1 /* 2 * fs/f2fs/inode.c 3 * 4 * Copyright (c) 2012 Samsung Electronics Co., Ltd. 5 * http://www.samsung.com/ 6 * 7 * This program is free software; you can redistribute it and/or modify 8 * it under the terms of the GNU General Public License version 2 as 9 * published by the Free Software Foundation. 10 */ 11 #include <linux/fs.h> 12 #include <linux/f2fs_fs.h> 13 #include <linux/buffer_head.h> 14 #include <linux/backing-dev.h> 15 #include <linux/writeback.h> 16 17 #include "f2fs.h" 18 #include "node.h" 19 20 #include <trace/events/f2fs.h> 21 22 void f2fs_mark_inode_dirty_sync(struct inode *inode) 23 { 24 if (f2fs_inode_dirtied(inode)) 25 return; 26 mark_inode_dirty_sync(inode); 27 } 28 29 void f2fs_set_inode_flags(struct inode *inode) 30 { 31 unsigned int flags = F2FS_I(inode)->i_flags; 32 unsigned int new_fl = 0; 33 34 if (flags & FS_SYNC_FL) 35 new_fl |= S_SYNC; 36 if (flags & FS_APPEND_FL) 37 new_fl |= S_APPEND; 38 if (flags & FS_IMMUTABLE_FL) 39 new_fl |= S_IMMUTABLE; 40 if (flags & FS_NOATIME_FL) 41 new_fl |= S_NOATIME; 42 if (flags & FS_DIRSYNC_FL) 43 new_fl |= S_DIRSYNC; 44 inode_set_flags(inode, new_fl, 45 S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC); 46 f2fs_mark_inode_dirty_sync(inode); 47 } 48 49 static void __get_inode_rdev(struct inode *inode, struct f2fs_inode *ri) 50 { 51 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) || 52 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) { 53 if (ri->i_addr[0]) 54 inode->i_rdev = 55 old_decode_dev(le32_to_cpu(ri->i_addr[0])); 56 else 57 inode->i_rdev = 58 new_decode_dev(le32_to_cpu(ri->i_addr[1])); 59 } 60 } 61 62 static bool __written_first_block(struct f2fs_inode *ri) 63 { 64 block_t addr = le32_to_cpu(ri->i_addr[0]); 65 66 if (addr != NEW_ADDR && addr != NULL_ADDR) 67 return true; 68 return false; 69 } 70 71 static void __set_inode_rdev(struct inode *inode, struct f2fs_inode *ri) 72 { 73 if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) { 74 if (old_valid_dev(inode->i_rdev)) { 75 ri->i_addr[0] = 76 cpu_to_le32(old_encode_dev(inode->i_rdev)); 77 ri->i_addr[1] = 0; 78 } else { 79 ri->i_addr[0] = 0; 80 ri->i_addr[1] = 81 cpu_to_le32(new_encode_dev(inode->i_rdev)); 82 ri->i_addr[2] = 0; 83 } 84 } 85 } 86 87 static void __recover_inline_status(struct inode *inode, struct page *ipage) 88 { 89 void *inline_data = inline_data_addr(ipage); 90 __le32 *start = inline_data; 91 __le32 *end = start + MAX_INLINE_DATA / sizeof(__le32); 92 93 while (start < end) { 94 if (*start++) { 95 f2fs_wait_on_page_writeback(ipage, NODE, true); 96 97 set_inode_flag(inode, FI_DATA_EXIST); 98 set_raw_inline(inode, F2FS_INODE(ipage)); 99 set_page_dirty(ipage); 100 return; 101 } 102 } 103 return; 104 } 105 106 static int do_read_inode(struct inode *inode) 107 { 108 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 109 struct f2fs_inode_info *fi = F2FS_I(inode); 110 struct page *node_page; 111 struct f2fs_inode *ri; 112 113 /* Check if ino is within scope */ 114 if (check_nid_range(sbi, inode->i_ino)) { 115 f2fs_msg(inode->i_sb, KERN_ERR, "bad inode number: %lu", 116 (unsigned long) inode->i_ino); 117 WARN_ON(1); 118 return -EINVAL; 119 } 120 121 node_page = get_node_page(sbi, inode->i_ino); 122 if (IS_ERR(node_page)) 123 return PTR_ERR(node_page); 124 125 ri = F2FS_INODE(node_page); 126 127 inode->i_mode = le16_to_cpu(ri->i_mode); 128 i_uid_write(inode, le32_to_cpu(ri->i_uid)); 129 i_gid_write(inode, le32_to_cpu(ri->i_gid)); 