Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 1 | /* |
| 2 | * inode.c - NILFS inode operations. |
| 3 | * |
| 4 | * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation. |
| 5 | * |
| 6 | * This program is free software; you can redistribute it and/or modify |
| 7 | * it under the terms of the GNU General Public License as published by |
| 8 | * the Free Software Foundation; either version 2 of the License, or |
| 9 | * (at your option) any later version. |
| 10 | * |
| 11 | * This program is distributed in the hope that it will be useful, |
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 14 | * GNU General Public License for more details. |
| 15 | * |
| 16 | * You should have received a copy of the GNU General Public License |
| 17 | * along with this program; if not, write to the Free Software |
| 18 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA |
| 19 | * |
| 20 | * Written by Ryusuke Konishi <ryusuke@osrg.net> |
| 21 | * |
| 22 | */ |
| 23 | |
| 24 | #include <linux/buffer_head.h> |
| 25 | #include <linux/mpage.h> |
| 26 | #include <linux/writeback.h> |
Ryusuke Konishi | f30bf3e | 2009-04-06 19:01:38 -0700 | [diff] [blame] | 27 | #include <linux/uio.h> |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 28 | #include "nilfs.h" |
| 29 | #include "segment.h" |
| 30 | #include "page.h" |
| 31 | #include "mdt.h" |
| 32 | #include "cpfile.h" |
| 33 | #include "ifile.h" |
| 34 | |
| 35 | |
| 36 | /** |
| 37 | * nilfs_get_block() - get a file block on the filesystem (callback function) |
| 38 | * @inode - inode struct of the target file |
| 39 | * @blkoff - file block number |
| 40 | * @bh_result - buffer head to be mapped on |
| 41 | * @create - indicate whether allocating the block or not when it has not |
| 42 | * been allocated yet. |
| 43 | * |
| 44 | * This function does not issue actual read request of the specified data |
| 45 | * block. It is done by VFS. |
| 46 | * Bulk read for direct-io is not supported yet. (should be supported) |
| 47 | */ |
| 48 | int nilfs_get_block(struct inode *inode, sector_t blkoff, |
| 49 | struct buffer_head *bh_result, int create) |
| 50 | { |
| 51 | struct nilfs_inode_info *ii = NILFS_I(inode); |
| 52 | unsigned long blknum = 0; |
| 53 | int err = 0, ret; |
| 54 | struct inode *dat = nilfs_dat_inode(NILFS_I_NILFS(inode)); |
| 55 | |
| 56 | /* This exclusion control is a workaround; should be revised */ |
| 57 | down_read(&NILFS_MDT(dat)->mi_sem); /* XXX */ |
| 58 | ret = nilfs_bmap_lookup(ii->i_bmap, (unsigned long)blkoff, &blknum); |
| 59 | up_read(&NILFS_MDT(dat)->mi_sem); /* XXX */ |
| 60 | if (ret == 0) { /* found */ |
| 61 | map_bh(bh_result, inode->i_sb, blknum); |
| 62 | goto out; |
| 63 | } |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 64 | /* data block was not found */ |
| 65 | if (ret == -ENOENT && create) { |
| 66 | struct nilfs_transaction_info ti; |
| 67 | |
| 68 | bh_result->b_blocknr = 0; |
| 69 | err = nilfs_transaction_begin(inode->i_sb, &ti, 1); |
| 70 | if (unlikely(err)) |
| 71 | goto out; |
| 72 | err = nilfs_bmap_insert(ii->i_bmap, (unsigned long)blkoff, |
| 73 | (unsigned long)bh_result); |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 74 | if (unlikely(err != 0)) { |
| 75 | if (err == -EEXIST) { |
| 76 | /* |
| 77 | * The get_block() function could be called |
| 78 | * from multiple callers for an inode. |
| 79 | * However, the page having this block must |
| 80 | * be locked in this case. |
| 81 | */ |
Ryusuke Konishi | 1f5abe7 | 2009-04-06 19:01:55 -0700 | [diff] [blame] | 82 | printk(KERN_WARNING |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 83 | "nilfs_get_block: a race condition " |
| 84 | "while inserting a data block. " |
| 85 | "(inode number=%lu, file block " |
| 86 | "offset=%llu)\n", |
| 87 | inode->i_ino, |
| 88 | (unsigned long long)blkoff); |
Ryusuke Konishi | 1f5abe7 | 2009-04-06 19:01:55 -0700 | [diff] [blame] | 89 | err = 0; |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 90 | } else if (err == -EINVAL) { |
