blob: 8375c922c0d56d3467c3c654ef84306531299032 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
3 */
4
Linus Torvalds1da177e2005-04-16 15:20:36 -07005#include <linux/time.h>
Al Virof466c6f2012-03-17 01:16:43 -04006#include "reiserfs.h"
Al Viroa3063ab2012-03-17 01:03:10 -04007#include "acl.h"
Al Viroc45ac882012-03-17 00:59:06 -04008#include "xattr.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#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.
Jeff Mahoney0222e652009-03-30 14:02:44 -040023**
Linus Torvalds1da177e2005-04-16 15:20:36 -070024** 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.
Jeff Mahoney0222e652009-03-30 14:02:44 -040028**
Linus Torvalds1da177e2005-04-16 15:20:36 -070029** We use reiserfs_truncate_file to pack the tail, since it already has
Jeff Mahoney0222e652009-03-30 14:02:44 -040030** all the conditions coded.
Linus Torvalds1da177e2005-04-16 15:20:36 -070031*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070032static int reiserfs_file_release(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -070033{
34
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070035 struct reiserfs_transaction_handle th;
36 int err;
37 int jbegin_failure = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
Eric Sesterhenn14a61442006-10-03 23:36:38 +020039 BUG_ON(!S_ISREG(inode->i_mode));
Linus Torvalds1da177e2005-04-16 15:20:36 -070040
Al Viro0e4f6a72010-07-04 12:18:57 +040041 if (atomic_add_unless(&REISERFS_I(inode)->openers, -1, 1))
42 return 0;
43
44 mutex_lock(&(REISERFS_I(inode)->tailpack));
45
46 if (!atomic_dec_and_test(&REISERFS_I(inode)->openers)) {
47 mutex_unlock(&(REISERFS_I(inode)->tailpack));
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070048 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070049 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070050
Al Viro0e4f6a72010-07-04 12:18:57 +040051 /* fast out for when nothing needs to be done */
52 if ((!(REISERFS_I(inode)->i_flags & i_pack_on_close_mask) ||
53 !tail_has_to_be_packed(inode)) &&
54 REISERFS_I(inode)->i_prealloc_count <= 0) {
55 mutex_unlock(&(REISERFS_I(inode)->tailpack));
56 return 0;
57 }
Vladimir Savelievde145692007-01-22 20:40:46 -080058
Chris Masonb5f39532006-08-05 12:15:08 -070059 reiserfs_write_lock(inode->i_sb);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070060 /* freeing preallocation only involves relogging blocks that
61 * are already in the current transaction. preallocation gets
62 * freed at the end of each transaction, so it is impossible for
63 * us to log any additional blocks (including quota blocks)
64 */
65 err = journal_begin(&th, inode->i_sb, 1);
66 if (err) {
67 /* uh oh, we can't allow the inode to go away while there
68 * is still preallocation blocks pending. Try to join the
69 * aborted transaction
70 */
71 jbegin_failure = err;
72 err = journal_join_abort(&th, inode->i_sb, 1);
73
74 if (err) {
75 /* hmpf, our choices here aren't good. We can pin the inode
76 * which will disallow unmount from every happening, we can
77 * do nothing, which will corrupt random memory on unmount,
78 * or we can forcibly remove the file from the preallocation
79 * list, which will leak blocks on disk. Lets pin the inode
80 * and let the admin know what is going on.
81 */
82 igrab(inode);
Jeff Mahoney45b03d52009-03-30 14:02:21 -040083 reiserfs_warning(inode->i_sb, "clm-9001",
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070084 "pinning inode %lu because the "
Alexey Dobriyan533221f2006-11-25 11:09:30 -080085 "preallocation can't be freed",
86 inode->i_ino);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070087 goto out;
88 }
89 }
90 reiserfs_update_inode_transaction(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -070091
92#ifdef REISERFS_PREALLOCATE
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070093 reiserfs_discard_prealloc(&th, inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -070094#endif
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070095 err = journal_end(&th, inode->i_sb, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -070096
Linus Torvaldsbd4c6252005-07-12 20:21:28 -070097 /* copy back the error code from journal_begin */
98 if (!err)
99 err = jbegin_failure;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
Al Viro0e4f6a72010-07-04 12:18:57 +0400101 if (!err &&
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700102 (REISERFS_I(inode)->i_flags & i_pack_on_close_mask) &&
103 tail_has_to_be_packed(inode)) {
Al Viro0e4f6a72010-07-04 12:18:57 +0400104
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700105 /* if regular file is released by last holder and it has been
106 appended (we append by unformatted node only) or its direct
107 item(s) had to be converted, then it may have to be
108 indirect2direct converted */
109 err = reiserfs_truncate_file(inode, 0);
110 }
111 out:
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700112 reiserfs_write_unlock(inode->i_sb);
Al Viro0e4f6a72010-07-04 12:18:57 +0400113 mutex_unlock(&(REISERFS_I(inode)->tailpack));
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700114 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115}
116
Al Viro0e4f6a72010-07-04 12:18:57 +0400117static int reiserfs_file_open(struct inode *inode, struct file *file)
Vladimir Savelievde145692007-01-22 20:40:46 -0800118{
Al Viro0e4f6a72010-07-04 12:18:57 +0400119 int err = dquot_file_open(inode, file);
120 if (!atomic_inc_not_zero(&REISERFS_I(inode)->openers)) {
121 /* somebody might be tailpacking on final close; wait for it */
122 mutex_lock(&(REISERFS_I(inode)->tailpack));
123 atomic_inc(&REISERFS_I(inode)->openers);
124 mutex_unlock(&(REISERFS_I(inode)->tailpack));
125 }
126 return err;
Vladimir Savelievde145692007-01-22 20:40:46 -0800127}
128
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700129static void reiserfs_vfs_truncate_file(struct inode *inode)
130{
Al Viro0e4f6a72010-07-04 12:18:57 +0400131 mutex_lock(&(REISERFS_I(inode)->tailpack));
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700132 reiserfs_truncate_file(inode, 1);
Al Viro0e4f6a72010-07-04 12:18:57 +0400133 mutex_unlock(&(REISERFS_I(inode)->tailpack));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134}
135
136/* Sync a reiserfs file. */
137
138/*
139 * FIXME: sync_mapping_buffers() never has anything to sync. Can
140 * be removed...