130 set_nlink(inode, le32_to_cpu(ri->i_links)); 131 inode->i_size = le64_to_cpu(ri->i_size); 132 inode->i_blocks = le64_to_cpu(ri->i_blocks); 133 134 inode->i_atime.tv_sec = le64_to_cpu(ri->i_atime); 135 inode->i_ctime.tv_sec = le64_to_cpu(ri->i_ctime); 136 inode->i_mtime.tv_sec = le64_to_cpu(ri->i_mtime); 137 inode->i_atime.tv_nsec = le32_to_cpu(ri->i_atime_nsec); 138 inode->i_ctime.tv_nsec = le32_to_cpu(ri->i_ctime_nsec); 139 inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec); 140 inode->i_generation = le32_to_cpu(ri->i_generation); 141 142 fi->i_current_depth = le32_to_cpu(ri->i_current_depth); 143 fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid); 144 fi->i_flags = le32_to_cpu(ri->i_flags); 145 fi->flags = 0; 146 fi->i_advise = ri->i_advise; 147 fi->i_pino = le32_to_cpu(ri->i_pino); 148 fi->i_dir_level = ri->i_dir_level; 149 150 if (f2fs_init_extent_tree(inode, &ri->i_ext)) 151 set_page_dirty(node_page); 152 153 get_inline_info(inode, ri); 154 155 /* check data exist */ 156 if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode)) 157 __recover_inline_status(inode, node_page); 158 159 /* get rdev by using inline_info */ 160 __get_inode_rdev(inode, ri); 161 162 if (__written_first_block(ri)) 163 set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN); 164 165 if (!need_inode_block_update(sbi, inode->i_ino)) 166 fi->last_disk_size = inode->i_size; 167 168 f2fs_put_page(node_page, 1); 169 170 stat_inc_inline_xattr(inode); 171 stat_inc_inline_inode(inode); 172 stat_inc_inline_dir(inode); 173 174 return 0; 175 } 176 177 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino) 178 { 179 struct f2fs_sb_info *sbi = F2FS_SB(sb); 180 struct inode *inode; 181 int ret = 0; 182 183 inode = iget_locked(sb, ino); 184 if (!inode) 185 return ERR_PTR(-ENOMEM); 186 187 if (!(inode->i_state & I_NEW)) { 188 trace_f2fs_iget(inode); 189 return inode; 190 } 191 if (ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi)) 192 goto make_now; 193 194 ret = do_read_inode(inode); 195 if (ret) 196 goto bad_inode; 197 make_now: 198 if (ino == F2FS_NODE_INO(sbi)) { 199 inode->i_mapping->a_ops = &f2fs_node_aops; 200 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO); 201 } else if (ino == F2FS_META_INO(sbi)) { 202 inode->i_mapping->a_ops = &f2fs_meta_aops; 203 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO); 204 } else if (S_ISREG(inode->i_mode)) { 205 inode->i_op = &f2fs_file_inode_operations; 206 inode->i_fop = &f2fs_file_operations; 207 inode->i_mapping->a_ops = &f2fs_dblock_aops; 208 } else if (S_ISDIR(inode->i_mode)) { 209 inode->i_op = &f2fs_dir_inode_operations; 210 inode->i_fop = &f2fs_dir_operations; 211 inode->i_mapping->a_ops = &f2fs_dblock_aops; 212 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_HIGH_ZERO); 213 } else if (S_ISLNK(inode->i_mode)) { 214 if (f2fs_encrypted_inode(inode)) 215 inode->i_op = &f2fs_encrypted_symlink_inode_operations; 216 else 217 inode->i_op = &f2fs_symlink_inode_operations; 218 inode_nohighmem(inode); 219 inode->i_mapping->a_ops = &f2fs_dblock_aops; 220 } else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) || 221 S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) { 222 inode->i_op = &f2fs_special_inode_operations; 223 init_special_inode(inode, inode->i_mode, inode->i_rdev); 224 } else { 225 ret = -EIO; 226 goto bad_inode; 227 } 228 unlock_new_inode(inode); 229 trace_f2fs_iget(inode); 