Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (C) International Business Machines Corp., 2000-2004 |
| 3 | * Portions Copyright (C) Christoph Hellwig, 2001-2002 |
| 4 | * |
| 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. |
| 9 | * |
| 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. |
| 14 | * |
| 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 |
| 18 | */ |
| 19 | |
| 20 | #include <linux/fs.h> |
| 21 | #include <linux/mpage.h> |
| 22 | #include <linux/buffer_head.h> |
| 23 | #include <linux/pagemap.h> |
| 24 | #include <linux/quotaops.h> |
| 25 | #include "jfs_incore.h" |
Dave Kleikamp | 1868f4a | 2005-05-04 15:29:35 -0500 | [diff] [blame] | 26 | #include "jfs_inode.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 27 | #include "jfs_filsys.h" |
| 28 | #include "jfs_imap.h" |
| 29 | #include "jfs_extent.h" |
| 30 | #include "jfs_unicode.h" |
| 31 | #include "jfs_debug.h" |
| 32 | |
| 33 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 34 | void jfs_read_inode(struct inode *inode) |
| 35 | { |
| 36 | if (diRead(inode)) { |
| 37 | make_bad_inode(inode); |
| 38 | return; |
| 39 | } |
| 40 | |
| 41 | if (S_ISREG(inode->i_mode)) { |
| 42 | inode->i_op = &jfs_file_inode_operations; |
| 43 | inode->i_fop = &jfs_file_operations; |
| 44 | inode->i_mapping->a_ops = &jfs_aops; |
| 45 | } else if (S_ISDIR(inode->i_mode)) { |
| 46 | inode->i_op = &jfs_dir_inode_operations; |
| 47 | inode->i_fop = &jfs_dir_operations; |
| 48 | } else if (S_ISLNK(inode->i_mode)) { |
| 49 | if (inode->i_size >= IDATASIZE) { |
| 50 | inode->i_op = &page_symlink_inode_operations; |
| 51 | inode->i_mapping->a_ops = &jfs_aops; |
| 52 | } else |
| 53 | inode->i_op = &jfs_symlink_inode_operations; |
| 54 | } else { |
| 55 | inode->i_op = &jfs_file_inode_operations; |
| 56 | init_special_inode(inode, inode->i_mode, inode->i_rdev); |
| 57 | } |
| 58 | } |
| 59 | |
| 60 | /* |
| 61 | * Workhorse of both fsync & write_inode |
| 62 | */ |
| 63 | int jfs_commit_inode(struct inode *inode, int wait) |
| 64 | { |
| 65 | int rc = 0; |
| 66 | tid_t tid; |
| 67 | static int noisy = 5; |
| 68 | |
| 69 | jfs_info("In jfs_commit_inode, inode = 0x%p", inode); |
| 70 | |
| 71 | /* |
| 72 | * Don't commit if inode has been committed since last being |
| 73 | * marked dirty, or if it has been deleted. |
| 74 | */ |
| 75 | if (inode->i_nlink == 0 || !test_cflag(COMMIT_Dirty, inode)) |
| 76 | return 0; |
| 77 | |
| 78 | if (isReadOnly(inode)) { |
| 79 | /* kernel allows writes to devices on read-only |
| 80 | * partitions and may think inode is dirty |
| 81 | */ |
| 82 | if (!special_file(inode->i_mode) && noisy) { |
| 83 | jfs_err("jfs_commit_inode(0x%p) called on " |
| 84 | "read-only volume", inode); |
| 85 | jfs_err("Is remount racy?"); |
| 86 | noisy--; |
| 87 | } |
| 88 | return 0; |
| 89 | } |
| 90 | |
| 91 | tid = txBegin(inode->i_sb, COMMIT_INODE); |
| 92 | down(&JFS_IP(inode)->commit_sem); |
| 93 | |
| 94 | /* |
| 95 | * Retest inode state after taking commit_sem |
| 96 | */ |
| 97 | if (inode->i_nlink && test_cflag(COMMIT_Dirty, inode)) |
| 98 | rc = txCommit(tid, 1, &inode, wait ? COMMIT_SYNC : 0); |
| 99 | |
| 100 | txEnd(tid); |
| 101 | up(&JFS_IP(inode)->commit_sem); |
| 102 | return rc; |
| 103 | } |
| 104 | |
