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