53 | |
54 | return 0; | |
55 | } | |
56 | ||
57 | static ssize_t | |
617ba13b | 58 | ext4_file_write(struct kiocb *iocb, const struct iovec *iov, |
ac27a0ec DK |
59 | unsigned long nr_segs, loff_t pos) |
60 | { | |
61 | struct file *file = iocb->ki_filp; | |
9d549890 | 62 | struct inode *inode = file->f_path.dentry->d_inode; |
ac27a0ec DK |
63 | ssize_t ret; |
64 | int err; | |
65 | ||
e2b46574 ES |
66 | /* |
67 | * If we have encountered a bitmap-format file, the size limit | |
68 | * is smaller than s_maxbytes, which is for extent-mapped files. | |
69 | */ | |
70 | ||
71 | if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL)) { | |
72 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
73 | size_t length = iov_length(iov, nr_segs); | |
ac27a0ec | 74 | |
e2b46574 ES |
75 | if (pos > sbi->s_bitmap_maxbytes) |
76 | return -EFBIG; | |
77 | ||
78 | if (pos + length > sbi->s_bitmap_maxbytes) { | |
79 | nr_segs = iov_shorten((struct iovec *)iov, nr_segs, | |
80 | sbi->s_bitmap_maxbytes - pos); | |
81 | } | |
82 | } | |
83 | ||
84 | ret = generic_file_aio_write(iocb, iov, nr_segs, pos); | |
ac27a0ec DK |
85 | /* |
86 | * Skip flushing if there was an error, or if nothing was written. | |
87 | */ | |
88 | if (ret <= 0) | |
89 | return ret; | |
90 | ||
91 | /* | |
92 | * If the inode is IS_SYNC, or is O_SYNC and we are doing data | |
93 | * journalling then we need to make sure that we force the transaction | |
94 | * to disk to keep all metadata uptodate synchronously. | |
95 | */ | |
96 | if (file->f_flags & O_SYNC) { | |
97 | /* | |
98 | * If we are non-data-journaled, then the dirty data has | |
99 | * already been flushed to backing store by generic_osync_inode, | |
100 | * and the inode has been flushed too if there have been any | |
101 | * modifications other than mere timestamp updates. | |
102 | * | |
103 | * Open question --- do we care about flushing timestamps too | |
104 | * if the inode is IS_SYNC? | |
105 | */ | |
617ba13b | 106 | if (!ext4_should_journal_data(inode)) |
ac27a0ec DK |
107 | return ret; |
108 | ||
109 | goto force_commit; | |
110 | } | |
111 | ||
112 | /* | |
113 | * So we know that there has been no forced data flush. If the inode | |
114 | * is marked IS_SYNC, we need to force one ourselves. | |
115 | */ | |
116 | if (!IS_SYNC(inode)) | |
117 | return ret; | |
118 | ||
119 | /* | |
120 | * Open question #2 --- should we force data to disk here too? If we | |
121 | * don't, the only impact is that data=writeback filesystems won't | |
122 | * flush data to disk automatically on IS_SYNC, only metadata (but | |
123 | * historically, that is what ext2 has done.) | |
124 | */ | |
125 | ||
126 | force_commit: | |
617ba13b | 127 | err = ext4_force_commit(inode->i_sb); |
ac27a0ec DK |
128 | if (err) |
129 | return err; | |
130 | return ret; | |
131 | } | |
132 | ||
2e9ee850 AK |
133 | static struct vm_operations_struct ext4_file_vm_ops = { |
134 | .fault = filemap_fault, | |
135 | .page_mkwrite = ext4_page_mkwrite, | |
136 | }; | |
137 | ||
138 | static int ext4_file_mmap(struct file *file, struct vm_area_struct *vma) | |
139 | { | |
140 | struct address_space *mapping = file->f_mapping; | |
141 | ||
142 | if (!mapping->a_ops->readpage) | |
143 | return -ENOEXEC; | |
144 | file_accessed(file); | |
145 | vma->vm_ops = &ext4_file_vm_ops; | |
146 | vma->vm_flags |= VM_CAN_NONLINEAR; | |
147 | return 0; | |
148 | } | |
149 | ||
bc0b0d6d TT |
150 | static int ext4_file_open(struct inode * inode, struct file * filp) |
151 | { | |
152 | struct super_block *sb = inode->i_sb; | |
153 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
154 | struct vfsmount *mnt = filp->f_path.mnt; | |
155 | struct path path; | |
156 | char buf[64], *cp; | |
157 | ||
158 | if (unlikely(!(sbi->s_mount_flags & EXT4_MF_MNTDIR_SAMPLED) && | |
159 | !(sb->s_flags & MS_RDONLY))) { | |
160 | sbi->s_mount_flags |= EXT4_MF_MNTDIR_SAMPLED; | |
161 | /* | |
162 | * Sample where the filesystem has been mounted and | |
163 | * store it in the superblock for sysadmin convenience | |
164 | * when trying to sort through large numbers of block | |
165 | * devices or filesystem images. | |
166 | */ | |
167 | memset(buf, 0, sizeof(buf)); | |
168 | path.mnt = mnt->mnt_parent; | |
169 | path.dentry = mnt->mnt_mountpoint; | |
170 | path_get(&path); | |
171 | cp = d_path(&path, buf, sizeof(buf)); | |
172 | path_put(&path); | |
173 | if (!IS_ERR(cp)) { | |
174 | memcpy(sbi->s_es->s_last_mounted, cp, | |
175 | sizeof(sbi->s_es->s_last_mounted)); | |
176 | sb->s_dirt = 1; | |
177 | } | |
178 | } | |
179 | return generic_file_open(inode, filp); | |
180 | } | |
181 | ||
617ba13b | 182 | const struct file_operations ext4_file_operations = { |
ac27a0ec DK |
183 | .llseek = generic_file_llseek, |
184 | .read = do_sync_read, | |
185 | .write = do_sync_write, | |
186 | .aio_read = generic_file_aio_read, | |
617ba13b | 187 | .aio_write = ext4_file_write, |
5cdd7b2d | 188 | .unlocked_ioctl = ext4_ioctl, |
ac27a0ec | 189 | #ifdef CONFIG_COMPAT |
617ba13b | 190 | .compat_ioctl = ext4_compat_ioctl, |
ac27a0ec | 191 | #endif |
2e9ee850 | 192 | .mmap = ext4_file_mmap, |
bc0b0d6d | 193 | .open = ext4_file_open, |
617ba13b MC |
194 | .release = ext4_release_file, |
195 | .fsync = ext4_sync_file, | |
ac27a0ec DK |
196 | .splice_read = generic_file_splice_read, |
197 | .splice_write = generic_file_splice_write, | |
198 | }; | |
199 | ||
754661f1 | 200 | const struct inode_operations ext4_file_inode_operations = { |
617ba13b MC |
201 | .truncate = ext4_truncate, |
202 | .setattr = ext4_setattr, | |
3e3398a0 | 203 | .getattr = ext4_getattr, |
03010a33 | 204 | #ifdef CONFIG_EXT4_FS_XATTR |
ac27a0ec DK |
205 | .setxattr = generic_setxattr, |
206 | .getxattr = generic_getxattr, | |
617ba13b | 207 | .listxattr = ext4_listxattr, |
ac27a0ec DK |
208 | .removexattr = generic_removexattr, |
209 | #endif | |
617ba13b | 210 | .permission = ext4_permission, |
a2df2a63 | 211 | .fallocate = ext4_fallocate, |
6873fa0d | 212 | .fiemap = ext4_fiemap, |
ac27a0ec DK |
213 | }; |
214 |