]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/reiserfs/file.c
reiserfs: rename p_s_inode to inode
[net-next-2.6.git] / fs / reiserfs / file.c
CommitLineData
1da177e4
LT
1/*
2 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
3 */
4
1da177e4
LT
5#include <linux/time.h>
6#include <linux/reiserfs_fs.h>
7#include <linux/reiserfs_acl.h>
8#include <linux/reiserfs_xattr.h>
1da177e4
LT
9#include <asm/uaccess.h>
10#include <linux/pagemap.h>
11#include <linux/swap.h>
12#include <linux/writeback.h>
13#include <linux/blkdev.h>
14#include <linux/buffer_head.h>
15#include <linux/quotaops.h>
16
17/*
18** We pack the tails of files on file close, not at the time they are written.
19** This implies an unnecessary copy of the tail and an unnecessary indirect item
20** insertion/balancing, for files that are written in one write.
21** It avoids unnecessary tail packings (balances) for files that are written in
22** multiple writes and are small enough to have tails.
0222e657 23**
1da177e4
LT
24** file_release is called by the VFS layer when the file is closed. If
25** this is the last open file descriptor, and the file
26** small enough to have a tail, and the tail is currently in an
27** unformatted node, the tail is converted back into a direct item.
0222e657 28**
1da177e4 29** We use reiserfs_truncate_file to pack the tail, since it already has
0222e657 30** all the conditions coded.
1da177e4 31*/
bd4c625c 32static int reiserfs_file_release(struct inode *inode, struct file *filp)
1da177e4
LT
33{
34
bd4c625c
LT
35 struct reiserfs_transaction_handle th;
36 int err;
37 int jbegin_failure = 0;
1da177e4 38
14a61442 39 BUG_ON(!S_ISREG(inode->i_mode));
1da177e4 40
bd4c625c
LT
41 /* fast out for when nothing needs to be done */
42 if ((atomic_read(&inode->i_count) > 1 ||
43 !(REISERFS_I(inode)->i_flags & i_pack_on_close_mask) ||
44 !tail_has_to_be_packed(inode)) &&
45 REISERFS_I(inode)->i_prealloc_count <= 0) {
46 return 0;
47 }
1da177e4 48
1b1dcc1b 49 mutex_lock(&inode->i_mutex);
de14569f
VS
50
51 mutex_lock(&(REISERFS_I(inode)->i_mmap));
52 if (REISERFS_I(inode)->i_flags & i_ever_mapped)
53 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
54
b5f3953c 55 reiserfs_write_lock(inode->i_sb);
bd4c625c
LT
56 /* freeing preallocation only involves relogging blocks that
57 * are already in the current transaction. preallocation gets
58 * freed at the end of each transaction, so it is impossible for
59 * us to log any additional blocks (including quota blocks)
60 */
61 err = journal_begin(&th, inode->i_sb, 1);
1da177e4 62 if (err) {
bd4c625c
LT
63 /* uh oh, we can't allow the inode to go away while there
64 * is still preallocation blocks pending. Try to join the
65 * aborted transaction
66 */
67 jbegin_failure = err;
68 err = journal_join_abort(&th, inode->i_sb, 1);
69
70 if (err) {
71 /* hmpf, our choices here aren't good. We can pin the inode
72 * which will disallow unmount from every happening, we can
73 * do nothing, which will corrupt random memory on unmount,
74 * or we can forcibly remove the file from the preallocation
75 * list, which will leak blocks on disk. Lets pin the inode
76 * and let the admin know what is going on.