| 91 | nilfs_error(inode->i_sb, __func__, |
| 92 | "broken bmap (inode=%lu)\n", |
| 93 | inode->i_ino); |
| 94 | err = -EIO; |
| 95 | } |
Ryusuke Konishi | 47420c7 | 2009-04-06 19:01:45 -0700 | [diff] [blame] | 96 | nilfs_transaction_abort(inode->i_sb); |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 97 | goto out; |
| 98 | } |
Ryusuke Konishi | 47420c7 | 2009-04-06 19:01:45 -0700 | [diff] [blame] | 99 | nilfs_transaction_commit(inode->i_sb); /* never fails */ |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 100 | /* Error handling should be detailed */ |
| 101 | set_buffer_new(bh_result); |
| 102 | map_bh(bh_result, inode->i_sb, 0); /* dbn must be changed |
| 103 | to proper value */ |
| 104 | } else if (ret == -ENOENT) { |
| 105 | /* not found is not error (e.g. hole); must return without |
| 106 | the mapped state flag. */ |
| 107 | ; |
| 108 | } else { |
| 109 | err = ret; |
| 110 | } |
| 111 | |
| 112 | out: |
| 113 | return err; |
| 114 | } |
| 115 | |
| 116 | /** |
| 117 | * nilfs_readpage() - implement readpage() method of nilfs_aops {} |
| 118 | * address_space_operations. |
| 119 | * @file - file struct of the file to be read |
| 120 | * @page - the page to be read |
| 121 | */ |
| 122 | static int nilfs_readpage(struct file *file, struct page *page) |
| 123 | { |
| 124 | return mpage_readpage(page, nilfs_get_block); |
| 125 | } |
| 126 | |
| 127 | /** |
| 128 | * nilfs_readpages() - implement readpages() method of nilfs_aops {} |
| 129 | * address_space_operations. |
| 130 | * @file - file struct of the file to be read |
| 131 | * @mapping - address_space struct used for reading multiple pages |
| 132 | * @pages - the pages to be read |
| 133 | * @nr_pages - number of pages to be read |
| 134 | */ |
| 135 | static int nilfs_readpages(struct file *file, struct address_space *mapping, |
| 136 | struct list_head *pages, unsigned nr_pages) |
| 137 | { |
| 138 | return mpage_readpages(mapping, pages, nr_pages, nilfs_get_block); |
| 139 | } |
| 140 | |
| 141 | static int nilfs_writepages(struct address_space *mapping, |
| 142 | struct writeback_control *wbc) |
| 143 | { |
Ryusuke Konishi | f30bf3e | 2009-04-06 19:01:38 -0700 | [diff] [blame] | 144 | struct inode *inode = mapping->host; |
| 145 | int err = 0; |
| 146 | |
| 147 | if (wbc->sync_mode == WB_SYNC_ALL) |
| 148 | err = nilfs_construct_dsync_segment(inode->i_sb, inode, |
| 149 | wbc->range_start, |
| 150 | wbc->range_end); |
| 151 | return err; |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 152 | } |
| 153 | |
| 154 | static int nilfs_writepage(struct page *page, struct writeback_control *wbc) |
| 155 | { |
| 156 | struct inode *inode = page->mapping->host; |
| 157 | int err; |
| 158 | |
| 159 | redirty_page_for_writepage(wbc, page); |
| 160 | unlock_page(page); |
| 161 | |
| 162 | if (wbc->sync_mode == WB_SYNC_ALL) { |
| 163 | err = nilfs_construct_segment(inode->i_sb); |
| 164 | if (unlikely(err)) |
| 165 | return err; |
| 166 | } else if (wbc->for_reclaim) |
| 167 | nilfs_flush_segment(inode->i_sb, inode->i_ino); |
| 168 | |
| 169 | return 0; |
| 170 | } |
| 171 | |
| 172 | static int nilfs_set_page_dirty(struct page *page) |
| 173 | { |
| 174 | int ret = __set_page_dirty_buffers(page); |
| 175 | |
| 176 | if (ret) { |
| 177 | struct inode *inode = page->mapping->host; |
| 178 | struct nilfs_sb_info *sbi = NILFS_SB(inode->i_sb); |
| 179 | unsigned nr_dirty = 1 << (PAGE_SHIFT - inode->i_blkbits); |
| 180 | |
| 181 | nilfs_set_file_dirty(sbi, inode, nr_dirty); |
| 182 | } |
| 183 | return ret; |
| 184 | } |
| 185 | |
| 186 | static int nilfs_write_begin(struct file *file, struct address_space *mapping, |
| 187 | loff_t pos, unsigned len, unsigned flags, |
| 188 | struct page **pagep, void **fsdata) |
| 189 | |
| 190 | { |
| 191 | struct inode *inode = mapping->host; |
| 192 | int err = nilfs_transaction_begin(inode->i_sb, NULL, 1); |
