2 * Copyright (C) International Business Machines Corp., 2000-2004
3 * Portions Copyright (C) Christoph Hellwig, 2001-2002
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
13 * the GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
21 #include <linux/mpage.h>
22 #include <linux/buffer_head.h>
23 #include <linux/pagemap.h>
24 #include <linux/quotaops.h>
25 #include <linux/writeback.h>
26 #include <linux/aio.h>
27 #include "jfs_incore.h"
28 #include "jfs_inode.h"
29 #include "jfs_filsys.h"
31 #include "jfs_extent.h"
32 #include "jfs_unicode.h"
33 #include "jfs_debug.h"
36 struct inode
*jfs_iget(struct super_block
*sb
, unsigned long ino
)
41 inode
= iget_locked(sb
, ino
);
43 return ERR_PTR(-ENOMEM
);
44 if (!(inode
->i_state
& I_NEW
))
53 if (S_ISREG(inode
->i_mode
)) {
54 inode
->i_op
= &jfs_file_inode_operations
;
55 inode
->i_fop
= &jfs_file_operations
;
56 inode
->i_mapping
->a_ops
= &jfs_aops
;
57 } else if (S_ISDIR(inode
->i_mode
)) {
58 inode
->i_op
= &jfs_dir_inode_operations
;
59 inode
->i_fop
= &jfs_dir_operations
;
60 } else if (S_ISLNK(inode
->i_mode
)) {
61 if (inode
->i_size
>= IDATASIZE
) {
62 inode
->i_op
= &page_symlink_inode_operations
;
63 inode
->i_mapping
->a_ops
= &jfs_aops
;
65 inode
->i_op
= &jfs_fast_symlink_inode_operations
;
67 * The inline data should be null-terminated, but
68 * don't let on-disk corruption crash the kernel
70 JFS_IP(inode
)->i_inline
[inode
->i_size
] = '\0';
73 inode
->i_op
= &jfs_file_inode_operations
;
74 init_special_inode(inode
, inode
->i_mode
, inode
->i_rdev
);
76 unlock_new_inode(inode
);
81 * Workhorse of both fsync & write_inode
83 int jfs_commit_inode(struct inode
*inode
, int wait
)
89 jfs_info("In jfs_commit_inode, inode = 0x%p", inode
);
92 * Don't commit if inode has been committed since last being
93 * marked dirty, or if it has been deleted.
95 if (inode
->i_nlink
== 0 || !test_cflag(COMMIT_Dirty
, inode
))
98 if (isReadOnly(inode
)) {
99 /* kernel allows writes to devices on read-only
100 * partitions and may think inode is dirty
102 if (!special_file(inode
->i_mode
) && noisy
) {
103 jfs_err("jfs_commit_inode(0x%p) called on "
104 "read-only volume", inode
);
105 jfs_err("Is remount racy?");
111 tid
= txBegin(inode
->i_sb
, COMMIT_INODE
);
112 mutex_lock(&JFS_IP(inode
)->commit_mutex
);
115 * Retest inode state after taking commit_mutex
117 if (inode
->i_nlink
&& test_cflag(COMMIT_Dirty
, inode
))
118 rc
= txCommit(tid
, 1, &inode
, wait
? COMMIT_SYNC
: 0);
121 mutex_unlock(&JFS_IP(inode
)->commit_mutex
);
125 int jfs_write_inode(struct inode
*inode
, struct writeback_control
*wbc
)
127 int wait
= wbc
->sync_mode
== WB_SYNC_ALL
;
129 if (inode
->i_nlink
== 0)
132 * If COMMIT_DIRTY is not set, the inode isn't really dirty.
133 * It has been committed since the last change, but was still
134 * on the dirty inode list.
