blob: e4095e988ebacb379a7784f72f52f78f51176ac4 [file] [log] [blame]
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001/*
Mingming Cao617ba132006-10-11 01:20:53 -07002 * linux/fs/ext4/file.c
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003 *
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 *
Mingming Cao617ba132006-10-11 01:20:53 -070015 * ext4 fs regular file handling primitives
Dave Kleikampac27a0e2006-10-11 01:20:50 -070016 *
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>
Mingming Caodab291a2006-10-11 01:21:01 -070023#include <linux/jbd2.h>
Theodore Ts'obc0b0d62009-06-13 10:09:48 -040024#include <linux/mount.h>
25#include <linux/path.h>
Christoph Hellwig871a2932010-03-03 09:05:07 -050026#include <linux/quotaops.h>
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040027#include "ext4.h"
28#include "ext4_jbd2.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070029#include "xattr.h"
30#include "acl.h"
31
32/*
33 * Called when an inode is released. Note that this is different
Mingming Cao617ba132006-10-11 01:20:53 -070034 * from ext4_file_open: open gets called at every open, but release
Dave Kleikampac27a0e2006-10-11 01:20:50 -070035 * gets called only when /all/ the files are closed.
36 */
Theodore Ts'oaf5bc922008-09-08 22:25:24 -040037static int ext4_release_file(struct inode *inode, struct file *filp)
Dave Kleikampac27a0e2006-10-11 01:20:50 -070038{
Theodore Ts'o19f5fb72010-01-24 14:34:07 -050039 if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE)) {
Theodore Ts'o7d8f9f72009-02-24 08:21:14 -050040 ext4_alloc_da_blocks(inode);
Theodore Ts'o19f5fb72010-01-24 14:34:07 -050041 ext4_clear_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE);
Theodore Ts'o7d8f9f72009-02-24 08:21:14 -050042 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -070043 /* if we are the last writer on the inode, drop the block reservation */
44 if ((filp->f_mode & FMODE_WRITE) &&
Aneesh Kumar K.Vd6014302009-03-27 22:36:43 -040045 (atomic_read(&inode->i_writecount) == 1) &&
46 !EXT4_I(inode)->i_reserved_data_blocks)
Dave Kleikampac27a0e2006-10-11 01:20:50 -070047 {
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -050048 down_write(&EXT4_I(inode)->i_data_sem);
Theodore Ts'oc2ea3fd2008-10-10 09:40:52 -040049 ext4_discard_preallocations(inode);
Aneesh Kumar K.V0e855ac2008-01-28 23:58:26 -050050 up_write(&EXT4_I(inode)->i_data_sem);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070051 }
52 if (is_dx(inode) && filp->private_data)
Mingming Cao617ba132006-10-11 01:20:53 -070053 ext4_htree_free_dir_info(filp->private_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070054
55 return 0;
56}
57
Eric Sandeene9e3bce2011-02-12 08:17:34 -050058static void ext4_aiodio_wait(struct inode *inode)
59{
60 wait_queue_head_t *wq = ext4_ioend_wq(inode);
61
62 wait_event(*wq, (atomic_read(&EXT4_I(inode)->i_aiodio_unwritten) == 0));
63}
64
65/*
66 * This tests whether the IO in question is block-aligned or not.
67 * Ext4 utilizes unwritten extents when hole-filling during direct IO, and they
68 * are converted to written only after the IO is complete. Until they are
69 * mapped, these blocks appear as holes, so dio_zero_block() will assume that
70 * it needs to zero out portions of the start and/or end block. If 2 AIO
71 * threads are at work on the same unwritten block, they must be synchronized
72 * or one thread will zero the other's data, causing corruption.
73 */
74static int
75ext4_unaligned_aio(struct inode *inode, const struct iovec *iov,
76 unsigned long nr_segs, loff_t pos)
77{
78 struct super_block *sb = inode->i_sb;
79 int blockmask = sb->s_blocksize - 1;
80 size_t count = iov_length(iov, nr_segs);
81 loff_t final_size = pos + count;
82
83 if (pos >= inode->i_size)
84 return 0;
85
86 if ((pos & blockmask) || (final_size & blockmask))
87 return 1;
88
89 return 0;
90}
91
Dave Kleikampac27a0e2006-10-11 01:20:50 -070092static ssize_t
Mingming Cao617ba132006-10-11 01:20:53 -070093ext4_file_write(struct kiocb *iocb, const struct iovec *iov,
Dave Kleikampac27a0e2006-10-11 01:20:50 -070094 unsigned long nr_segs, loff_t pos)
95{
Jan Kara0d34ec62009-08-18 17:48:27 +020096 struct inode *inode = iocb->ki_filp->f_path.dentry->d_inode;
Eric Sandeene9e3bce2011-02-12 08:17:34 -050097 int unaligned_aio = 0;
98 int ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -070099
Eric Sandeene2b46572008-01-28 23:58:27 -0500100 /*
101 * If we have encountered a bitmap-format file, the size limit
102 * is smaller than s_maxbytes, which is for extent-mapped files.
