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/backing-dev.h>
  15 #include <linux/writeback.h>
  16
  17 #include "f2fs.h"
  18 #include "node.h"
  19 #include "segment.h"
  20
  21 #include <trace/events/f2fs.h>
  22
  23 void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
  24 {
  25         if (f2fs_inode_dirtied(inode, sync))
  26                 return;
  27
  28         mark_inode_dirty_sync(inode);
  29 }
  30
  31 void f2fs_set_inode_flags(struct inode *inode)
  32 {
  33         unsigned int flags = F2FS_I(inode)->i_flags;
  34         unsigned int new_fl = 0;
  35
  36         if (flags & FS_SYNC_FL)
  37                 new_fl |= S_SYNC;
  38         if (flags & FS_APPEND_FL)
  39                 new_fl |= S_APPEND;
  40         if (flags & FS_IMMUTABLE_FL)
  41                 new_fl |= S_IMMUTABLE;
  42         if (flags & FS_NOATIME_FL)
  43                 new_fl |= S_NOATIME;
  44         if (flags & FS_DIRSYNC_FL)
  45                 new_fl |= S_DIRSYNC;
  46         inode_set_flags(inode, new_fl,
  47                         S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
  48 }
  49
  50 static void __get_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
  51 {
  52         int extra_size = get_extra_isize(inode);
  53
  54         if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
  55                         S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
  56                 if (ri->i_addr[extra_size])
  57                         inode->i_rdev = old_decode_dev(
  58                                 le32_to_cpu(ri->i_addr[extra_size]));
  59                 else
  60                         inode->i_rdev = new_decode_dev(
  61                                 le32_to_cpu(ri->i_addr[extra_size + 1]));
  62         }
  63 }
  64
  65 static bool __written_first_block(struct f2fs_inode *ri)
  66 {
  67         block_t addr = le32_to_cpu(ri->i_addr[offset_in_addr(ri)]);
  68
  69         if (addr != NEW_ADDR && addr != NULL_ADDR)
  70                 return true;
  71         return false;
  72 }
  73
  74 static void __set_inode_rdev(struct inode *inode, struct f2fs_inode *ri)
  75 {
  76         int extra_size = get_extra_isize(inode);
  77
  78         if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) {
  79                 if (old_valid_dev(inode->i_rdev)) {
  80                         ri->i_addr[extra_size] =
  81                                 cpu_to_le32(old_encode_dev(inode->i_rdev));
  82                         ri->i_addr[extra_size + 1] = 0;
  83                 } else {
  84                         ri->i_addr[extra_size] = 0;
  85                         ri->i_addr[extra_size + 1] =
  86                                 cpu_to_le32(new_encode_dev(inode->i_rdev));
  87                         ri->i_addr[extra_size + 2] = 0;
  88                 }
  89         }
  90 }
  91
  92 static void __recover_inline_status(struct inode *inode, struct page *ipage)
  93 {
  94         void *inline_data = inline_data_addr(inode, ipage);
  95         __le32 *start = inline_data;
  96         __le32 *end = start + MAX_INLINE_DATA(inode) / sizeof(__le32);
  97
  98         while (start < end) {
  99                 if (*start++) {
 100                         f2fs_wait_on_page_writeback(ipage, NODE, true);
 101
 102                         set_inode_flag(inode, FI_DATA_EXIST);
 103                         set_raw_inline(inode, F2FS_INODE(ipage));
 104                         set_page_dirty(ipage);
 105                         return;
 106                 }
 107         }
 108         return;
 109 }
 110
 111 static int do_read_inode(struct inode *inode)
 112 {
 113         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 114         struct f2fs_inode_info *fi = F2FS_I(inode);
