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/uio.h>
26 #include <linux/writeback.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_nohighmem(inode);
64 inode->i_mapping->a_ops = &jfs_aops;
66 inode->i_op = &jfs_fast_symlink_inode_operations;
67 inode->i_link = JFS_IP(inode)->i_inline;
69 * The inline data should be null-terminated, but
70 * don't let on-disk corruption crash the kernel
72 inode->i_link[inode->i_size] = '\0';
75 inode->i_op = &jfs_file_inode_operations;
76 init_special_inode(inode, inode->i_mode, inode->i_rdev);
78 unlock_new_inode(inode);
83 * Workhorse of both fsync & write_inode
85 int jfs_commit_inode(struct inode *inode, int wait)
91 jfs_info("In jfs_commit_inode, inode = 0x%p", inode);
94 * Don't commit if inode has been committed since last being
95 * marked dirty, or if it has been deleted.
97 if (inode->i_nlink == 0 || !test_cflag(COMMIT_Dirty, inode))
100 if (isReadOnly(inode)) {
101 /* kernel allows writes to devices on read-only
102 * partitions and may think inode is dirty
104 if (!special_file(inode->i_mode) && noisy) {
105 jfs_err("jfs_commit_inode(0x%p) called on read-only volume",
107 jfs_err("Is remount racy?");
113 tid = txBegin(inode->i_sb, COMMIT_INODE);
114 mutex_lock(&JFS_IP(inode)->commit_mutex);
117 * Retest inode state after taking commit_mutex
119 if (inode->i_nlink && test_cflag(COMMIT_Dirty, inode))
120 rc = txCommit(tid, 1, &inode, wait ? COMMIT_SYNC : 0);
123 mutex_unlock(&JFS_IP(inode)->commit_mutex);
127 int jfs_write_inode(struct inode *inode, struct writeback_control *wbc)
129 int wait = wbc->sync_mode == WB_SYNC_ALL;
131 if (inode->i_nlink == 0)
134 * If COMMIT_DIRTY is not set, the inode isn't really dirty.
135 * It has been committed since the last change, but was still
136 * on the dirty inode list.
138 if (!test_cflag(COMMIT_Dirty, inode)) {
139 /* Make sure committed changes hit the disk */
140 jfs_flush_journal(JFS_SBI(inode->i_sb)->log, wait);
144 if (jfs_commit_inode(inode, wait)) {
145 jfs_err("jfs_write_inode: jfs_commit_inode failed!");
151 void jfs_evict_inode(struct inode *inode)
153 jfs_info("In jfs_evict_inode, inode = 0x%p", inode);
155 if (!inode->i_nlink && !is_bad_inode(inode)) {
156 dquot_initialize(inode);
158 if (JFS_IP(inode)->fileset == FILESYSTEM_I) {
159 struct inode *ipimap = JFS_SBI(inode->i_sb)->ipimap;
160 truncate_inode_pages_final(&inode->i_data);
162 if (test_cflag(COMMIT_Freewmap, inode))
163 jfs_free_zero_link(inode);
165 if (ipimap && JFS_IP(ipimap)->i_imap)
169 * Free the inode from the quota allocation.
171 dquot_initialize(inode);
172 dquot_free_inode(inode);
175 truncate_inode_pages_final(&inode->i_data);
181 void jfs_dirty_inode(struct inode *inode, int flags)
183 static int noisy = 5;
185 if (isReadOnly(inode)) {
186 if (!special_file(inode->i_mode) && noisy) {
187 /* kernel allows writes to devices on read-only
188 * partitions and may try to mark inode dirty
190 jfs_err("jfs_dirty_inode called on read-only volume");
191 jfs_err("Is remount racy?");
197 set_cflag(COMMIT_Dirty, inode);
200 int jfs_get_block(struct inode *ip, sector_t lblock,
201 struct buffer_head *bh_result, int create)
203 s64 lblock64 = lblock;
208 s32 xlen = bh_result->b_size >> ip->i_blkbits;
211 * Take appropriate lock on inode
214 IWRITE_LOCK(ip, RDWRLOCK_NORMAL);
216 IREAD_LOCK(ip, RDWRLOCK_NORMAL);
218 if (((lblock64 << ip->i_sb->s_blocksize_bits) < ip->i_size) &&
219 (!xtLookup(ip, lblock64, xlen, &xflag, &xaddr, &xlen, 0)) &&
221 if (xflag & XAD_NOTRECORDED) {
224 * Allocated but not recorded, read treats
229 XADoffset(&xad, lblock64);
230 XADlength(&xad, xlen);
231 XADaddress(&xad, xaddr);
234 * As long as block size = 4K, this isn't a problem.