| 193 | |
| 194 | if (unlikely(err)) |
| 195 | return err; |
| 196 | |
| 197 | *pagep = NULL; |
| 198 | err = block_write_begin(file, mapping, pos, len, flags, pagep, |
| 199 | fsdata, nilfs_get_block); |
| 200 | if (unlikely(err)) |
Ryusuke Konishi | 47420c7 | 2009-04-06 19:01:45 -0700 | [diff] [blame] | 201 | nilfs_transaction_abort(inode->i_sb); |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 202 | return err; |
| 203 | } |
| 204 | |
| 205 | static int nilfs_write_end(struct file *file, struct address_space *mapping, |
| 206 | loff_t pos, unsigned len, unsigned copied, |
| 207 | struct page *page, void *fsdata) |
| 208 | { |
| 209 | struct inode *inode = mapping->host; |
| 210 | unsigned start = pos & (PAGE_CACHE_SIZE - 1); |
| 211 | unsigned nr_dirty; |
| 212 | int err; |
| 213 | |
| 214 | nr_dirty = nilfs_page_count_clean_buffers(page, start, |
| 215 | start + copied); |
| 216 | copied = generic_write_end(file, mapping, pos, len, copied, page, |
| 217 | fsdata); |
| 218 | nilfs_set_file_dirty(NILFS_SB(inode->i_sb), inode, nr_dirty); |
Ryusuke Konishi | 47420c7 | 2009-04-06 19:01:45 -0700 | [diff] [blame] | 219 | err = nilfs_transaction_commit(inode->i_sb); |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 220 | return err ? : copied; |
| 221 | } |
| 222 | |
| 223 | static ssize_t |
| 224 | nilfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, |
| 225 | loff_t offset, unsigned long nr_segs) |
| 226 | { |
| 227 | struct file *file = iocb->ki_filp; |
| 228 | struct inode *inode = file->f_mapping->host; |
| 229 | ssize_t size; |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 230 | |
| 231 | if (rw == WRITE) |
| 232 | return 0; |
| 233 | |
| 234 | /* Needs synchronization with the cleaner */ |
| 235 | size = blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov, |
| 236 | offset, nr_segs, nilfs_get_block, NULL); |
| 237 | return size; |
| 238 | } |
| 239 | |
| 240 | struct address_space_operations nilfs_aops = { |
| 241 | .writepage = nilfs_writepage, |
| 242 | .readpage = nilfs_readpage, |
| 243 | /* .sync_page = nilfs_sync_page, */ |
| 244 | .writepages = nilfs_writepages, |
| 245 | .set_page_dirty = nilfs_set_page_dirty, |
| 246 | .readpages = nilfs_readpages, |
| 247 | .write_begin = nilfs_write_begin, |
| 248 | .write_end = nilfs_write_end, |
| 249 | /* .releasepage = nilfs_releasepage, */ |
| 250 | .invalidatepage = block_invalidatepage, |
| 251 | .direct_IO = nilfs_direct_IO, |
| 252 | }; |
| 253 | |
| 254 | struct inode *nilfs_new_inode(struct inode *dir, int mode) |
| 255 | { |
| 256 | struct super_block *sb = dir->i_sb; |
| 257 | struct nilfs_sb_info *sbi = NILFS_SB(sb); |
| 258 | struct inode *inode; |
| 259 | struct nilfs_inode_info *ii; |
| 260 | int err = -ENOMEM; |
| 261 | ino_t ino; |
| 262 | |
| 263 | inode = new_inode(sb); |
| 264 | if (unlikely(!inode)) |
| 265 | goto failed; |
| 266 | |
| 267 | mapping_set_gfp_mask(inode->i_mapping, |
| 268 | mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS); |
| 269 | |
| 270 | ii = NILFS_I(inode); |
| 271 | ii->i_state = 1 << NILFS_I_NEW; |
| 272 | |
| 273 | err = nilfs_ifile_create_inode(sbi->s_ifile, &ino, &ii->i_bh); |
| 274 | if (unlikely(err)) |
| 275 | goto failed_ifile_create_inode; |
| 276 | /* reference count of i_bh inherits from nilfs_mdt_read_block() */ |
| 277 | |
| 278 | atomic_inc(&sbi->s_inodes_count); |
| 279 | |
| 280 | inode->i_uid = current_fsuid(); |
| 281 | if (dir->i_mode & S_ISGID) { |
| 282 | inode->i_gid = dir->i_gid; |
| 283 | if (S_ISDIR(mode)) |
| 284 | mode |= S_ISGID; |
| 285 | } else |
| 286 | inode->i_gid = current_fsgid(); |
| 287 | |
| 288 | inode->i_mode = mode; |
| 289 | inode->i_ino = ino; |
| 290 | inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; |
| 291 | |
| 292 | if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)) { |
| 293 | err = nilfs_bmap_read(ii->i_bmap, NULL); |
| 294 | if (err < 0) |
| 295 | goto failed_bmap; |
| 296 | |