 115         struct page *node_page;
 116         struct f2fs_inode *ri;
 117
 118         /* Check if ino is within scope */
 119         if (check_nid_range(sbi, inode->i_ino)) {
 120                 f2fs_msg(inode->i_sb, KERN_ERR, "bad inode number: %lu",
 121                          (unsigned long) inode->i_ino);
 122                 WARN_ON(1);
 123                 return -EINVAL;
 124         }
 125
 126         node_page = get_node_page(sbi, inode->i_ino);
 127         if (IS_ERR(node_page))
 128                 return PTR_ERR(node_page);
 129
 130         ri = F2FS_INODE(node_page);
 131
 132         inode->i_mode = le16_to_cpu(ri->i_mode);
 133         i_uid_write(inode, le32_to_cpu(ri->i_uid));
 134         i_gid_write(inode, le32_to_cpu(ri->i_gid));
 135         set_nlink(inode, le32_to_cpu(ri->i_links));
 136         inode->i_size = le64_to_cpu(ri->i_size);
 137         inode->i_blocks = SECTOR_FROM_BLOCK(le64_to_cpu(ri->i_blocks) - 1);
 138
 139         inode->i_atime.tv_sec = le64_to_cpu(ri->i_atime);
 140         inode->i_ctime.tv_sec = le64_to_cpu(ri->i_ctime);
 141         inode->i_mtime.tv_sec = le64_to_cpu(ri->i_mtime);
 142         inode->i_atime.tv_nsec = le32_to_cpu(ri->i_atime_nsec);
 143         inode->i_ctime.tv_nsec = le32_to_cpu(ri->i_ctime_nsec);
 144         inode->i_mtime.tv_nsec = le32_to_cpu(ri->i_mtime_nsec);
 145         inode->i_generation = le32_to_cpu(ri->i_generation);
 146
 147         fi->i_current_depth = le32_to_cpu(ri->i_current_depth);
 148         fi->i_xattr_nid = le32_to_cpu(ri->i_xattr_nid);
 149         fi->i_flags = le32_to_cpu(ri->i_flags);
 150         fi->flags = 0;
 151         fi->i_advise = ri->i_advise;
 152         fi->i_pino = le32_to_cpu(ri->i_pino);
 153         fi->i_dir_level = ri->i_dir_level;
 154
 155         if (f2fs_init_extent_tree(inode, &ri->i_ext))
 156                 set_page_dirty(node_page);
 157
 158         get_inline_info(inode, ri);
 159
 160         fi->i_extra_isize = f2fs_has_extra_attr(inode) ?
 161                                         le16_to_cpu(ri->i_extra_isize) : 0;
 162
 163         /* check data exist */
 164         if (f2fs_has_inline_data(inode) && !f2fs_exist_data(inode))
 165                 __recover_inline_status(inode, node_page);
 166
 167         /* get rdev by using inline_info */
 168         __get_inode_rdev(inode, ri);
 169
 170         if (__written_first_block(ri))
 171                 set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
 172
 173         if (!need_inode_block_update(sbi, inode->i_ino))
 174                 fi->last_disk_size = inode->i_size;
 175
 176         f2fs_put_page(node_page, 1);
 177
 178         stat_inc_inline_xattr(inode);
 179         stat_inc_inline_inode(inode);
 180         stat_inc_inline_dir(inode);
 181
 182         return 0;
 183 }
 184
 185 struct inode *f2fs_iget(struct super_block *sb, unsigned long ino)
 186 {
 187         struct f2fs_sb_info *sbi = F2FS_SB(sb);
 188         struct inode *inode;
 189         int ret = 0;
 190
 191         inode = iget_locked(sb, ino);
 192         if (!inode)
 193                 return ERR_PTR(-ENOMEM);
 194
 195         if (!(inode->i_state & I_NEW)) {
 196                 trace_f2fs_iget(inode);
 197                 return inode;
 198         }
 199         if (ino == F2FS_NODE_INO(sbi) || ino == F2FS_META_INO(sbi))
 200                 goto make_now;
 201
 202         ret = do_read_inode(inode);
 203         if (ret)
 204                 goto bad_inode;
 205 make_now:
 206         if (ino == F2FS_NODE_INO(sbi)) {
 207                 inode->i_mapping->a_ops = &f2fs_node_aops;
 208                 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO);
 209         } else if (ino == F2FS_META_INO(sbi)) {
 210                 inode->i_mapping->a_ops = &f2fs_meta_aops;