136 if (!test_cflag(COMMIT_Dirty
, inode
)) {
137 /* Make sure committed changes hit the disk */
138 jfs_flush_journal(JFS_SBI(inode
->i_sb
)->log
, wait
);
142 if (jfs_commit_inode(inode
, wait
)) {
143 jfs_err("jfs_write_inode: jfs_commit_inode failed!");
149 void jfs_evict_inode(struct inode
*inode
)
151 jfs_info("In jfs_evict_inode, inode = 0x%p", inode
);
153 if (!inode
->i_nlink
&& !is_bad_inode(inode
)) {
154 dquot_initialize(inode
);
156 if (JFS_IP(inode
)->fileset
== FILESYSTEM_I
) {
157 truncate_inode_pages(&inode
->i_data
, 0);
159 if (test_cflag(COMMIT_Freewmap
, inode
))
160 jfs_free_zero_link(inode
);
165 * Free the inode from the quota allocation.
167 dquot_initialize(inode
);
168 dquot_free_inode(inode
);
171 truncate_inode_pages(&inode
->i_data
, 0);
177 void jfs_dirty_inode(struct inode
*inode
, int flags
)
179 static int noisy
= 5;
181 if (isReadOnly(inode
)) {
182 if (!special_file(inode
->i_mode
) && noisy
) {
183 /* kernel allows writes to devices on read-only
184 * partitions and may try to mark inode dirty
186 jfs_err("jfs_dirty_inode called on read-only volume");
187 jfs_err("Is remount racy?");
193 set_cflag(COMMIT_Dirty
, inode
);
196 int jfs_get_block(struct inode
*ip
, sector_t lblock
,
197 struct buffer_head
*bh_result
, int create
)
199 s64 lblock64
= lblock
;
204 s32 xlen
= bh_result
->b_size
>> ip
->i_blkbits
;
207 * Take appropriate lock on inode
210 IWRITE_LOCK(ip
, RDWRLOCK_NORMAL
);
212 IREAD_LOCK(ip
, RDWRLOCK_NORMAL
);
214 if (((lblock64
<< ip
->i_sb
->s_blocksize_bits
) < ip
->i_size
) &&
215 (!xtLookup(ip
, lblock64
, xlen
, &xflag
, &xaddr
, &xlen
, 0)) &&
217 if (xflag
& XAD_NOTRECORDED
) {
220 * Allocated but not recorded, read treats
225 XADoffset(&xad
, lblock64
);
226 XADlength(&xad
, xlen
);
227 XADaddress(&xad
, xaddr
);
230 * As long as block size = 4K, this isn't a problem.
231 * We should mark the whole page not ABNR, but how
232 * will we know to mark the other blocks BH_New?
236 rc
= extRecord(ip
, &xad
);
239 set_buffer_new(bh_result
);
242 map_bh(bh_result
, ip
->i_sb
, xaddr
);
243 bh_result
->b_size
= xlen
<< ip
->i_blkbits
;
250 * Allocate a new block
253 if ((rc
= extHint(ip
, lblock64
<< ip
->i_sb
->s_blocksize_bits
, &xad
)))
255 rc
= extAlloc(ip
, xlen
, lblock64
, &xad
, false);
259 set_buffer_new(bh_result
);
260 map_bh(bh_result
, ip
->i_sb
, addressXAD(&xad
));
261 bh_result
->b_size
= lengthXAD(&xad
) << ip
->i_blkbits
;
265 * We need to do whatever it takes to keep all but the last buffers
266 * in 4K pages - see jfs_write.c
273 * Release lock on inode
282 static int jfs_writepage(struct page
*page
, struct writeback_control
*wbc
)
284 return block_write_full_page(page
, jfs_get_block
, wbc
);
287 static int jfs_writepages(struct address_space
*mapping
,
288 struct writeback_control
*wbc
)
290 return mpage_writepages(mapping
, wbc
, jfs_get_block
);
293 static int jfs_readpage(struct file
*file
, struct page
*page
)
295 return mpage_readpage(page
