fs/f2fs/inode.c

   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/writeback.h>
  15
  16 #include "f2fs.h"
  17 #include "node.h"
  18
  19 #include <trace/events/f2fs.h>
  20
  21 void f2fs_set_inode_flags(struct inode *inode)
  22 {
  23         unsigned int flags = F2FS_I(inode)->i_flags;
  24         unsigned int new_fl = 0;
  25
  26         if (flags & FS_SYNC_FL)
  27                 new_fl |= S_SYNC;
  28         if (flags & FS_APPEND_FL)
  29                 new_fl |= S_APPEND;
  30         if (flags & FS_IMMUTABLE_FL)
  31                 new_fl |= S_IMMUTABLE;
  32         if (flags & FS_NOATIME_FL)
  33                 new_fl |= S_NOATIME;
  34         if (flags & FS_DIRSYNC_FL)
  35                 new_fl |= S_DIRSYNC;
  36         inode_set_flags(inode, new_fl,
  37                         S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
  38 }
  39
  40 static void __get_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
  41 {
  42         if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
  43                         S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
  44                 if (ri->i_addr[0])
  45                         inode->i_rdev =
  46                                 old_decode_dev(le32_to_cpu(ri->i_addr[0]));
  47                 else
  48                         inode->i_rdev =
  49                                 new_decode_dev(le32_to_cpu(ri->i_addr[1]));
  50         }
  51 }
  52
  53 static int __written_first_block(struct f2fs_sb_info *sbi,
  54                                         struct f2fs_inode *ri)
  55 {
  56         block_t addr = le32_to_cpu(ri->i_addr[0]);
  57
  58         if (!__is_valid_data_blkaddr(addr))
  59                 return 1;
  60         if (!f2fs_is_valid_blkaddr(sbi, addr, DATA_GENERIC))
  61                 return -EFAULT;
  62         return 0;
  63 }
  64
  65 static void __set_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
  66 {
  67         if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
  68                 if (old_valid_dev(inode->i_rdev)) {
  69                         ri->i_addr[0] =
  70                                 cpu_to_le32(old_encode_dev(inode->i_rdev));
  71                         ri->i_addr[1] = 0;
  72                 } else {
  73                         ri->i_addr[0] = 0;
  74                         ri->i_addr[1] =
  75                                 cpu_to_le32(new_encode_dev(inode->i_rdev));
  76                         ri->i_addr[2] = 0;
  77                 }
  78         }
  79 }
  80
  81 static void __recover_inline_status(struct inode *inode, struct page *ipage)
  82 {
  83         void *inline_data = inline_data_addr(ipage);
  84         __le32 *start = inline_data;
  85         __le32 *end = start + MAX_INLINE_DATA / sizeof(__le32);
  86
  87         while (start < end) {
  88                 if (*start++) {
  89                         f2fs_wait_on_page_writeback(ipage, NODE);
  90
  91                         set_inode_flag(F2FS_I(inode), FI_DATA_EXIST);
  92                         set_raw_inline(F2FS_I(inode), F2FS_INODE(ipage));
  93                         set_page_dirty(ipage);
  94                         return;
  95                 }
  96         }
  97         return;
  98 }
  99
 100 static bool sanity_check_inode(struct inode *inode, struct page *node_page)
 101 {
 102         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 103         unsigned long long iblocks;
 104
 105         iblocks = le64_to_cpu(F2FS_INODE(node_page)->i_blocks);
 106         if (!iblocks) {
 107                 set_sbi_flag(sbi, SBI_NEED_FSCK);
 108                 f2fs_msg(sbi->sb, KERN_WARNING,
 109                         "%s: corrupted inode i_blocks i_ino=%lx iblocks=%llu, "
 110                         "run fsck to fix.",
 111                         __func__, inode->i_ino, iblocks);
 112                 return false;
 113         }
 114
 115         if (ino_of_node(node_page) != nid_of_node(node_page)) {
 116                 set_sbi_flag(sbi, SBI_NEED_FSCK);
 117                 f2fs_msg(sbi->sb, KERN_WARNING,
 118                         "%s: corrupted inode footer i_ino=%lx, ino,nid: "
 119                         "[%u, %u] run fsck to fix.",
 120                         __func__, inode->i_ino,
 121                         ino_of_node(node_page), nid_of_node(node_page));
 122                 return false;
 123         }
 124
 125         if (F2FS_I(inode)->extent_tree) {
 126                 struct extent_info *ei = &F2FS_I(inode)->extent_tree->largest;
 127
 128                 if (ei->len &&
 129                         (!f2fs_is_valid_blkaddr(sbi, ei->blk, DATA_GENERIC) ||
 130                         !f2fs_is_valid_blkaddr(sbi, ei->blk + ei->len - 1,