| 105 | int jfs_write_inode(struct inode *inode, int wait) |
| 106 | { |
| 107 | if (test_cflag(COMMIT_Nolink, inode)) |
| 108 | return 0; |
| 109 | /* |
| 110 | * If COMMIT_DIRTY is not set, the inode isn't really dirty. |
| 111 | * It has been committed since the last change, but was still |
| 112 | * on the dirty inode list. |
| 113 | */ |
| 114 | if (!test_cflag(COMMIT_Dirty, inode)) { |
| 115 | /* Make sure committed changes hit the disk */ |
| 116 | jfs_flush_journal(JFS_SBI(inode->i_sb)->log, wait); |
| 117 | return 0; |
| 118 | } |
| 119 | |
| 120 | if (jfs_commit_inode(inode, wait)) { |
| 121 | jfs_err("jfs_write_inode: jfs_commit_inode failed!"); |
| 122 | return -EIO; |
| 123 | } else |
| 124 | return 0; |
| 125 | } |
| 126 | |
| 127 | void jfs_delete_inode(struct inode *inode) |
| 128 | { |
| 129 | jfs_info("In jfs_delete_inode, inode = 0x%p", inode); |
| 130 | |
Dave Kleikamp | b1b5d7f | 2005-08-30 14:28:56 -0500 | [diff] [blame] | 131 | if (!is_bad_inode(inode) && |
Dave Kleikamp | 6cb1269 | 2005-09-15 23:25:41 -0500 | [diff] [blame^] | 132 | (JFS_IP(inode)->fileset == FILESYSTEM_I)) { |
Linus Torvalds | 3298369 | 2005-09-11 10:14:54 -0700 | [diff] [blame] | 133 | truncate_inode_pages(&inode->i_data, 0); |
Mark Fasheh | fef2665 | 2005-09-09 13:01:31 -0700 | [diff] [blame] | 134 | |
Dave Kleikamp | b1b5d7f | 2005-08-30 14:28:56 -0500 | [diff] [blame] | 135 | if (test_cflag(COMMIT_Freewmap, inode)) |
| 136 | jfs_free_zero_link(inode); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 137 | |
Dave Kleikamp | b1b5d7f | 2005-08-30 14:28:56 -0500 | [diff] [blame] | 138 | diFree(inode); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 139 | |
Dave Kleikamp | b1b5d7f | 2005-08-30 14:28:56 -0500 | [diff] [blame] | 140 | /* |
| 141 | * Free the inode from the quota allocation. |
| 142 | */ |
| 143 | DQUOT_INIT(inode); |
| 144 | DQUOT_FREE_INODE(inode); |
| 145 | DQUOT_DROP(inode); |
| 146 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 147 | |
| 148 | clear_inode(inode); |
| 149 | } |
| 150 | |
| 151 | void jfs_dirty_inode(struct inode *inode) |
| 152 | { |
| 153 | static int noisy = 5; |
| 154 | |
| 155 | if (isReadOnly(inode)) { |
| 156 | if (!special_file(inode->i_mode) && noisy) { |
| 157 | /* kernel allows writes to devices on read-only |
| 158 | * partitions and may try to mark inode dirty |
| 159 | */ |
| 160 | jfs_err("jfs_dirty_inode called on read-only volume"); |
| 161 | jfs_err("Is remount racy?"); |
| 162 | noisy--; |
| 163 | } |
| 164 | return; |
| 165 | } |
| 166 | |
| 167 | set_cflag(COMMIT_Dirty, inode); |
| 168 | } |
| 169 | |
| 170 | static int |
| 171 | jfs_get_blocks(struct inode *ip, sector_t lblock, unsigned long max_blocks, |
| 172 | struct buffer_head *bh_result, int create) |
| 173 | { |
| 174 | s64 lblock64 = lblock; |
| 175 | int rc = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 176 | xad_t xad; |
| 177 | s64 xaddr; |
| 178 | int xflag; |
Dave Kleikamp | 6628465 | 2005-05-02 12:25:13 -0600 | [diff] [blame] | 179 | s32 xlen = max_blocks; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 180 | |
| 181 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 182 | * Take appropriate lock on inode |
| 183 | */ |