, jfs_get_block
);
298 static int jfs_readpages(struct file
*file
, struct address_space
*mapping
,
299 struct list_head
*pages
, unsigned nr_pages
)
301 return mpage_readpages(mapping
, pages
, nr_pages
, jfs_get_block
);
304 static void jfs_write_failed(struct address_space
*mapping
, loff_t to
)
306 struct inode
*inode
= mapping
->host
;
308 if (to
> inode
->i_size
) {
309 truncate_pagecache(inode
, to
, inode
->i_size
);
314 static int jfs_write_begin(struct file
*file
, struct address_space
*mapping
,
315 loff_t pos
, unsigned len
, unsigned flags
,
316 struct page
**pagep
, void **fsdata
)
320 ret
= nobh_write_begin(mapping
, pos
, len
, flags
, pagep
, fsdata
,
323 jfs_write_failed(mapping
, pos
+ len
);
328 static sector_t
jfs_bmap(struct address_space
*mapping
, sector_t block
)
330 return generic_block_bmap(mapping
, block
, jfs_get_block
);
333 static ssize_t
jfs_direct_IO(int rw
, struct kiocb
*iocb
,
334 const struct iovec
*iov
, loff_t offset
, unsigned long nr_segs
)
336 struct file
*file
= iocb
->ki_filp
;
337 struct address_space
*mapping
= file
->f_mapping
;
338 struct inode
*inode
= file
->f_mapping
->host
;
341 ret
= blockdev_direct_IO(rw
, iocb
, inode
, iov
, offset
, nr_segs
,
345 * In case of error extending write may have instantiated a few
346 * blocks outside i_size. Trim these off again.
348 if (unlikely((rw
& WRITE
) && ret
< 0)) {
349 loff_t isize
= i_size_read(inode
);
350 loff_t end
= offset
+ iov_length(iov
, nr_segs
);
353 jfs_write_failed(mapping
, end
);
359 const struct address_space_operations jfs_aops
= {
360 .readpage
= jfs_readpage
,
361 .readpages
= jfs_readpages
,
362 .writepage
= jfs_writepage
,
363 .writepages
= jfs_writepages
,
364 .write_begin
= jfs_write_begin
,
365 .write_end
= nobh_write_end
,
367 .direct_IO
= jfs_direct_IO
,
371 * Guts of jfs_truncate. Called with locks already held. Can be called
372 * with directory for truncating directory index table.
374 void jfs_truncate_nolock(struct inode
*ip
, loff_t length
)
381 if (test_cflag(COMMIT_Nolink
, ip
)) {
382 xtTruncate(0, ip
, length
, COMMIT_WMAP
);
387 tid
= txBegin(ip
->i_sb
, 0);
390 * The commit_mutex cannot be taken before txBegin.
391 * txBegin may block and there is a chance the inode
392 * could be marked dirty and need to be committed
393 * before txBegin unblocks
395 mutex_lock(&JFS_IP(ip
)->commit_mutex
);
397 newsize
= xtTruncate(tid
, ip
, length
,
398 COMMIT_TRUNCATE
| COMMIT_PWMAP
);
401 mutex_unlock(&JFS_IP(ip
)->commit_mutex
);
405 ip
->i_mtime
= ip
->i_ctime
= CURRENT_TIME
;
406 mark_inode_dirty(ip
);
408 txCommit(tid
, 1, &ip
, 0);
410 mutex_unlock(&JFS_IP(ip
)->commit_mutex
);
411 } while (newsize
> length
); /* Truncate isn't always atomic */
414 void jfs_truncate(struct inode
*ip
)
416 jfs_info("jfs_truncate: size = 0x%lx", (ulong
) ip
->i_size
);
418 nobh_truncate_page(ip
->i_mapping
, ip
->i_size
, jfs_get_block
);
420 IWRITE_LOCK(ip
, RDWRLOCK_NORMAL
);
421 jfs_truncate_nolock(ip
, ip
->i_size
);