230 return inode; 231 232 bad_inode: 233 iget_failed(inode); 234 trace_f2fs_iget_exit(inode, ret); 235 return ERR_PTR(ret); 236 } 237 238 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino) 239 { 240 struct inode *inode; 241 retry: 242 inode = f2fs_iget(sb, ino); 243 if (IS_ERR(inode)) { 244 if (PTR_ERR(inode) == -ENOMEM) { 245 congestion_wait(BLK_RW_ASYNC, HZ/50); 246 goto retry; 247 } 248 } 249 return inode; 250 } 251 252 int update_inode(struct inode *inode, struct page *node_page) 253 { 254 struct f2fs_inode *ri; 255 struct extent_tree *et = F2FS_I(inode)->extent_tree; 256 257 f2fs_inode_synced(inode); 258 259 f2fs_wait_on_page_writeback(node_page, NODE, true); 260 261 ri = F2FS_INODE(node_page); 262 263 ri->i_mode = cpu_to_le16(inode->i_mode); 264 ri->i_advise = F2FS_I(inode)->i_advise; 265 ri->i_uid = cpu_to_le32(i_uid_read(inode)); 266 ri->i_gid = cpu_to_le32(i_gid_read(inode)); 267 ri->i_links = cpu_to_le32(inode->i_nlink); 268 ri->i_size = cpu_to_le64(i_size_read(inode)); 269 ri->i_blocks = cpu_to_le64(inode->i_blocks); 270 271 if (et) { 272 read_lock(&et->lock); 273 set_raw_extent(&et->largest, &ri->i_ext); 274 read_unlock(&et->lock); 275 } else { 276 memset(&ri->i_ext, 0, sizeof(ri->i_ext)); 277 } 278 set_raw_inline(inode, ri); 279 280 ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec); 281 ri->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec); 282 ri->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec); 283 ri->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec); 284 ri->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); 285 ri->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec); 286 ri->i_current_depth = cpu_to_le32(F2FS_I(inode)->i_current_depth); 287 ri->i_xattr_nid = cpu_to_le32(F2FS_I(inode)->i_xattr_nid); 288 ri->i_flags = cpu_to_le32(F2FS_I(inode)->i_flags); 289 ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino); 290 ri->i_generation = cpu_to_le32(inode->i_generation); 291 ri->i_dir_level = F2FS_I(inode)->i_dir_level; 292 293 __set_inode_rdev(inode, ri); 294 set_cold_node(inode, node_page); 295 296 /* deleted inode */ 297 if (inode->i_nlink == 0) 298 clear_inline_node(node_page); 299 300 return set_page_dirty(node_page); 301 } 302 303 int update_inode_page(struct inode *inode) 304 { 305 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 306 struct page *node_page; 307 int ret = 0; 308 retry: 309 node_page = get_node_page(sbi, inode->i_ino); 310 if (IS_ERR(node_page)) { 311 int err = PTR_ERR(node_page); 312 if (err == -ENOMEM) { 313 cond_resched(); 314 goto retry; 315 } else if (err != -ENOENT) { 316 f2fs_stop_checkpoint(sbi, false); 317 } 318 f2fs_inode_synced(inode); 319 return 0; 320 } 321 ret = update_inode(inode, node_page); 322 f2fs_put_page(node_page, 1); 323 return ret; 324 } 325 326 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc) 327 { 328 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 329 330 if (inode->i_ino == F2FS_NODE_INO(sbi) || 331 inode->i_ino == F2FS_META_INO(sbi)) 332 return 0; 333 334 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) 335 return 0; 336 337 /* 338 * We need to balance fs here to prevent from producing dirty node pages 339 * during the urgent cleaning time when runing out of free sections. 