77 */
78 igrab(inode);
45b03d5e 79 reiserfs_warning(inode->i_sb, "clm-9001",
bd4c625c 80 "pinning inode %lu because the "
533221fb
AD
81 "preallocation can't be freed",
82 inode->i_ino);
bd4c625c
LT
83 goto out;
84 }
1da177e4 85 }
bd4c625c 86 reiserfs_update_inode_transaction(inode);
1da177e4
LT
87
88#ifdef REISERFS_PREALLOCATE
bd4c625c 89 reiserfs_discard_prealloc(&th, inode);
1da177e4 90#endif
bd4c625c
LT
91 err = journal_end(&th, inode->i_sb, 1);
92
93 /* copy back the error code from journal_begin */
94 if (!err)
95 err = jbegin_failure;
96
97 if (!err && atomic_read(&inode->i_count) <= 1 &&
98 (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) &&
99 tail_has_to_be_packed(inode)) {
100 /* if regular file is released by last holder and it has been
101 appended (we append by unformatted node only) or its direct
102 item(s) had to be converted, then it may have to be
103 indirect2direct converted */
104 err = reiserfs_truncate_file(inode, 0);
105 }
106 out:
de14569f 107 mutex_unlock(&(REISERFS_I(inode)->i_mmap));
1b1dcc1b 108 mutex_unlock(&inode->i_mutex);
bd4c625c
LT
109 reiserfs_write_unlock(inode->i_sb);
110 return err;
1da177e4
LT
111}
112
de14569f
VS
113static int reiserfs_file_mmap(struct file *file, struct vm_area_struct *vma)
114{
115 struct inode *inode;
116
117 inode = file->f_path.dentry->d_inode;
118 mutex_lock(&(REISERFS_I(inode)->i_mmap));
119 REISERFS_I(inode)->i_flags |= i_ever_mapped;
120 mutex_unlock(&(REISERFS_I(inode)->i_mmap));
121
122 return generic_file_mmap(file, vma);
123}
124
bd4c625c
LT
125static void reiserfs_vfs_truncate_file(struct inode *inode)
126{
127 reiserfs_truncate_file(inode, 1);
1da177e4
LT
128}
129
130/* Sync a reiserfs file. */
131
132/*
133 * FIXME: sync_mapping_buffers() never has anything to sync. Can
134 * be removed...
135 */
136
bd4c625c
LT
137static int reiserfs_sync_file(struct file *p_s_filp,
138 struct dentry *p_s_dentry, int datasync)
139{
995c762e 140 struct inode *inode = p_s_dentry->d_inode;
bd4c625c
LT
141 int n_err;
142 int barrier_done;
143
995c762e
JM
144 BUG_ON(!S_ISREG(inode->i_mode));
145 n_err = sync_mapping_buffers(inode->i_mapping);
146 reiserfs_write_lock(inode->i_sb);
147 barrier_done = reiserfs_commit_for_inode(inode);
148 reiserfs_write_unlock(inode->i_sb);
149 if (barrier_done != 1 && reiserfs_barrier_flush(inode->i_sb))
150 blkdev_issue_flush(inode->i_sb->s_bdev, NULL);
bd4c625c
LT
151 if (barrier_done < 0)
152 return barrier_done;
153 return (n_err < 0) ? -EIO : 0;
1da177e4
LT
154}
155
1da177e4
LT
156/* taken fs/buffer.c:__block_commit_write */
157int reiserfs_commit_page(struct inode *inode, struct page *page,
bd4c625c 158 unsigned from, unsigned to)
1da177e4 159{
bd4c625c
LT
160 unsigned block_start, block_end;
161 int partial = 0;
162 unsigned blocksize;
163 struct buffer_head *bh, *head;
164 unsigned long i_size_index = inode->i_size >> PAGE_CACHE_SHIFT;
165 int new;
166 int logit = reiserfs_file_data_log(inode);
167 struct super_block *s = inode->i_sb;
168 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
169 struct reiserfs_transaction_handle th;
170 int ret = 0;
171
172 th.t_trans_id = 0;
173 blocksize = 1 << inode->i_blkbits;
174
175 if (logit) {
176 reiserfs_write_lock(s);
177 ret = journal_begin(&th, s, bh_per_page + 1);
178 if (ret)
179 goto drop_write_lock;
180 reiserfs_update_inode_transaction(inode);
181 }
182 for (bh = head = page_buffers(page), block_start = 0;
183 bh != head || !block_start;
184 block_start = block_end, bh = bh->b_this_page) {
185
186 new = buffer_new(bh);
187 clear_buffer_new(bh);
188 block_end = block_start + blocksize;
189 if (block_end <= from || block_start >= to) {
190 if (!buffer_uptodate(bh))
191 partial = 1;
192 } else {
193 set_buffer_uptodate(bh);
194 if (logit) {
195 reiserfs_prepare_for_journal(s, bh, 1);
196 journal_mark_dirty(&th, s, bh);
197 } else if (!buffer_dirty(bh)) {
198 mark_buffer_dirty(bh);
199 /* do data=ordered on any page past the end
200 * of file and any buffer marked BH_New.
201 */
202 if (reiserfs_data_ordered(inode->i_sb) &&
203 (new || page->index >= i_size_index)) {
204 reiserfs_add_ordered_list(inode, bh);
205 }
206 }
207 }
1da177e4 208 }
bd4c625c
LT
209 if (logit) {
210 ret = journal_end(&th, s, bh_per_page + 1);
211 drop_write_lock:
212 reiserfs_write_unlock(s);
213 }
214 /*
215 * If this is a partial write which happened to make all buffers
216 * uptodate then we can optimize away a bogus readpage() for
217 * the next read(). Here we 'discover' whether the page went
218 * uptodate as a result of this (potentially partial) write.