| 297 | set_bit(NILFS_I_BMAP, &ii->i_state); |
| 298 | /* No lock is needed; iget() ensures it. */ |
| 299 | } |
| 300 | |
| 301 | ii->i_flags = NILFS_I(dir)->i_flags; |
| 302 | if (S_ISLNK(mode)) |
| 303 | ii->i_flags &= ~(NILFS_IMMUTABLE_FL | NILFS_APPEND_FL); |
| 304 | if (!S_ISDIR(mode)) |
| 305 | ii->i_flags &= ~NILFS_DIRSYNC_FL; |
| 306 | |
| 307 | /* ii->i_file_acl = 0; */ |
| 308 | /* ii->i_dir_acl = 0; */ |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 309 | ii->i_dir_start_lookup = 0; |
| 310 | #ifdef CONFIG_NILFS_FS_POSIX_ACL |
| 311 | ii->i_acl = NULL; |
| 312 | ii->i_default_acl = NULL; |
| 313 | #endif |
| 314 | ii->i_cno = 0; |
| 315 | nilfs_set_inode_flags(inode); |
| 316 | spin_lock(&sbi->s_next_gen_lock); |
| 317 | inode->i_generation = sbi->s_next_generation++; |
| 318 | spin_unlock(&sbi->s_next_gen_lock); |
| 319 | insert_inode_hash(inode); |
| 320 | |
| 321 | err = nilfs_init_acl(inode, dir); |
| 322 | if (unlikely(err)) |
| 323 | goto failed_acl; /* never occur. When supporting |
| 324 | nilfs_init_acl(), proper cancellation of |
| 325 | above jobs should be considered */ |
| 326 | |
| 327 | mark_inode_dirty(inode); |
| 328 | return inode; |
| 329 | |
| 330 | failed_acl: |
| 331 | failed_bmap: |
| 332 | inode->i_nlink = 0; |
| 333 | iput(inode); /* raw_inode will be deleted through |
| 334 | generic_delete_inode() */ |
| 335 | goto failed; |
| 336 | |
| 337 | failed_ifile_create_inode: |
| 338 | make_bad_inode(inode); |
| 339 | iput(inode); /* if i_nlink == 1, generic_forget_inode() will be |
| 340 | called */ |
| 341 | failed: |
| 342 | return ERR_PTR(err); |
| 343 | } |
| 344 | |
| 345 | void nilfs_free_inode(struct inode *inode) |
| 346 | { |
| 347 | struct super_block *sb = inode->i_sb; |
| 348 | struct nilfs_sb_info *sbi = NILFS_SB(sb); |
| 349 | |
| 350 | clear_inode(inode); |
| 351 | /* XXX: check error code? Is there any thing I can do? */ |
| 352 | (void) nilfs_ifile_delete_inode(sbi->s_ifile, inode->i_ino); |
| 353 | atomic_dec(&sbi->s_inodes_count); |
| 354 | } |
| 355 | |
| 356 | void nilfs_set_inode_flags(struct inode *inode) |
| 357 | { |
| 358 | unsigned int flags = NILFS_I(inode)->i_flags; |
| 359 | |
| 360 | inode->i_flags &= ~(S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | |
| 361 | S_DIRSYNC); |
| 362 | if (flags & NILFS_SYNC_FL) |
| 363 | inode->i_flags |= S_SYNC; |
| 364 | if (flags & NILFS_APPEND_FL) |
| 365 | inode->i_flags |= S_APPEND; |
| 366 | if (flags & NILFS_IMMUTABLE_FL) |
| 367 | inode->i_flags |= S_IMMUTABLE; |
| 368 | #ifndef NILFS_ATIME_DISABLE |
| 369 | if (flags & NILFS_NOATIME_FL) |
| 370 | #endif |
| 371 | inode->i_flags |= S_NOATIME; |
| 372 | if (flags & NILFS_DIRSYNC_FL) |
| 373 | inode->i_flags |= S_DIRSYNC; |
| 374 | mapping_set_gfp_mask(inode->i_mapping, |
| 375 | mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS); |
| 376 | } |
| 377 | |
| 378 | int nilfs_read_inode_common(struct inode *inode, |
| 379 | struct nilfs_inode *raw_inode) |
| 380 | { |
| 381 | struct nilfs_inode_info *ii = NILFS_I(inode); |
| 382 | int err; |
| 383 | |
| 384 | inode->i_mode = le16_to_cpu(raw_inode->i_mode); |
| 385 | inode->i_uid = (uid_t)le32_to_cpu(raw_inode->i_uid); |
| 386 | inode->i_gid = (gid_t)le32_to_cpu(raw_inode->i_gid); |
| 387 | inode->i_nlink = le16_to_cpu(raw_inode->i_links_count); |
| 388 | inode->i_size = le64_to_cpu(raw_inode->i_size); |
| 389 | inode->i_atime.tv_sec = le64_to_cpu(raw_inode->i_mtime); |
| 390 | inode->i_ctime.tv_sec = le64_to_cpu(raw_inode->i_ctime); |
| 391 | inode->i_mtime.tv_sec = le64_to_cpu(raw_inode->i_mtime); |
Ryusuke Konishi | 6123923 | 2009-04-06 19:02:00 -0700 | [diff] [blame] | 392 | inode->i_atime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); |
| 393 | inode->i_ctime.tv_nsec = le32_to_cpu(raw_inode->i_ctime_nsec); |
| 394 | inode->i_mtime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec); |
| 395 | if (inode->i_nlink == 0 && inode->i_mode == 0) |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 396 | return -EINVAL; /* this inode is deleted */ |