 211                 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_ZERO);
 212         } else if (S_ISREG(inode->i_mode)) {
 213                 inode->i_op = &f2fs_file_inode_operations;
 214                 inode->i_fop = &f2fs_file_operations;
 215                 inode->i_mapping->a_ops = &f2fs_dblock_aops;
 216         } else if (S_ISDIR(inode->i_mode)) {
 217                 inode->i_op = &f2fs_dir_inode_operations;
 218                 inode->i_fop = &f2fs_dir_operations;
 219                 inode->i_mapping->a_ops = &f2fs_dblock_aops;
 220                 mapping_set_gfp_mask(inode->i_mapping, GFP_F2FS_HIGH_ZERO);
 221         } else if (S_ISLNK(inode->i_mode)) {
 222                 if (f2fs_encrypted_inode(inode))
 223                         inode->i_op = &f2fs_encrypted_symlink_inode_operations;
 224                 else
 225                         inode->i_op = &f2fs_symlink_inode_operations;
 226                 inode_nohighmem(inode);
 227                 inode->i_mapping->a_ops = &f2fs_dblock_aops;
 228         } else if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode) ||
 229                         S_ISFIFO(inode->i_mode) || S_ISSOCK(inode->i_mode)) {
 230                 inode->i_op = &f2fs_special_inode_operations;
 231                 init_special_inode(inode, inode->i_mode, inode->i_rdev);
 232         } else {
 233                 ret = -EIO;
 234                 goto bad_inode;
 235         }
 236         f2fs_set_inode_flags(inode);
 237         unlock_new_inode(inode);
 238         trace_f2fs_iget(inode);
 239         return inode;
 240
 241 bad_inode:
 242         iget_failed(inode);
 243         trace_f2fs_iget_exit(inode, ret);
 244         return ERR_PTR(ret);
 245 }
 246
 247 struct inode *f2fs_iget_retry(struct super_block *sb, unsigned long ino)
 248 {
 249         struct inode *inode;
 250 retry:
 251         inode = f2fs_iget(sb, ino);
 252         if (IS_ERR(inode)) {
 253                 if (PTR_ERR(inode) == -ENOMEM) {
 254                         congestion_wait(BLK_RW_ASYNC, HZ/50);
 255                         goto retry;
 256                 }
 257         }
 258         return inode;
 259 }
 260
 261 int update_inode(struct inode *inode, struct page *node_page)
 262 {
 263         struct f2fs_inode *ri;
 264         struct extent_tree *et = F2FS_I(inode)->extent_tree;
 265
 266         f2fs_inode_synced(inode);
 267
 268         f2fs_wait_on_page_writeback(node_page, NODE, true);
 269
 270         ri = F2FS_INODE(node_page);
 271
 272         ri->i_mode = cpu_to_le16(inode->i_mode);
 273         ri->i_advise = F2FS_I(inode)->i_advise;
 274         ri->i_uid = cpu_to_le32(i_uid_read(inode));
 275         ri->i_gid = cpu_to_le32(i_gid_read(inode));
 276         ri->i_links = cpu_to_le32(inode->i_nlink);
 277         ri->i_size = cpu_to_le64(i_size_read(inode));
 278         ri->i_blocks = cpu_to_le64(SECTOR_TO_BLOCK(inode->i_blocks) + 1);
 279
 280         if (et) {
 281                 read_lock(&et->lock);
 282                 set_raw_extent(&et->largest, &ri->i_ext);
 283                 read_unlock(&et->lock);
 284         } else {
 285                 memset(&ri->i_ext, 0, sizeof(ri->i_ext));
 286         }
 287         set_raw_inline(inode, ri);
 288
 289         ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
 290         ri->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
 291         ri->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
 292         ri->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
 293         ri->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
 294         ri->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
 295         ri->i_current_depth = cpu_to_le32(F2FS_I(inode)->i_current_depth);