Dave Kleikamp | 7fab479 | 2005-05-02 12:25:02 -0600 | [diff] [blame] | 184 | if (create) |
| 185 | IWRITE_LOCK(ip); |
| 186 | else |
| 187 | IREAD_LOCK(ip); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 188 | |
| 189 | if (((lblock64 << ip->i_sb->s_blocksize_bits) < ip->i_size) && |
Dave Kleikamp | 7fab479 | 2005-05-02 12:25:02 -0600 | [diff] [blame] | 190 | (!xtLookup(ip, lblock64, max_blocks, &xflag, &xaddr, &xlen, 0)) && |
Dave Kleikamp | 6628465 | 2005-05-02 12:25:13 -0600 | [diff] [blame] | 191 | xaddr) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 192 | if (xflag & XAD_NOTRECORDED) { |
| 193 | if (!create) |
| 194 | /* |
| 195 | * Allocated but not recorded, read treats |
| 196 | * this as a hole |
| 197 | */ |
| 198 | goto unlock; |
| 199 | #ifdef _JFS_4K |
| 200 | XADoffset(&xad, lblock64); |
| 201 | XADlength(&xad, xlen); |
| 202 | XADaddress(&xad, xaddr); |
| 203 | #else /* _JFS_4K */ |
| 204 | /* |
| 205 | * As long as block size = 4K, this isn't a problem. |
| 206 | * We should mark the whole page not ABNR, but how |
| 207 | * will we know to mark the other blocks BH_New? |
| 208 | */ |
| 209 | BUG(); |
| 210 | #endif /* _JFS_4K */ |
| 211 | rc = extRecord(ip, &xad); |
| 212 | if (rc) |
| 213 | goto unlock; |
| 214 | set_buffer_new(bh_result); |
| 215 | } |
| 216 | |
| 217 | map_bh(bh_result, ip->i_sb, xaddr); |
| 218 | bh_result->b_size = xlen << ip->i_blkbits; |
| 219 | goto unlock; |
| 220 | } |
| 221 | if (!create) |
| 222 | goto unlock; |
| 223 | |
| 224 | /* |
| 225 | * Allocate a new block |
| 226 | */ |
| 227 | #ifdef _JFS_4K |
| 228 | if ((rc = extHint(ip, lblock64 << ip->i_sb->s_blocksize_bits, &xad))) |
| 229 | goto unlock; |
Dave Kleikamp | 6628465 | 2005-05-02 12:25:13 -0600 | [diff] [blame] | 230 | rc = extAlloc(ip, xlen, lblock64, &xad, FALSE); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 231 | if (rc) |
| 232 | goto unlock; |
| 233 | |
| 234 | set_buffer_new(bh_result); |
| 235 | map_bh(bh_result, ip->i_sb, addressXAD(&xad)); |
| 236 | bh_result->b_size = lengthXAD(&xad) << ip->i_blkbits; |
| 237 | |
| 238 | #else /* _JFS_4K */ |
| 239 | /* |
| 240 | * We need to do whatever it takes to keep all but the last buffers |
| 241 | * in 4K pages - see jfs_write.c |
| 242 | */ |
| 243 | BUG(); |
| 244 | #endif /* _JFS_4K */ |
| 245 | |
| 246 | unlock: |
| 247 | /* |
| 248 | * Release lock on inode |
| 249 | */ |
Dave Kleikamp | 7fab479 | 2005-05-02 12:25:02 -0600 | [diff] [blame] | 250 | if (create) |
| 251 | IWRITE_UNLOCK(ip); |
| 252 | else |
| 253 | IREAD_UNLOCK(ip); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 254 | return rc; |
| 255 | } |
| 256 | |
| 257 | static int jfs_get_block(struct inode *ip, sector_t lblock, |
| 258 | struct buffer_head *bh_result, int create) |
| 259 | { |
| 260 | return jfs_get_blocks(ip, lblock, 1, bh_result, create); |
| 261 | } |
| 262 | |
| 263 | static int jfs_writepage(struct page *page, struct writeback_control *wbc) |
| 264 | { |
| 265 | return nobh_writepage(page, jfs_get_block, wbc); |
| 266 | } |
| 267 | |
| 268 | static int jfs_writepages(struct address_space *mapping, |
| 269 | struct writeback_control *wbc) |
| 270 | { |
| 271 | return mpage_writepages(mapping, wbc, jfs_get_block); |
| 272 | } |
| 273 | |
| 274 | static int jfs_readpage(struct file *file, struct page *page) |