103 */
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700104
Dmitry Monakhov12e9b892010-05-16 22:00:00 -0400105 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) {
Eric Sandeene2b46572008-01-28 23:58:27 -0500106 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
107 size_t length = iov_length(iov, nr_segs);
108
Toshiyuki Okajimad889dc82010-07-27 11:56:07 -0400109 if ((pos > sbi->s_bitmap_maxbytes ||
110 (pos == sbi->s_bitmap_maxbytes && length > 0)))
Eric Sandeene2b46572008-01-28 23:58:27 -0500111 return -EFBIG;
112
113 if (pos + length > sbi->s_bitmap_maxbytes) {
114 nr_segs = iov_shorten((struct iovec *)iov, nr_segs,
115 sbi->s_bitmap_maxbytes - pos);
116 }
Eric Sandeene9e3bce2011-02-12 08:17:34 -0500117 } else if (unlikely((iocb->ki_filp->f_flags & O_DIRECT) &&
118 !is_sync_kiocb(iocb))) {
119 unaligned_aio = ext4_unaligned_aio(inode, iov, nr_segs, pos);
Eric Sandeene2b46572008-01-28 23:58:27 -0500120 }
121
Eric Sandeene9e3bce2011-02-12 08:17:34 -0500122 /* Unaligned direct AIO must be serialized; see comment above */
123 if (unaligned_aio) {
124 static unsigned long unaligned_warn_time;
125
126 /* Warn about this once per day */
127 if (printk_timed_ratelimit(&unaligned_warn_time, 60*60*24*HZ))
128 ext4_msg(inode->i_sb, KERN_WARNING,
129 "Unaligned AIO/DIO on inode %ld by %s; "
130 "performance will be poor.",
131 inode->i_ino, current->comm);
132 mutex_lock(ext4_aio_mutex(inode));
133 ext4_aiodio_wait(inode);
134 }
135
136 ret = generic_file_aio_write(iocb, iov, nr_segs, pos);
137
138 if (unaligned_aio)
139 mutex_unlock(ext4_aio_mutex(inode));
140
141 return ret;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700142}
143
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +0400144static const struct vm_operations_struct ext4_file_vm_ops = {
Aneesh Kumar K.V2e9ee852008-07-11 19:27:31 -0400145 .fault = filemap_fault,
146 .page_mkwrite = ext4_page_mkwrite,
147};
148
149static int ext4_file_mmap(struct file *file, struct vm_area_struct *vma)
150{
151 struct address_space *mapping = file->f_mapping;
152
153 if (!mapping->a_ops->readpage)
154 return -ENOEXEC;
155 file_accessed(file);
156 vma->vm_ops = &ext4_file_vm_ops;
157 vma->vm_flags |= VM_CAN_NONLINEAR;
158 return 0;
159}
160
Theodore Ts'obc0b0d62009-06-13 10:09:48 -0400161static int ext4_file_open(struct inode * inode, struct file * filp)
162{
163 struct super_block *sb = inode->i_sb;
164 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500165 struct ext4_inode_info *ei = EXT4_I(inode);
Theodore Ts'obc0b0d62009-06-13 10:09:48 -0400166 struct vfsmount *mnt = filp->f_path.mnt;
167 struct path path;
168 char buf[64], *cp;
169
170 if (unlikely(!(sbi->s_mount_flags & EXT4_MF_MNTDIR_SAMPLED) &&
171 !(sb->s_flags & MS_RDONLY))) {
172 sbi->s_mount_flags |= EXT4_MF_MNTDIR_SAMPLED;
173 /*
174 * Sample where the filesystem has been mounted and
175 * store it in the superblock for sysadmin convenience
176 * when trying to sort through large numbers of block
177 * devices or filesystem images.
178 */
179 memset(buf, 0, sizeof(buf));
Al Viro38991672010-01-23 20:10:29 -0500180 path.mnt = mnt;
181 path.dentry = mnt->mnt_root;
Theodore Ts'obc0b0d62009-06-13 10:09:48 -0400182 cp = d_path(&path, buf, sizeof(buf));
Theodore Ts'obc0b0d62009-06-13 10:09:48 -0400183 if (!IS_ERR(cp)) {
184 memcpy(sbi->s_es->s_last_mounted, cp,
185 sizeof(sbi->s_es->s_last_mounted));
Theodore Ts'oa0375152010-06-11 23:14:04 -0400186 ext4_mark_super_dirty(sb);
Theodore Ts'obc0b0d62009-06-13 10:09:48 -0400187 }
188 }
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500189 /*
190 * Set up the jbd2_inode if we are opening the inode for
191 * writing and the journal is present
192 */
193 if (sbi->s_journal && !ei->jinode && (filp->f_mode & FMODE_WRITE)) {
194 struct jbd2_inode *jinode = jbd2_alloc_inode(GFP_KERNEL);
195
196 spin_lock(&inode->i_lock);
197 if (!ei->jinode) {
198 if (!jinode) {
199 spin_unlock(&inode->i_lock);
200 return -ENOMEM;
201 }
202 ei->jinode = jinode;
203 jbd2_journal_init_jbd_inode(ei->jinode, inode);
204 jinode = NULL;
205 }
206 spin_unlock(&inode->i_lock);
207 if (unlikely(jinode != NULL))
208 jbd2_free_inode(jinode);
209 }
Christoph Hellwig907f4552010-03-03 09:05:06 -0500210 return dquot_file_open(inode, filp);
Theodore Ts'obc0b0d62009-06-13 10:09:48 -0400211}
212
Toshiyuki Okajimae0d10bf2010-10-27 21:30:06 -0400213/*
214 * ext4_llseek() copied from generic_file_llseek() to handle both
215 * block-mapped and extent-mapped maxbytes values. This should
216 * otherwise be identical with generic_file_llseek().