 296         ri->i_xattr_nid = cpu_to_le32(F2FS_I(inode)->i_xattr_nid);
 297         ri->i_flags = cpu_to_le32(F2FS_I(inode)->i_flags);
 298         ri->i_pino = cpu_to_le32(F2FS_I(inode)->i_pino);
 299         ri->i_generation = cpu_to_le32(inode->i_generation);
 300         ri->i_dir_level = F2FS_I(inode)->i_dir_level;
 301
 302         if (f2fs_has_extra_attr(inode))
 303                 ri->i_extra_isize = cpu_to_le16(F2FS_I(inode)->i_extra_isize);
 304
 305         __set_inode_rdev(inode, ri);
 306         set_cold_node(inode, node_page);
 307
 308         /* deleted inode */
 309         if (inode->i_nlink == 0)
 310                 clear_inline_node(node_page);
 311
 312         return set_page_dirty(node_page);
 313 }
 314
 315 int update_inode_page(struct inode *inode)
 316 {
 317         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 318         struct page *node_page;
 319         int ret = 0;
 320 retry:
 321         node_page = get_node_page(sbi, inode->i_ino);
 322         if (IS_ERR(node_page)) {
 323                 int err = PTR_ERR(node_page);
 324                 if (err == -ENOMEM) {
 325                         cond_resched();
 326                         goto retry;
 327                 } else if (err != -ENOENT) {
 328                         f2fs_stop_checkpoint(sbi, false);
 329                 }
 330                 return 0;
 331         }
 332         ret = update_inode(inode, node_page);
 333         f2fs_put_page(node_page, 1);
 334         return ret;
 335 }
 336
 337 int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
 338 {
 339         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 340
 341         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 342                         inode->i_ino == F2FS_META_INO(sbi))
 343                 return 0;
 344
 345         if (!is_inode_flag_set(inode, FI_DIRTY_INODE))
 346                 return 0;
 347
 348         /*
 349          * We need to balance fs here to prevent from producing dirty node pages
 350          * during the urgent cleaning time when runing out of free sections.
 351          */
 352         update_inode_page(inode);
 353         if (wbc && wbc->nr_to_write)
 354                 f2fs_balance_fs(sbi, true);
 355         return 0;
 356 }
 357
 358 /*
 359  * Called at the last iput() if i_nlink is zero
 360  */
 361 void f2fs_evict_inode(struct inode *inode)
 362 {
 363         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 364         nid_t xnid = F2FS_I(inode)->i_xattr_nid;
 365         int err = 0;
 366
 367         /* some remained atomic pages should discarded */
 368         if (f2fs_is_atomic_file(inode))
 369                 drop_inmem_pages(inode);
 370
 371         trace_f2fs_evict_inode(inode);
 372         truncate_inode_pages_final(&inode->i_data);
 373
 374         if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 375                         inode->i_ino == F2FS_META_INO(sbi))
 376                 goto out_clear;
 377
 378         f2fs_bug_on(sbi, get_dirty_pages(inode));
 379         remove_dirty_inode(inode);
 380
 381         f2fs_destroy_extent_tree(inode);
 382
 383         if (inode->i_nlink || is_bad_inode(inode))
 384                 goto no_delete;
 385
 386         dquot_initialize(inode);
 387
 388         remove_ino_entry(sbi, inode->i_ino, APPEND_INO);
 389         remove_ino_entry(sbi, inode->i_ino, UPDATE_INO);
 390
 391         sb_start_intwrite(inode->i_sb);
 392         set_inode_flag(inode, FI_NO_ALLOC);
 393         i_size_write(inode, 0);
 394 retry:
 395         if (F2FS_HAS_BLOCKS(inode))
 396                 err = f2fs_truncate(inode);
 397
 398 #ifdef CONFIG_F2FS_FAULT_INJECTION
 399         if (time_to_inject(sbi, FAULT_EVICT_INODE)) {