| 397 | |
| 398 | inode->i_blocks = le64_to_cpu(raw_inode->i_blocks); |
| 399 | ii->i_flags = le32_to_cpu(raw_inode->i_flags); |
| 400 | #if 0 |
| 401 | ii->i_file_acl = le32_to_cpu(raw_inode->i_file_acl); |
| 402 | ii->i_dir_acl = S_ISREG(inode->i_mode) ? |
| 403 | 0 : le32_to_cpu(raw_inode->i_dir_acl); |
| 404 | #endif |
| 405 | ii->i_cno = 0; |
| 406 | inode->i_generation = le32_to_cpu(raw_inode->i_generation); |
| 407 | |
| 408 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || |
| 409 | S_ISLNK(inode->i_mode)) { |
| 410 | err = nilfs_bmap_read(ii->i_bmap, raw_inode); |
| 411 | if (err < 0) |
| 412 | return err; |
| 413 | set_bit(NILFS_I_BMAP, &ii->i_state); |
| 414 | /* No lock is needed; iget() ensures it. */ |
| 415 | } |
| 416 | return 0; |
| 417 | } |
| 418 | |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 419 | static int __nilfs_read_inode(struct super_block *sb, unsigned long ino, |
| 420 | struct inode *inode) |
| 421 | { |
| 422 | struct nilfs_sb_info *sbi = NILFS_SB(sb); |
| 423 | struct inode *dat = nilfs_dat_inode(sbi->s_nilfs); |
| 424 | struct buffer_head *bh; |
| 425 | struct nilfs_inode *raw_inode; |
| 426 | int err; |
| 427 | |
| 428 | down_read(&NILFS_MDT(dat)->mi_sem); /* XXX */ |
| 429 | err = nilfs_ifile_get_inode_block(sbi->s_ifile, ino, &bh); |
| 430 | if (unlikely(err)) |
| 431 | goto bad_inode; |
| 432 | |
| 433 | raw_inode = nilfs_ifile_map_inode(sbi->s_ifile, ino, bh); |
| 434 | |
| 435 | #ifdef CONFIG_NILFS_FS_POSIX_ACL |
| 436 | ii->i_acl = NILFS_ACL_NOT_CACHED; |
| 437 | ii->i_default_acl = NILFS_ACL_NOT_CACHED; |
| 438 | #endif |
| 439 | if (nilfs_read_inode_common(inode, raw_inode)) |
| 440 | goto failed_unmap; |
| 441 | |
| 442 | if (S_ISREG(inode->i_mode)) { |
| 443 | inode->i_op = &nilfs_file_inode_operations; |
| 444 | inode->i_fop = &nilfs_file_operations; |
| 445 | inode->i_mapping->a_ops = &nilfs_aops; |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 446 | } else if (S_ISDIR(inode->i_mode)) { |
| 447 | inode->i_op = &nilfs_dir_inode_operations; |
| 448 | inode->i_fop = &nilfs_dir_operations; |
| 449 | inode->i_mapping->a_ops = &nilfs_aops; |
| 450 | } else if (S_ISLNK(inode->i_mode)) { |
| 451 | inode->i_op = &nilfs_symlink_inode_operations; |
| 452 | inode->i_mapping->a_ops = &nilfs_aops; |
| 453 | } else { |
| 454 | inode->i_op = &nilfs_special_inode_operations; |
| 455 | init_special_inode( |
| 456 | inode, inode->i_mode, |
| 457 | new_decode_dev(le64_to_cpu(raw_inode->i_device_code))); |
| 458 | } |
| 459 | nilfs_ifile_unmap_inode(sbi->s_ifile, ino, bh); |
| 460 | brelse(bh); |
| 461 | up_read(&NILFS_MDT(dat)->mi_sem); /* XXX */ |
| 462 | nilfs_set_inode_flags(inode); |
| 463 | return 0; |
| 464 | |
| 465 | failed_unmap: |
| 466 | nilfs_ifile_unmap_inode(sbi->s_ifile, ino, bh); |
| 467 | brelse(bh); |
| 468 | |
| 469 | bad_inode: |
| 470 | up_read(&NILFS_MDT(dat)->mi_sem); /* XXX */ |
| 471 | return err; |
| 472 | } |
| 473 | |
| 474 | struct inode *nilfs_iget(struct super_block *sb, unsigned long ino) |
| 475 | { |
| 476 | struct inode *inode; |
| 477 | int err; |
| 478 | |
| 479 | inode = iget_locked(sb, ino); |
| 480 | if (unlikely(!inode)) |
| 481 | return ERR_PTR(-ENOMEM); |
| 482 | if (!(inode->i_state & I_NEW)) |
| 483 | return inode; |
| 484 | |
| 485 | err = __nilfs_read_inode(sb, ino, inode); |
| 486 | if (unlikely(err)) { |
| 487 | iget_failed(inode); |
| 488 | return ERR_PTR(err); |
| 489 | } |
| 490 | unlock_new_inode(inode); |
| 491 | return inode; |
| 492 | } |
| 493 | |
| 494 | void nilfs_write_inode_common(struct inode *inode, |
| 495 | struct nilfs_inode *raw_inode, int has_bmap) |
| 496 | { |
| 497 | struct nilfs_inode_info *ii = NILFS_I(inode); |
| 498 | |
| 499 | raw_inode->i_mode = cpu_to_le16(inode->i_mode); |
| 500 | raw_inode->i_uid = cpu_to_le32(inode->i_uid); |
| 501 | raw_inode->i_gid = cpu_to_le32(inode->i_gid); |