| 275 | { |
| 276 | return mpage_readpage(page, jfs_get_block); |
| 277 | } |
| 278 | |
| 279 | static int jfs_readpages(struct file *file, struct address_space *mapping, |
| 280 | struct list_head *pages, unsigned nr_pages) |
| 281 | { |
| 282 | return mpage_readpages(mapping, pages, nr_pages, jfs_get_block); |
| 283 | } |
| 284 | |
| 285 | static int jfs_prepare_write(struct file *file, |
| 286 | struct page *page, unsigned from, unsigned to) |
| 287 | { |
| 288 | return nobh_prepare_write(page, from, to, jfs_get_block); |
| 289 | } |
| 290 | |
| 291 | static sector_t jfs_bmap(struct address_space *mapping, sector_t block) |
| 292 | { |
| 293 | return generic_block_bmap(mapping, block, jfs_get_block); |
| 294 | } |
| 295 | |
| 296 | static ssize_t jfs_direct_IO(int rw, struct kiocb *iocb, |
| 297 | const struct iovec *iov, loff_t offset, unsigned long nr_segs) |
| 298 | { |
| 299 | struct file *file = iocb->ki_filp; |
| 300 | struct inode *inode = file->f_mapping->host; |
| 301 | |
| 302 | return blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov, |
| 303 | offset, nr_segs, jfs_get_blocks, NULL); |
| 304 | } |
| 305 | |
| 306 | struct address_space_operations jfs_aops = { |
| 307 | .readpage = jfs_readpage, |
| 308 | .readpages = jfs_readpages, |
| 309 | .writepage = jfs_writepage, |
| 310 | .writepages = jfs_writepages, |
| 311 | .sync_page = block_sync_page, |
| 312 | .prepare_write = jfs_prepare_write, |
| 313 | .commit_write = nobh_commit_write, |
| 314 | .bmap = jfs_bmap, |
| 315 | .direct_IO = jfs_direct_IO, |
| 316 | }; |
| 317 | |
| 318 | /* |
| 319 | * Guts of jfs_truncate. Called with locks already held. Can be called |
| 320 | * with directory for truncating directory index table. |
| 321 | */ |
| 322 | void jfs_truncate_nolock(struct inode *ip, loff_t length) |
| 323 | { |
| 324 | loff_t newsize; |
| 325 | tid_t tid; |
| 326 | |
| 327 | ASSERT(length >= 0); |
| 328 | |
| 329 | if (test_cflag(COMMIT_Nolink, ip)) { |
| 330 | xtTruncate(0, ip, length, COMMIT_WMAP); |
| 331 | return; |
| 332 | } |
| 333 | |
| 334 | do { |
| 335 | tid = txBegin(ip->i_sb, 0); |
| 336 | |
| 337 | /* |
| 338 | * The commit_sem cannot be taken before txBegin. |
| 339 | * txBegin may block and there is a chance the inode |
| 340 | * could be marked dirty and need to be committed |
| 341 | * before txBegin unblocks |
| 342 | */ |
| 343 | down(&JFS_IP(ip)->commit_sem); |
| 344 | |
| 345 | newsize = xtTruncate(tid, ip, length, |
| 346 | COMMIT_TRUNCATE | COMMIT_PWMAP); |
| 347 | if (newsize < 0) { |
| 348 | txEnd(tid); |
| 349 | up(&JFS_IP(ip)->commit_sem); |
| 350 | break; |
| 351 | } |
| 352 | |
| 353 | ip->i_mtime = ip->i_ctime = CURRENT_TIME; |
| 354 | mark_inode_dirty(ip); |
| 355 | |
| 356 | txCommit(tid, 1, &ip, 0); |
| 357 | txEnd(tid); |
| 358 | up(&JFS_IP(ip)->commit_sem); |
| 359 | } while (newsize > length); /* Truncate isn't always atomic */ |
| 360 | } |
| 361 | |
| 362 | void jfs_truncate(struct inode *ip) |
| 363 | { |
| 364 | jfs_info("jfs_truncate: size = 0x%lx", (ulong) ip->i_size); |
| 365 | |
| 366 | nobh_truncate_page(ip->i_mapping, ip->i_size); |
| 367 | |
| 368 | IWRITE_LOCK(ip); |
| 369 | jfs_truncate_nolock(ip, ip->i_size); |
| 370 | IWRITE_UNLOCK(ip); |
| 371 | } |