141 */
142
Josef Bacik02c24a82011-07-16 20:44:56 -0400143static int reiserfs_sync_file(struct file *filp, loff_t start, loff_t end,
144 int datasync)
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700145{
Christoph Hellwig7ea80852010-05-26 17:53:25 +0200146 struct inode *inode = filp->f_mapping->host;
Jeff Mahoneyee939612009-03-30 14:02:50 -0400147 int err;
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700148 int barrier_done;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149
Josef Bacik02c24a82011-07-16 20:44:56 -0400150 err = filemap_write_and_wait_range(inode->i_mapping, start, end);
151 if (err)
152 return err;
153
154 mutex_lock(&inode->i_mutex);
Jeff Mahoney995c7622009-03-30 14:02:47 -0400155 BUG_ON(!S_ISREG(inode->i_mode));
Jeff Mahoneyee939612009-03-30 14:02:50 -0400156 err = sync_mapping_buffers(inode->i_mapping);
Jeff Mahoney995c7622009-03-30 14:02:47 -0400157 reiserfs_write_lock(inode->i_sb);
158 barrier_done = reiserfs_commit_for_inode(inode);
159 reiserfs_write_unlock(inode->i_sb);
160 if (barrier_done != 1 && reiserfs_barrier_flush(inode->i_sb))
Christoph Hellwigdd3932e2010-09-16 20:51:46 +0200161 blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
Josef Bacik02c24a82011-07-16 20:44:56 -0400162 mutex_unlock(&inode->i_mutex);
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700163 if (barrier_done < 0)
164 return barrier_done;
Jeff Mahoneyee939612009-03-30 14:02:50 -0400165 return (err < 0) ? -EIO : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166}
167
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168/* taken fs/buffer.c:__block_commit_write */
169int reiserfs_commit_page(struct inode *inode, struct page *page,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700170 unsigned from, unsigned to)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171{
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700172 unsigned block_start, block_end;
173 int partial = 0;
174 unsigned blocksize;
175 struct buffer_head *bh, *head;
176 unsigned long i_size_index = inode->i_size >> PAGE_CACHE_SHIFT;
177 int new;
178 int logit = reiserfs_file_data_log(inode);
179 struct super_block *s = inode->i_sb;
180 int bh_per_page = PAGE_CACHE_SIZE / s->s_blocksize;
181 struct reiserfs_transaction_handle th;
182 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700184 th.t_trans_id = 0;
185 blocksize = 1 << inode->i_blkbits;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700187 if (logit) {
188 reiserfs_write_lock(s);
189 ret = journal_begin(&th, s, bh_per_page + 1);
190 if (ret)
191 goto drop_write_lock;
192 reiserfs_update_inode_transaction(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 }
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700194 for (bh = head = page_buffers(page), block_start = 0;
195 bh != head || !block_start;
196 block_start = block_end, bh = bh->b_this_page) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700198 new = buffer_new(bh);
199 clear_buffer_new(bh);
200 block_end = block_start + blocksize;
201 if (block_end <= from || block_start >= to) {
202 if (!buffer_uptodate(bh))
203 partial = 1;
204 } else {
205 set_buffer_uptodate(bh);
206 if (logit) {
207 reiserfs_prepare_for_journal(s, bh, 1);
208 journal_mark_dirty(&th, s, bh);
209 } else if (!buffer_dirty(bh)) {
210 mark_buffer_dirty(bh);
211 /* do data=ordered on any page past the end
212 * of file and any buffer marked BH_New.
213 */
214 if (reiserfs_data_ordered(inode->i_sb) &&
215 (new || page->index >= i_size_index)) {
216 reiserfs_add_ordered_list(inode, bh);
217 }
218 }
219 }
220 }
221 if (logit) {
222 ret = journal_end(&th, s, bh_per_page + 1);
223 drop_write_lock:
224 reiserfs_write_unlock(s);
225 }
226 /*
227 * If this is a partial write which happened to make all buffers
228 * uptodate then we can optimize away a bogus readpage() for
229 * the next read(). Here we 'discover' whether the page went
230 * uptodate as a result of this (potentially partial) write.