 131                                                         DATA_GENERIC))) {
 132                         set_sbi_flag(sbi, SBI_NEED_FSCK);
 133                         f2fs_msg(sbi->sb, KERN_WARNING,
 134                                 "%s: inode (ino=%lx) extent info [%u, %u, %u] "
 135                                 "is incorrect, run fsck to fix",
 136                                 __func__, inode->i_ino,
 137                                 ei->blk, ei->fofs, ei->len);
 138                         return false;
 139                 }
 140         }
 141         return true;
 142 }
 143
 144 static int do_read_inode(struct inode *inode)
 145 {
 146         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 147         struct f2fs_inode_info *fi = F2FS_I(inode);
 148         struct page *node_page;
 149         struct f2fs_inode *ri;
 150         int err;
 151
 152         /* Check if ino is within scope */
 153         if (check_nid_range(sbi, inode->i_ino)) {
 154                 f2fs_msg(inode->i_sb, KERN_ERR, "bad inode number: %lu",
 155                          (unsigned long) inode->i_ino);
 156                 WARN_ON(1);
 157                 return -EINVAL;
 158         }
 159
 160         node_page = get_node_page(sbi, inode->i_ino);
 161         if (IS_ERR(node_page))
 162                 return PTR_ERR(node_page);
 163
 164         ri = F2FS_INODE(node_page);
 165
 166         inode->i_mode = le16_to_cpu(ri->i_mode);
 167         i_uid_write(inode, le32_to_cpu(ri->i_uid));
 168         i_gid_write(inode, le32_to_cpu(ri->i_gid));
 169         set_nlink(inode, le32_to_cpu(ri->i_links));
 170         inode->i_size = le64_to_cpu(ri->i_size);
 171         inode->i_blocks = le64_to_cpu(ri->i_blocks);
 172
 173         inode->i_atime.tv_sec = le64_to_cpu(ri->i_atime);
 174         inode->i_ctime.tv_sec = le64_to_cpu(ri->i_ctime);
 175         inode->i_mtime.tv_sec = le64_to_cpu(ri->i_mtime);
 176         inode->i_atime.tv_nsec = le32_to_cpu(ri->i_atime_nsec);
 177         inode->i_ctime.tv_nsec = le32_to_cpu(ri->i_ctime_nsec);
 178         inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec);
 179         inode->i_generation = le32_to_cpu(ri->i_generation);
 180
 181         fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
 182         fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
 183         fi->i_flags = le32_to_cpu(ri->i_flags);
 184         fi->flags = 0;
 185         fi->i_advise = ri->i_advise;
 186         fi->i_pino = le32_to_cpu(ri->i_pino);
 187         fi->i_dir_level = ri->i_dir_level;
 188
 189         f2fs_init_extent_tree(inode, &ri->i_ext);
 190
 191         get_inline_info(fi, ri);
 192
 193         if (!sanity_check_inode(inode, node_page)) {
 194                 f2fs_put_page(node_page, 1);
 195                 return -EINVAL;
 196         }
 197
 198         /* check data exist */
 199         if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
 200                 __recover_inline_status(inode, node_page);
 201
 202         /* get rdev by using inline_info */
 203         __get_inode_rdev(inode, ri);
 204
 205         err = __written_first_block(sbi, ri);
 206         if (err < 0) {
 207                 f2fs_put_page(node_page, 1);
 208                 return err;
 209         }
 210         if (!err)
 211                 set_inode_flag(F2FS_I(inode), FI_FIRST_BLOCK_WRITTEN);
 212
 213         f2fs_put_page(node_page, 1);
 214
 215         stat_inc_inline_xattr(inode);
 216         stat_inc_inline_inode(inode);
 217         stat_inc_inline_dir(inode);
 218
 219         return 0;
 220 }
 221
 222 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
 223 {
 224         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 225         struct inode *inode;
 226         int ret = 0;
 227
 228         inode = iget_locked(sb, ino);
 229         if (!inode)
 230                 return ERR_PTR(-ENOMEM);
 231
 232         if (!(inode->i_state & I_NEW)) {
 233                 trace_f2fs_iget(inode);
 234                 return inode;
 235         }
 236         if (ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi))
 237                 goto make_now;
 238
 239         ret = do_read_inode(inode);
 240         if (ret)
 241                 goto bad_inode;
 242 make_now:
 243         if (ino == F2FS_NODE_INO(sbi)) {
 244                 inode->i_mapping->a_ops = &f2fs_node_aops;
 245                 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO);
 246         } else if (ino == F2FS_META_INO(sbi)) {
 247                 inode->i_mapping->a_ops = &f2fs_meta_aops;
 248                 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO);
 249         } else if (S_ISREG(inode->i_mode)) {
 250                 inode->i_op = &f2fs_file_inode_operations;