219 */
220 if (!partial)
221 SetPageUptodate(page);
222 return ret;
1da177e4
LT
223}
224
1da177e4 225/* Write @count bytes at position @ppos in a file indicated by @file
0222e657 226 from the buffer @buf.
1da177e4
LT
227
228 generic_file_write() is only appropriate for filesystems that are not seeking to optimize performance and want
229 something simple that works. It is not for serious use by general purpose filesystems, excepting the one that it was
230 written for (ext2/3). This is for several reasons:
231
232 * It has no understanding of any filesystem specific optimizations.
233
234 * It enters the filesystem repeatedly for each page that is written.
235
236 * It depends on reiserfs_get_block() function which if implemented by reiserfs performs costly search_by_key
237 * operation for each page it is supplied with. By contrast reiserfs_file_write() feeds as much as possible at a time
238 * to reiserfs which allows for fewer tree traversals.
239
240 * Each indirect pointer insertion takes a lot of cpu, because it involves memory moves inside of blocks.
241
242 * Asking the block allocation code for blocks one at a time is slightly less efficient.
243
244 All of these reasons for not using only generic file write were understood back when reiserfs was first miscoded to
245 use it, but we were in a hurry to make code freeze, and so it couldn't be revised then. This new code should make
246 things right finally.
247
248 Future Features: providing search_by_key with hints.
249
250*/
bd4c625c
LT
251static ssize_t reiserfs_file_write(struct file *file, /* the file we are going to write into */
252 const char __user * buf, /* pointer to user supplied data
253 (in userspace) */
254 size_t count, /* amount of bytes to write */
255 loff_t * ppos /* pointer to position in file that we start writing at. Should be updated to
256 * new current position before returning. */
257 )
1da177e4 258{
1fc5adbd 259 struct inode *inode = file->f_path.dentry->d_inode; // Inode of the file that we are writing to.
bd4c625c
LT
260 /* To simplify coding at this time, we store
261 locked pages in array for now */
bd4c625c
LT
262 struct reiserfs_transaction_handle th;
263 th.t_trans_id = 0;
264
fa385bef
JM
265 /* If a filesystem is converted from 3.5 to 3.6, we'll have v3.5 items
266 * lying around (most of the disk, in fact). Despite the filesystem
267 * now being a v3.6 format, the old items still can't support large
268 * file sizes. Catch this case here, as the rest of the VFS layer is
269 * oblivious to the different limitations between old and new items.
270 * reiserfs_setattr catches this for truncates. This chunk is lifted
271 * from generic_write_checks. */
272 if (get_inode_item_key_version (inode) == KEY_FORMAT_3_5 &&
273 *ppos + count > MAX_NON_LFS) {
274 if (*ppos >= MAX_NON_LFS) {
fa385bef
JM
275 return -EFBIG;
276 }
277 if (count > MAX_NON_LFS - (unsigned long)*ppos)
278 count = MAX_NON_LFS - (unsigned long)*ppos;
279 }
280
797b4cff 281 return do_sync_write(file, buf, count, ppos);
1da177e4
LT
282}
283
4b6f5d20 284const struct file_operations reiserfs_file_operations = {
027445c3 285 .read = do_sync_read,
bd4c625c
LT
286 .write = reiserfs_file_write,
287 .ioctl = reiserfs_ioctl,
52b499c4
DH
288#ifdef CONFIG_COMPAT
289 .compat_ioctl = reiserfs_compat_ioctl,
290#endif
de14569f 291 .mmap = reiserfs_file_mmap,
5a2618e6 292 .open = generic_file_open,
bd4c625c
LT
293 .release = reiserfs_file_release,
294 .fsync = reiserfs_sync_file,
bd4c625c 295 .aio_read = generic_file_aio_read,
9637f28f 296 .aio_write = generic_file_aio_write,
5274f052
JA
297 .splice_read = generic_file_splice_read,
298 .splice_write = generic_file_splice_write,
91efc167 299 .llseek = generic_file_llseek,
1da177e4
LT
300};
301
c5ef1c42 302const struct inode_operations reiserfs_file_inode_operations = {
bd4c625c
LT
303 .truncate = reiserfs_vfs_truncate_file,
304 .setattr = reiserfs_setattr,
305 .setxattr = reiserfs_setxattr,
306 .getxattr = reiserfs_getxattr,
307 .listxattr = reiserfs_listxattr,
308 .removexattr = reiserfs_removexattr,
309 .permission = reiserfs_permission,
1da177e4 310};