217 */
218loff_t ext4_llseek(struct file *file, loff_t offset, int origin)
219{
220 struct inode *inode = file->f_mapping->host;
221 loff_t maxbytes;
222
223 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
224 maxbytes = EXT4_SB(inode->i_sb)->s_bitmap_maxbytes;
225 else
226 maxbytes = inode->i_sb->s_maxbytes;
227 mutex_lock(&inode->i_mutex);
228 switch (origin) {
229 case SEEK_END:
230 offset += inode->i_size;
231 break;
232 case SEEK_CUR:
233 if (offset == 0) {
234 mutex_unlock(&inode->i_mutex);
235 return file->f_pos;
236 }
237 offset += file->f_pos;
238 break;
Josef Bacikc334b112011-07-18 13:21:37 -0400239 case SEEK_DATA:
240 /*
241 * In the generic case the entire file is data, so as long as
242 * offset isn't at the end of the file then the offset is data.
243 */
244 if (offset >= inode->i_size) {
245 mutex_unlock(&inode->i_mutex);
246 return -ENXIO;
247 }
248 break;
249 case SEEK_HOLE:
250 /*
251 * There is a virtual hole at the end of the file, so as long as
252 * offset isn't i_size or larger, return i_size.
253 */
254 if (offset >= inode->i_size) {
255 mutex_unlock(&inode->i_mutex);
256 return -ENXIO;
257 }
258 offset = inode->i_size;
259 break;
Toshiyuki Okajimae0d10bf2010-10-27 21:30:06 -0400260 }
261
262 if (offset < 0 || offset > maxbytes) {
263 mutex_unlock(&inode->i_mutex);
264 return -EINVAL;
265 }
266
267 if (offset != file->f_pos) {
268 file->f_pos = offset;
269 file->f_version = 0;
270 }
271 mutex_unlock(&inode->i_mutex);
272
273 return offset;
274}
275
Mingming Cao617ba132006-10-11 01:20:53 -0700276const struct file_operations ext4_file_operations = {
Toshiyuki Okajimae0d10bf2010-10-27 21:30:06 -0400277 .llseek = ext4_llseek,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700278 .read = do_sync_read,
279 .write = do_sync_write,
280 .aio_read = generic_file_aio_read,
Mingming Cao617ba132006-10-11 01:20:53 -0700281 .aio_write = ext4_file_write,
Andi Kleen5cdd7b22008-04-29 22:03:54 -0400282 .unlocked_ioctl = ext4_ioctl,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700283#ifdef CONFIG_COMPAT
Mingming Cao617ba132006-10-11 01:20:53 -0700284 .compat_ioctl = ext4_compat_ioctl,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700285#endif
Aneesh Kumar K.V2e9ee852008-07-11 19:27:31 -0400286 .mmap = ext4_file_mmap,
Theodore Ts'obc0b0d62009-06-13 10:09:48 -0400287 .open = ext4_file_open,
Mingming Cao617ba132006-10-11 01:20:53 -0700288 .release = ext4_release_file,
289 .fsync = ext4_sync_file,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700290 .splice_read = generic_file_splice_read,
291 .splice_write = generic_file_splice_write,
Christoph Hellwig2fe17c12011-01-14 13:07:43 +0100292 .fallocate = ext4_fallocate,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700293};
294
Arjan van de Ven754661f2007-02-12 00:55:38 -0800295const struct inode_operations ext4_file_inode_operations = {
Mingming Cao617ba132006-10-11 01:20:53 -0700296 .setattr = ext4_setattr,
Mingming Cao3e3398a2008-07-11 19:27:31 -0400297 .getattr = ext4_getattr,
Theodore Ts'o03010a32008-10-10 20:02:48 -0400298#ifdef CONFIG_EXT4_FS_XATTR
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700299 .setxattr = generic_setxattr,
300 .getxattr = generic_getxattr,
Mingming Cao617ba132006-10-11 01:20:53 -0700301 .listxattr = ext4_listxattr,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700302 .removexattr = generic_removexattr,
303#endif
Christoph Hellwig4e34e712011-07-23 17:37:31 +0200304 .get_acl = ext4_get_acl,
Eric Sandeen6873fa02008-10-07 00:46:36 -0400305 .fiemap = ext4_fiemap,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700306};
307