 251                 inode->i_fop = &f2fs_file_operations;
 252                 inode->i_mapping->a_ops = &f2fs_dblock_aops;
 253         } else if (S_ISDIR(inode->i_mode)) {
 254                 inode->i_op = &f2fs_dir_inode_operations;
 255                 inode->i_fop = &f2fs_dir_operations;
 256                 inode->i_mapping->a_ops = &f2fs_dblock_aops;
 257                 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_HIGH_ZERO);
 258         } else if (S_ISLNK(inode->i_mode)) {
 259                 if (f2fs_encrypted_inode(inode))
 260                         inode->i_op = &f2fs_encrypted_symlink_inode_operations;
 261                 else
 262                         inode->i_op = &f2fs_symlink_inode_operations;
 263                 inode_nohighmem(inode);
 264                 inode->i_mapping->a_ops = &f2fs_dblock_aops;
 265         } else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
 266                         S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
 267                 inode->i_op = &f2fs_special_inode_operations;
 268                 init_special_inode(inode, inode->i_mode, inode->i_rdev);
 269         } else {
 270                 ret = -EIO;
 271                 goto bad_inode;
 272         }
 273         unlock_new_inode(inode);
 274         trace_f2fs_iget(inode);
 275         return inode;
 276
 277 bad_inode:
 278         iget_failed(inode);
 279         trace_f2fs_iget_exit(inode, ret);
 280         return ERR_PTR(ret);
 281 }
 282
 283 void update_inode(struct inode *inode, struct page *node_page)
 284 {
 285         struct f2fs_inode *ri;
 286
 287         f2fs_wait_on_page_writeback(node_page, NODE);
 288
 289         ri = F2FS_INODE(node_page);
 290
 291         ri->i_mode = cpu_to_le16(inode->i_mode);
 292         ri->i_advise = F2FS_I(inode)->i_advise;
 293         ri->i_uid = cpu_to_le32(i_uid_read(inode));
 294         ri->i_gid = cpu_to_le32(i_gid_read(inode));
 295         ri->i_links = cpu_to_le32(inode->i_nlink);
 296         ri->i_size = cpu_to_le64(i_size_read(inode));
 297         ri->i_blocks = cpu_to_le64(inode->i_blocks);
 298
 299         if (F2FS_I(inode)->extent_tree)
 300                 set_raw_extent(&F2FS_I(inode)->extent_tree->largest,
 301                                                         &ri->i_ext);
 302         else
 303                 memset(&ri->i_ext, 0, sizeof(ri->i_ext));
 304         set_raw_inline(F2FS_I(inode), ri);
 305
 306         ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
 307         ri->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
 308         ri->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
 309         ri->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
 310         ri->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
 311         ri->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
 312         ri->i_current_depth = cpu_to_le32(F2FS_I(inode)->i_current_depth);
 313         ri->i_xattr_nid = cpu_to_le32(F2FS_I(inode)->i_xattr_nid);
 314         ri->i_flags = cpu_to_le32(F2FS_I(inode)->i_flags);
 315         ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino);
 316         ri->i_generation = cpu_to_le32(inode->i_generation);
 317         ri->i_dir_level = F2FS_I(inode)->i_dir_level;
 318
 319         __set_inode_rdev(inode, ri);
 320         set_cold_node(inode, node_page);
 321         set_page_dirty(node_page);
 322
 323         clear_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
 324 }
 325
 326 void update_inode_page(struct inode *inode)
 327 {
 328         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 329         struct page *node_page;
 330 retry:
 331         node_page = get_node_page(sbi, inode->i_ino);
 332         if (IS_ERR(node_page)) {
 333                 int err = PTR_ERR(node_page);
 334                 if (err == -ENOMEM) {
 335                         cond_resched();
 336                         goto retry;
 337                 } else if (err != -ENOENT) {
 338                         f2fs_stop_checkpoint(sbi);
 339                 }
 340                 return;
 341         }
 342         update_inode(inode, node_page);
 343         f2fs_put_page(node_page, 1);
 344 }
 345
 346 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
 347 {
 348         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 349
 350         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 351                         inode->i_ino == F2FS_META_INO(sbi))
 352                 return 0;
 353
 354         if (!is_inode_flag_set(F2FS_I(inode), FI_DIRTY_INODE))
 355                 return 0;
 356
 357         /*
 358          * We need to balance fs here to prevent from producing dirty node pages
 359          * during the urgent cleaning time when runing out of free sections.