 400                 f2fs_show_injection_info(FAULT_EVICT_INODE);
 401                 err = -EIO;
 402         }
 403 #endif
 404         if (!err) {
 405                 f2fs_lock_op(sbi);
 406                 err = remove_inode_page(inode);
 407                 f2fs_unlock_op(sbi);
 408                 if (err == -ENOENT)
 409                         err = 0;
 410         }
 411
 412         /* give more chances, if ENOMEM case */
 413         if (err == -ENOMEM) {
 414                 err = 0;
 415                 goto retry;
 416         }
 417
 418         if (err)
 419                 update_inode_page(inode);
 420         dquot_free_inode(inode);
 421         sb_end_intwrite(inode->i_sb);
 422 no_delete:
 423         dquot_drop(inode);
 424
 425         stat_dec_inline_xattr(inode);
 426         stat_dec_inline_dir(inode);
 427         stat_dec_inline_inode(inode);
 428
 429         /* ino == 0, if f2fs_new_inode() was failed t*/
 430         if (inode->i_ino)
 431                 invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino,
 432                                                         inode->i_ino);
 433         if (xnid)
 434                 invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
 435         if (inode->i_nlink) {
 436                 if (is_inode_flag_set(inode, FI_APPEND_WRITE))
 437                         add_ino_entry(sbi, inode->i_ino, APPEND_INO);
 438                 if (is_inode_flag_set(inode, FI_UPDATE_WRITE))
 439                         add_ino_entry(sbi, inode->i_ino, UPDATE_INO);
 440         }
 441         if (is_inode_flag_set(inode, FI_FREE_NID)) {
 442                 alloc_nid_failed(sbi, inode->i_ino);
 443                 clear_inode_flag(inode, FI_FREE_NID);
 444         } else {
 445                 f2fs_bug_on(sbi, err &&
 446                         !exist_written_data(sbi, inode->i_ino, ORPHAN_INO));
 447         }
 448 out_clear:
 449         fscrypt_put_encryption_info(inode, NULL);
 450         clear_inode(inode);
 451 }
 452
 453 /* caller should call f2fs_lock_op() */
 454 void handle_failed_inode(struct inode *inode)
 455 {
 456         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 457         struct node_info ni;
 458
 459         /*
 460          * clear nlink of inode in order to release resource of inode
 461          * immediately.
 462          */
 463         clear_nlink(inode);
 464
 465         /*
 466          * we must call this to avoid inode being remained as dirty, resulting
 467          * in a panic when flushing dirty inodes in gdirty_list.
 468          */
 469         update_inode_page(inode);
 470         f2fs_inode_synced(inode);
 471
 472         /* don't make bad inode, since it becomes a regular file. */
 473         unlock_new_inode(inode);
 474
 475         /*
 476          * Note: we should add inode to orphan list before f2fs_unlock_op()
 477          * so we can prevent losing this orphan when encoutering checkpoint
 478          * and following suddenly power-off.
 479          */
 480         get_node_info(sbi, inode->i_ino, &ni);
 481
 482         if (ni.blk_addr != NULL_ADDR) {
 483                 int err = acquire_orphan_inode(sbi);
 484                 if (err) {
 485                         set_sbi_flag(sbi, SBI_NEED_FSCK);
 486                         f2fs_msg(sbi->sb, KERN_WARNING,
 487                                 "Too many orphan inodes, run fsck to fix.");
 488                 } else {
 489                         add_orphan_inode(inode);
 490                 }
 491                 alloc_nid_done(sbi, inode->i_ino);
 492         } else {
 493                 set_inode_flag(inode, FI_FREE_NID);
 494         }
 495
 496         f2fs_unlock_op(sbi);
 497
 498         /* iput will drop the inode object */
 499         iput(inode);
 500 }