| 502 | raw_inode->i_links_count = cpu_to_le16(inode->i_nlink); |
| 503 | raw_inode->i_size = cpu_to_le64(inode->i_size); |
| 504 | raw_inode->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec); |
| 505 | raw_inode->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec); |
Ryusuke Konishi | 6123923 | 2009-04-06 19:02:00 -0700 | [diff] [blame] | 506 | raw_inode->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); |
| 507 | raw_inode->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec); |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 508 | raw_inode->i_blocks = cpu_to_le64(inode->i_blocks); |
| 509 | |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 510 | raw_inode->i_flags = cpu_to_le32(ii->i_flags); |
| 511 | raw_inode->i_generation = cpu_to_le32(inode->i_generation); |
| 512 | |
| 513 | if (has_bmap) |
| 514 | nilfs_bmap_write(ii->i_bmap, raw_inode); |
| 515 | else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) |
| 516 | raw_inode->i_device_code = |
| 517 | cpu_to_le64(new_encode_dev(inode->i_rdev)); |
| 518 | /* When extending inode, nilfs->ns_inode_size should be checked |
| 519 | for substitutions of appended fields */ |
| 520 | } |
| 521 | |
| 522 | void nilfs_update_inode(struct inode *inode, struct buffer_head *ibh) |
| 523 | { |
| 524 | ino_t ino = inode->i_ino; |
| 525 | struct nilfs_inode_info *ii = NILFS_I(inode); |
| 526 | struct super_block *sb = inode->i_sb; |
| 527 | struct nilfs_sb_info *sbi = NILFS_SB(sb); |
| 528 | struct nilfs_inode *raw_inode; |
| 529 | |
| 530 | raw_inode = nilfs_ifile_map_inode(sbi->s_ifile, ino, ibh); |
| 531 | |
| 532 | /* The buffer is guarded with lock_buffer() by the caller */ |
| 533 | if (test_and_clear_bit(NILFS_I_NEW, &ii->i_state)) |
| 534 | memset(raw_inode, 0, NILFS_MDT(sbi->s_ifile)->mi_entry_size); |
| 535 | set_bit(NILFS_I_INODE_DIRTY, &ii->i_state); |
| 536 | |
| 537 | nilfs_write_inode_common(inode, raw_inode, 0); |
| 538 | /* XXX: call with has_bmap = 0 is a workaround to avoid |
| 539 | deadlock of bmap. This delays update of i_bmap to just |
| 540 | before writing */ |
| 541 | nilfs_ifile_unmap_inode(sbi->s_ifile, ino, ibh); |
| 542 | } |
| 543 | |
| 544 | #define NILFS_MAX_TRUNCATE_BLOCKS 16384 /* 64MB for 4KB block */ |
| 545 | |
| 546 | static void nilfs_truncate_bmap(struct nilfs_inode_info *ii, |
| 547 | unsigned long from) |
| 548 | { |
| 549 | unsigned long b; |
| 550 | int ret; |
| 551 | |
| 552 | if (!test_bit(NILFS_I_BMAP, &ii->i_state)) |
| 553 | return; |
| 554 | repeat: |
| 555 | ret = nilfs_bmap_last_key(ii->i_bmap, &b); |
| 556 | if (ret == -ENOENT) |
| 557 | return; |
| 558 | else if (ret < 0) |
| 559 | goto failed; |
| 560 | |
| 561 | if (b < from) |
| 562 | return; |
| 563 | |
| 564 | b -= min_t(unsigned long, NILFS_MAX_TRUNCATE_BLOCKS, b - from); |
| 565 | ret = nilfs_bmap_truncate(ii->i_bmap, b); |
| 566 | nilfs_relax_pressure_in_lock(ii->vfs_inode.i_sb); |
| 567 | if (!ret || (ret == -ENOMEM && |
| 568 | nilfs_bmap_truncate(ii->i_bmap, b) == 0)) |
| 569 | goto repeat; |
| 570 | |
| 571 | failed: |
| 572 | if (ret == -EINVAL) |
| 573 | nilfs_error(ii->vfs_inode.i_sb, __func__, |
| 574 | "bmap is broken (ino=%lu)", ii->vfs_inode.i_ino); |
| 575 | else |
| 576 | nilfs_warning(ii->vfs_inode.i_sb, __func__, |
| 577 | "failed to truncate bmap (ino=%lu, err=%d)", |
| 578 | ii->vfs_inode.i_ino, ret); |
| 579 | } |
| 580 | |
| 581 | void nilfs_truncate(struct inode *inode) |
| 582 | { |
| 583 | unsigned long blkoff; |
| 584 | unsigned int blocksize; |
| 585 | struct nilfs_transaction_info ti; |
| 586 | struct super_block *sb = inode->i_sb; |
| 587 | struct nilfs_inode_info *ii = NILFS_I(inode); |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 588 | |
| 589 | if (!test_bit(NILFS_I_BMAP, &ii->i_state)) |
| 590 | return; |
| 591 | if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) |
| 592 | return; |
| 593 | |
| 594 | blocksize = sb->s_blocksize; |