231 */
232 if (!partial)
233 SetPageUptodate(page);
234 return ret;
235}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237/* Write @count bytes at position @ppos in a file indicated by @file
Jeff Mahoney0222e652009-03-30 14:02:44 -0400238 from the buffer @buf.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
240 generic_file_write() is only appropriate for filesystems that are not seeking to optimize performance and want
241 something simple that works. It is not for serious use by general purpose filesystems, excepting the one that it was
242 written for (ext2/3). This is for several reasons:
243
244 * It has no understanding of any filesystem specific optimizations.
245
246 * It enters the filesystem repeatedly for each page that is written.
247
248 * It depends on reiserfs_get_block() function which if implemented by reiserfs performs costly search_by_key
249 * operation for each page it is supplied with. By contrast reiserfs_file_write() feeds as much as possible at a time
250 * to reiserfs which allows for fewer tree traversals.
251
252 * Each indirect pointer insertion takes a lot of cpu, because it involves memory moves inside of blocks.
253
254 * Asking the block allocation code for blocks one at a time is slightly less efficient.
255
256 All of these reasons for not using only generic file write were understood back when reiserfs was first miscoded to
257 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
258 things right finally.
259
260 Future Features: providing search_by_key with hints.
261
262*/
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700263static ssize_t reiserfs_file_write(struct file *file, /* the file we are going to write into */
264 const char __user * buf, /* pointer to user supplied data
265 (in userspace) */
266 size_t count, /* amount of bytes to write */
267 loff_t * ppos /* pointer to position in file that we start writing at. Should be updated to
268 * new current position before returning. */
269 )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270{
Josef Sipek1fc5adb2006-12-08 02:37:33 -0800271 struct inode *inode = file->f_path.dentry->d_inode; // Inode of the file that we are writing to.
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700272 /* To simplify coding at this time, we store
273 locked pages in array for now */
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700274 struct reiserfs_transaction_handle th;
275 th.t_trans_id = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276
Jeff Mahoneyfa385be2006-02-01 03:06:51 -0800277 /* If a filesystem is converted from 3.5 to 3.6, we'll have v3.5 items
278 * lying around (most of the disk, in fact). Despite the filesystem
279 * now being a v3.6 format, the old items still can't support large
280 * file sizes. Catch this case here, as the rest of the VFS layer is
281 * oblivious to the different limitations between old and new items.
282 * reiserfs_setattr catches this for truncates. This chunk is lifted
283 * from generic_write_checks. */
284 if (get_inode_item_key_version (inode) == KEY_FORMAT_3_5 &&
285 *ppos + count > MAX_NON_LFS) {
286 if (*ppos >= MAX_NON_LFS) {
Jeff Mahoneyfa385be2006-02-01 03:06:51 -0800287 return -EFBIG;
288 }
289 if (count > MAX_NON_LFS - (unsigned long)*ppos)
290 count = MAX_NON_LFS - (unsigned long)*ppos;
291 }
292
Vladimir Saveliev797b4cf2007-10-16 01:25:12 -0700293 return do_sync_write(file, buf, count, ppos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294}
295
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800296const struct file_operations reiserfs_file_operations = {
Badari Pulavarty027445c2006-09-30 23:28:46 -0700297 .read = do_sync_read,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700298 .write = reiserfs_file_write,
Frederic Weisbecker205cb372009-10-14 23:22:17 +0200299 .unlocked_ioctl = reiserfs_ioctl,
David Howells52b499c2006-08-29 19:06:18 +0100300#ifdef CONFIG_COMPAT
301 .compat_ioctl = reiserfs_compat_ioctl,
302#endif
Al Viro0e4f6a72010-07-04 12:18:57 +0400303 .mmap = generic_file_mmap,
304 .open = reiserfs_file_open,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700305 .release = reiserfs_file_release,
306 .fsync = reiserfs_sync_file,
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700307 .aio_read = generic_file_aio_read,
Alexey Dobriyan9637f282006-06-26 00:24:57 -0700308 .aio_write = generic_file_aio_write,
Jens Axboe5274f052006-03-30 15:15:30 +0200309 .splice_read = generic_file_splice_read,
310 .splice_write = generic_file_splice_write,
Christoph Hellwig91efc162008-09-08 19:42:50 +0200311 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312};
313
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800314const struct inode_operations reiserfs_file_inode_operations = {
Linus Torvaldsbd4c6252005-07-12 20:21:28 -0700315 .truncate = reiserfs_vfs_truncate_file,
316 .setattr = reiserfs_setattr,
317 .setxattr = reiserfs_setxattr,
318 .getxattr = reiserfs_getxattr,
319 .listxattr = reiserfs_listxattr,
320 .removexattr = reiserfs_removexattr,
321 .permission = reiserfs_permission,
Christoph Hellwig4e34e712011-07-23 17:37:31 +0200322 .get_acl = reiserfs_get_acl,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323};