 360          */
 361         update_inode_page(inode);
 362
 363         f2fs_balance_fs(sbi);
 364         return 0;
 365 }
 366
 367 /*
 368  * Called at the last iput() if i_nlink is zero
 369  */
 370 void f2fs_evict_inode(struct inode *inode)
 371 {
 372         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 373         struct f2fs_inode_info *fi = F2FS_I(inode);
 374         nid_t xnid = fi->i_xattr_nid;
 375         int err = 0;
 376
 377         /* some remained atomic pages should discarded */
 378         if (f2fs_is_atomic_file(inode))
 379                 commit_inmem_pages(inode, true);
 380
 381         trace_f2fs_evict_inode(inode);
 382         truncate_inode_pages_final(&inode->i_data);
 383
 384         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 385                         inode->i_ino == F2FS_META_INO(sbi))
 386                 goto out_clear;
 387
 388         f2fs_bug_on(sbi, get_dirty_pages(inode));
 389         remove_dirty_dir_inode(inode);
 390
 391         f2fs_destroy_extent_tree(inode);
 392
 393         if (inode->i_nlink || is_bad_inode(inode))
 394                 goto no_delete;
 395
 396         sb_start_intwrite(inode->i_sb);
 397         set_inode_flag(fi, FI_NO_ALLOC);
 398         i_size_write(inode, 0);
 399
 400         if (F2FS_HAS_BLOCKS(inode))
 401                 err = f2fs_truncate(inode, true);
 402
 403         if (!err) {
 404                 f2fs_lock_op(sbi);
 405                 err = remove_inode_page(inode);
 406                 f2fs_unlock_op(sbi);
 407         }
 408
 409         sb_end_intwrite(inode->i_sb);
 410 no_delete:
 411         stat_dec_inline_xattr(inode);
 412         stat_dec_inline_dir(inode);
 413         stat_dec_inline_inode(inode);
 414
 415         invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino, inode->i_ino);
 416         if (xnid)
 417                 invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
 418         if (is_inode_flag_set(fi, FI_APPEND_WRITE))
 419                 add_dirty_inode(sbi, inode->i_ino, APPEND_INO);
 420         if (is_inode_flag_set(fi, FI_UPDATE_WRITE))
 421                 add_dirty_inode(sbi, inode->i_ino, UPDATE_INO);
 422         if (is_inode_flag_set(fi, FI_FREE_NID)) {
 423                 if (err && err != -ENOENT)
 424                         alloc_nid_done(sbi, inode->i_ino);
 425                 else
 426                         alloc_nid_failed(sbi, inode->i_ino);
 427                 clear_inode_flag(fi, FI_FREE_NID);
 428         }
 429
 430         if (err && err != -ENOENT) {
 431                 if (!exist_written_data(sbi, inode->i_ino, ORPHAN_INO)) {
 432                         /*
 433                          * get here because we failed to release resource
 434                          * of inode previously, reminder our user to run fsck
 435                          * for fixing.
 436                          */
 437                         set_sbi_flag(sbi, SBI_NEED_FSCK);
 438                         f2fs_msg(sbi->sb, KERN_WARNING,
 439                                 "inode (ino:%lu) resource leak, run fsck "
 440                                 "to fix this issue!", inode->i_ino);
 441                 }
 442         }
 443 out_clear:
 444 #ifdef CONFIG_F2FS_FS_ENCRYPTION
 445         if (fi->i_crypt_info)
 446                 f2fs_free_encryption_info(inode, fi->i_crypt_info);
 447 #endif
 448         clear_inode(inode);
 449 }
 450
 451 /* caller should call f2fs_lock_op() */
 452 void handle_failed_inode(struct inode *inode)
 453 {
 454         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 455         int err = 0;
 456
 457         clear_nlink(inode);
 458         make_bad_inode(inode);
 459         unlock_new_inode(inode);
 460
 461         i_size_write(inode, 0);
 462         if (F2FS_HAS_BLOCKS(inode))
 463                 err = f2fs_truncate(inode, false);
 464
 465         if (!err)
 466                 err = remove_inode_page(inode);
 467
 468         /*
 469          * if we skip truncate_node in remove_inode_page bacause we failed
 470          * before, it's better to find another way to release resource of
 471          * this inode (e.g. valid block count, node block or nid). Here we
 472          * choose to add this inode to orphan list, so that we can call iput
 473          * for releasing in orphan recovery flow.
 474          *
 475          * Note: we should add inode to orphan list before f2fs_unlock_op()
 476          * so we can prevent losing this orphan when encoutering checkpoint
 477          * and following suddenly power-off.
 478          */
 479         if (err && err != -ENOENT) {
 480                 err = acquire_orphan_inode(sbi);
 481                 if (!err)
 482                         add_orphan_inode(sbi, inode->i_ino);
 483         }
 484
 485         set_inode_flag(F2FS_I(inode), FI_FREE_NID);
 486         f2fs_unlock_op(sbi);
 487
 488         /* iput will drop the inode object */
 489         iput(inode);
 490 }