| 595 | blkoff = (inode->i_size + blocksize - 1) >> sb->s_blocksize_bits; |
Ryusuke Konishi | 1f5abe7 | 2009-04-06 19:01:55 -0700 | [diff] [blame] | 596 | nilfs_transaction_begin(sb, &ti, 0); /* never fails */ |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 597 | |
| 598 | block_truncate_page(inode->i_mapping, inode->i_size, nilfs_get_block); |
| 599 | |
| 600 | nilfs_truncate_bmap(ii, blkoff); |
| 601 | |
| 602 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; |
| 603 | if (IS_SYNC(inode)) |
| 604 | nilfs_set_transaction_flag(NILFS_TI_SYNC); |
| 605 | |
| 606 | nilfs_set_file_dirty(NILFS_SB(sb), inode, 0); |
Ryusuke Konishi | 47420c7 | 2009-04-06 19:01:45 -0700 | [diff] [blame] | 607 | nilfs_transaction_commit(sb); |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 608 | /* May construct a logical segment and may fail in sync mode. |
| 609 | But truncate has no return value. */ |
| 610 | } |
| 611 | |
| 612 | void nilfs_delete_inode(struct inode *inode) |
| 613 | { |
| 614 | struct nilfs_transaction_info ti; |
| 615 | struct super_block *sb = inode->i_sb; |
| 616 | struct nilfs_inode_info *ii = NILFS_I(inode); |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 617 | |
| 618 | if (unlikely(is_bad_inode(inode))) { |
| 619 | if (inode->i_data.nrpages) |
| 620 | truncate_inode_pages(&inode->i_data, 0); |
| 621 | clear_inode(inode); |
| 622 | return; |
| 623 | } |
Ryusuke Konishi | 1f5abe7 | 2009-04-06 19:01:55 -0700 | [diff] [blame] | 624 | nilfs_transaction_begin(sb, &ti, 0); /* never fails */ |
| 625 | |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 626 | if (inode->i_data.nrpages) |
| 627 | truncate_inode_pages(&inode->i_data, 0); |
| 628 | |
| 629 | nilfs_truncate_bmap(ii, 0); |
| 630 | nilfs_free_inode(inode); |
| 631 | /* nilfs_free_inode() marks inode buffer dirty */ |
| 632 | if (IS_SYNC(inode)) |
| 633 | nilfs_set_transaction_flag(NILFS_TI_SYNC); |
Ryusuke Konishi | 47420c7 | 2009-04-06 19:01:45 -0700 | [diff] [blame] | 634 | nilfs_transaction_commit(sb); |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 635 | /* May construct a logical segment and may fail in sync mode. |
| 636 | But delete_inode has no return value. */ |
| 637 | } |
| 638 | |
| 639 | int nilfs_setattr(struct dentry *dentry, struct iattr *iattr) |
| 640 | { |
| 641 | struct nilfs_transaction_info ti; |
| 642 | struct inode *inode = dentry->d_inode; |
| 643 | struct super_block *sb = inode->i_sb; |
Ryusuke Konishi | 47420c7 | 2009-04-06 19:01:45 -0700 | [diff] [blame] | 644 | int err; |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 645 | |
| 646 | err = inode_change_ok(inode, iattr); |
| 647 | if (err) |
| 648 | return err; |
| 649 | |
| 650 | err = nilfs_transaction_begin(sb, &ti, 0); |
| 651 | if (unlikely(err)) |
| 652 | return err; |
| 653 | err = inode_setattr(inode, iattr); |
| 654 | if (!err && (iattr->ia_valid & ATTR_MODE)) |
| 655 | err = nilfs_acl_chmod(inode); |
Ryusuke Konishi | 47420c7 | 2009-04-06 19:01:45 -0700 | [diff] [blame] | 656 | if (likely(!err)) |
| 657 | err = nilfs_transaction_commit(sb); |
| 658 | else |
| 659 | nilfs_transaction_abort(sb); |
| 660 | |
| 661 | return err; |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 662 | } |
| 663 | |
| 664 | int nilfs_load_inode_block(struct nilfs_sb_info *sbi, struct inode *inode, |
| 665 | struct buffer_head **pbh) |
| 666 | { |
| 667 | struct nilfs_inode_info *ii = NILFS_I(inode); |
| 668 | int err; |
| 669 | |
| 670 | spin_lock(&sbi->s_inode_lock); |
| 671 | /* Caller of this function MUST lock s_inode_lock */ |
| 672 | if (ii->i_bh == NULL) { |
| 673 | spin_unlock(&sbi->s_inode_lock); |
| 674 | err = nilfs_ifile_get_inode_block(sbi->s_ifile, inode->i_ino, |
| 675 | pbh); |
| 676 | if (unlikely(err)) |
| 677 | return err; |
| 678 | spin_lock(&sbi->s_inode_lock); |
| 679 | if (ii->i_bh == NULL) |
| 680 | ii->i_bh = *pbh; |
| 681 | else { |
| 682 | brelse(*pbh); |
| 683 | *pbh = ii->i_bh; |
| 684 | } |
| 685 | } else |