340 */ 341 if (update_inode_page(inode)) 342 f2fs_balance_fs(sbi, true); 343 return 0; 344 } 345 346 /* 347 * Called at the last iput() if i_nlink is zero 348 */ 349 void f2fs_evict_inode(struct inode *inode) 350 { 351 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 352 nid_t xnid = F2FS_I(inode)->i_xattr_nid; 353 int err = 0; 354 355 /* some remained atomic pages should discarded */ 356 if (f2fs_is_atomic_file(inode)) 357 drop_inmem_pages(inode); 358 359 trace_f2fs_evict_inode(inode); 360 truncate_inode_pages_final(&inode->i_data); 361 362 if (inode->i_ino == F2FS_NODE_INO(sbi) || 363 inode->i_ino == F2FS_META_INO(sbi)) 364 goto out_clear; 365 366 f2fs_bug_on(sbi, get_dirty_pages(inode)); 367 remove_dirty_inode(inode); 368 369 f2fs_destroy_extent_tree(inode); 370 371 if (inode->i_nlink || is_bad_inode(inode)) 372 goto no_delete; 373 374 #ifdef CONFIG_F2FS_FAULT_INJECTION 375 if (time_to_inject(sbi, FAULT_EVICT_INODE)) 376 goto no_delete; 377 #endif 378 379 sb_start_intwrite(inode->i_sb); 380 set_inode_flag(inode, FI_NO_ALLOC); 381 i_size_write(inode, 0); 382 retry: 383 if (F2FS_HAS_BLOCKS(inode)) 384 err = f2fs_truncate(inode); 385 386 if (!err) { 387 f2fs_lock_op(sbi); 388 err = remove_inode_page(inode); 389 f2fs_unlock_op(sbi); 390 if (err == -ENOENT) 391 err = 0; 392 } 393 394 /* give more chances, if ENOMEM case */ 395 if (err == -ENOMEM) { 396 err = 0; 397 goto retry; 398 } 399 400 if (err) 401 update_inode_page(inode); 402 sb_end_intwrite(inode->i_sb); 403 no_delete: 404 stat_dec_inline_xattr(inode); 405 stat_dec_inline_dir(inode); 406 stat_dec_inline_inode(inode); 407 408 invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino, inode->i_ino); 409 if (xnid) 410 invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid); 411 if (is_inode_flag_set(inode, FI_APPEND_WRITE)) 412 add_ino_entry(sbi, inode->i_ino, APPEND_INO); 413 if (is_inode_flag_set(inode, FI_UPDATE_WRITE)) 414 add_ino_entry(sbi, inode->i_ino, UPDATE_INO); 415 if (is_inode_flag_set(inode, FI_FREE_NID)) { 416 alloc_nid_failed(sbi, inode->i_ino); 417 clear_inode_flag(inode, FI_FREE_NID); 418 } 419 f2fs_bug_on(sbi, err && 420 !exist_written_data(sbi, inode->i_ino, ORPHAN_INO)); 421 out_clear: 422 fscrypt_put_encryption_info(inode, NULL); 423 clear_inode(inode); 424 } 425 426 /* caller should call f2fs_lock_op() */ 427 void handle_failed_inode(struct inode *inode) 428 { 429 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 430 struct node_info ni; 431 432 /* 433 * clear nlink of inode in order to release resource of inode 434 * immediately. 435 */ 436 clear_nlink(inode); 437 438 /* 439 * we must call this to avoid inode being remained as dirty, resulting 440 * in a panic when flushing dirty inodes in gdirty_list. 441 */ 442 update_inode_page(inode); 443 444 /* don't make bad inode, since it becomes a regular file. */ 445 unlock_new_inode(inode); 446 447 /* 448 * Note: we should add inode to orphan list before f2fs_unlock_op() 449 * so we can prevent losing this orphan when encoutering checkpoint 450 * and following suddenly power-off. 451 */ 452 get_node_info(sbi, inode->i_ino, &ni); 453 454 if (ni.blk_addr != NULL_ADDR) { 455 int err = acquire_orphan_inode(sbi); 456 if (err) { 457 set_sbi_flag(sbi, SBI_NEED_FSCK); 458 f2fs_msg(sbi->sb, KERN_WARNING, 459 "Too many orphan inodes, run fsck to fix."); 460 } else { 461 add_orphan_inode(inode); 462 } 463 alloc_nid_done(sbi, inode->i_ino); 464 } else { 465 set_inode_flag(inode, FI_FREE_NID); 466 } 467 468 f2fs_unlock_op(sbi); 469 470 /* iput will drop the inode object */ 471 iput(inode); 472 } 473