235 * We should mark the whole page not ABNR, but how
236 * will we know to mark the other blocks BH_New?
240 rc = extRecord(ip, &xad);
243 set_buffer_new(bh_result);
246 map_bh(bh_result, ip->i_sb, xaddr);
247 bh_result->b_size = xlen << ip->i_blkbits;
254 * Allocate a new block
257 if ((rc = extHint(ip, lblock64 << ip->i_sb->s_blocksize_bits, &xad)))
259 rc = extAlloc(ip, xlen, lblock64, &xad, false);
263 set_buffer_new(bh_result);
264 map_bh(bh_result, ip->i_sb, addressXAD(&xad));
265 bh_result->b_size = lengthXAD(&xad) << ip->i_blkbits;
269 * We need to do whatever it takes to keep all but the last buffers
270 * in 4K pages - see jfs_write.c
277 * Release lock on inode
286 static int jfs_writepage(struct page *page, struct writeback_control *wbc)
288 return block_write_full_page(page, jfs_get_block, wbc);
291 static int jfs_writepages(struct address_space *mapping,
292 struct writeback_control *wbc)
294 return mpage_writepages(mapping, wbc, jfs_get_block);
297 static int jfs_readpage(struct file *file, struct page *page)
299 return mpage_readpage(page, jfs_get_block);
302 static int jfs_readpages(struct file *file, struct address_space *mapping,
303 struct list_head *pages, unsigned nr_pages)
305 return mpage_readpages(mapping, pages, nr_pages, jfs_get_block);
308 static void jfs_write_failed(struct address_space *mapping, loff_t to)
310 struct inode *inode = mapping->host;
312 if (to > inode->i_size) {
313 truncate_pagecache(inode, inode->i_size);
318 static int jfs_write_begin(struct file *file, struct address_space *mapping,
319 loff_t pos, unsigned len, unsigned flags,
320 struct page **pagep, void **fsdata)
324 ret = nobh_write_begin(mapping, pos, len, flags, pagep, fsdata,
327 jfs_write_failed(mapping, pos + len);
332 static sector_t jfs_bmap(struct address_space *mapping, sector_t block)
334 return generic_block_bmap(mapping, block, jfs_get_block);
337 static ssize_t jfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
339 struct file *file = iocb->ki_filp;
340 struct address_space *mapping = file->f_mapping;
341 struct inode *inode = file->f_mapping->host;
342 size_t count = iov_iter_count(iter);
345 ret = blockdev_direct_IO(iocb, inode, iter, jfs_get_block);
348 * In case of error extending write may have instantiated a few
349 * blocks outside i_size. Trim these off again.
351 if (unlikely(iov_iter_rw(iter) == WRITE && ret < 0)) {
352 loff_t isize = i_size_read(inode);
353 loff_t end = iocb->ki_pos + count;
356 jfs_write_failed(mapping, end);
362 const struct address_space_operations jfs_aops = {
363 .readpage = jfs_readpage,
364 .readpages = jfs_readpages,
365 .writepage = jfs_writepage,
366 .writepages = jfs_writepages,
367 .write_begin = jfs_write_begin,
368 .write_end = nobh_write_end,
370 .direct_IO = jfs_direct_IO,
374 * Guts of jfs_truncate. Called with locks already held. Can be called
375 * with directory for truncating directory index table.
377 void jfs_truncate_nolock(struct inode *ip, loff_t length)
384 if (test_cflag(COMMIT_Nolink, ip)) {
385 xtTruncate(0, ip, length, COMMIT_WMAP);
390 tid = txBegin(ip->i_sb, 0);
393 * The commit_mutex cannot be taken before txBegin.
394 * txBegin may block and there is a chance the inode
395 * could be marked dirty and need to be committed
396 * before txBegin unblocks
398 mutex_lock(&JFS_IP(ip)->commit_mutex);
400 newsize = xtTruncate(tid, ip, length,
401 COMMIT_TRUNCATE | COMMIT_PWMAP);
404 mutex_unlock(&JFS_IP(ip)->commit_mutex);
408 ip->i_mtime = ip->i_ctime = current_time(ip);
409 mark_inode_dirty(ip);
411 txCommit(tid, 1, &ip, 0);
413 mutex_unlock(&JFS_IP(ip)->commit_mutex);
414 } while (newsize > length); /* Truncate isn't always atomic */
417 void jfs_truncate(struct inode *ip)
419 jfs_info("jfs_truncate: size = 0x%lx", (ulong) ip->i_size);
421 nobh_truncate_page(ip->i_mapping, ip->i_size, jfs_get_block);
423 IWRITE_LOCK(ip, RDWRLOCK_NORMAL);
424 jfs_truncate_nolock(ip, ip->i_size);