| 686 | *pbh = ii->i_bh; |
| 687 | |
| 688 | get_bh(*pbh); |
| 689 | spin_unlock(&sbi->s_inode_lock); |
| 690 | return 0; |
| 691 | } |
| 692 | |
| 693 | int nilfs_inode_dirty(struct inode *inode) |
| 694 | { |
| 695 | struct nilfs_inode_info *ii = NILFS_I(inode); |
| 696 | struct nilfs_sb_info *sbi = NILFS_SB(inode->i_sb); |
| 697 | int ret = 0; |
| 698 | |
| 699 | if (!list_empty(&ii->i_dirty)) { |
| 700 | spin_lock(&sbi->s_inode_lock); |
| 701 | ret = test_bit(NILFS_I_DIRTY, &ii->i_state) || |
| 702 | test_bit(NILFS_I_BUSY, &ii->i_state); |
| 703 | spin_unlock(&sbi->s_inode_lock); |
| 704 | } |
| 705 | return ret; |
| 706 | } |
| 707 | |
| 708 | int nilfs_set_file_dirty(struct nilfs_sb_info *sbi, struct inode *inode, |
| 709 | unsigned nr_dirty) |
| 710 | { |
| 711 | struct nilfs_inode_info *ii = NILFS_I(inode); |
| 712 | |
| 713 | atomic_add(nr_dirty, &sbi->s_nilfs->ns_ndirtyblks); |
| 714 | |
Ryusuke Konishi | 458c5b0 | 2009-04-06 19:01:56 -0700 | [diff] [blame] | 715 | if (test_and_set_bit(NILFS_I_DIRTY, &ii->i_state)) |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 716 | return 0; |
| 717 | |
| 718 | spin_lock(&sbi->s_inode_lock); |
| 719 | if (!test_bit(NILFS_I_QUEUED, &ii->i_state) && |
| 720 | !test_bit(NILFS_I_BUSY, &ii->i_state)) { |
| 721 | /* Because this routine may race with nilfs_dispose_list(), |
| 722 | we have to check NILFS_I_QUEUED here, too. */ |
| 723 | if (list_empty(&ii->i_dirty) && igrab(inode) == NULL) { |
| 724 | /* This will happen when somebody is freeing |
| 725 | this inode. */ |
| 726 | nilfs_warning(sbi->s_super, __func__, |
| 727 | "cannot get inode (ino=%lu)\n", |
| 728 | inode->i_ino); |
| 729 | spin_unlock(&sbi->s_inode_lock); |
| 730 | return -EINVAL; /* NILFS_I_DIRTY may remain for |
| 731 | freeing inode */ |
| 732 | } |
| 733 | list_del(&ii->i_dirty); |
| 734 | list_add_tail(&ii->i_dirty, &sbi->s_dirty_files); |
| 735 | set_bit(NILFS_I_QUEUED, &ii->i_state); |
| 736 | } |
| 737 | spin_unlock(&sbi->s_inode_lock); |
| 738 | return 0; |
| 739 | } |
| 740 | |
| 741 | int nilfs_mark_inode_dirty(struct inode *inode) |
| 742 | { |
| 743 | struct nilfs_sb_info *sbi = NILFS_SB(inode->i_sb); |
| 744 | struct buffer_head *ibh; |
| 745 | int err; |
| 746 | |
| 747 | err = nilfs_load_inode_block(sbi, inode, &ibh); |
| 748 | if (unlikely(err)) { |
| 749 | nilfs_warning(inode->i_sb, __func__, |
| 750 | "failed to reget inode block.\n"); |
| 751 | return err; |
| 752 | } |
| 753 | lock_buffer(ibh); |
| 754 | nilfs_update_inode(inode, ibh); |
| 755 | unlock_buffer(ibh); |
| 756 | nilfs_mdt_mark_buffer_dirty(ibh); |
| 757 | nilfs_mdt_mark_dirty(sbi->s_ifile); |
| 758 | brelse(ibh); |
| 759 | return 0; |
| 760 | } |
| 761 | |
| 762 | /** |
| 763 | * nilfs_dirty_inode - reflect changes on given inode to an inode block. |
| 764 | * @inode: inode of the file to be registered. |
| 765 | * |
| 766 | * nilfs_dirty_inode() loads a inode block containing the specified |
| 767 | * @inode and copies data from a nilfs_inode to a corresponding inode |
| 768 | * entry in the inode block. This operation is excluded from the segment |
| 769 | * construction. This function can be called both as a single operation |
| 770 | * and as a part of indivisible file operations. |
| 771 | */ |
| 772 | void nilfs_dirty_inode(struct inode *inode) |
| 773 | { |
| 774 | struct nilfs_transaction_info ti; |
| 775 | |
| 776 | if (is_bad_inode(inode)) { |
| 777 | nilfs_warning(inode->i_sb, __func__, |
| 778 | "tried to mark bad_inode dirty. ignored.\n"); |
| 779 | dump_stack(); |
| 780 | return; |
| 781 | } |
| 782 | nilfs_transaction_begin(inode->i_sb, &ti, 0); |
Ryusuke Konishi | 458c5b0 | 2009-04-06 19:01:56 -0700 | [diff] [blame] | 783 | nilfs_mark_inode_dirty(inode); |
Ryusuke Konishi | 47420c7 | 2009-04-06 19:01:45 -0700 | [diff] [blame] | 784 | nilfs_transaction_commit(inode->i_sb); /* never fails */ |
Ryusuke Konishi | 05fe58f | 2009-04-06 19:01:32 -0700 | [diff] [blame] | 785 | } |