fs/erofs/super.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  * Copyright (C) 2017-2018 HUAWEI, Inc.
   4  *             https://www.huawei.com/
   5  * Created by Gao Xiang <gaoxiang25@huawei.com>
   6  */
   7 #include <linux/module.h>
   8 #include <linux/buffer_head.h>
   9 #include <linux/statfs.h>
  10 #include <linux/parser.h>
  11 #include <linux/seq_file.h>
  12 #include <linux/crc32c.h>
  13 #include <linux/fs_context.h>
  14 #include <linux/fs_parser.h>
  15 #include "xattr.h"
  16
  17 #define CREATE_TRACE_POINTS
  18 #include <trace/events/erofs.h>
  19
  20 static struct kmem_cache *erofs_inode_cachep __read_mostly;
  21
  22 void _erofs_err(struct super_block *sb, const char *function,
  23                 const char *fmt, ...)
  24 {
  25         struct va_format vaf;
  26         va_list args;
  27
  28         va_start(args, fmt);
  29
  30         vaf.fmt = fmt;
  31         vaf.va = &args;
  32
  33         pr_err("(device %s): %s: %pV", sb->s_id, function, &vaf);
  34         va_end(args);
  35 }
  36
  37 void _erofs_info(struct super_block *sb, const char *function,
  38                  const char *fmt, ...)
  39 {
  40         struct va_format vaf;
  41         va_list args;
  42
  43         va_start(args, fmt);
  44
  45         vaf.fmt = fmt;
  46         vaf.va = &args;
  47
  48         pr_info("(device %s): %pV", sb->s_id, &vaf);
  49         va_end(args);
  50 }
  51
  52 static int erofs_superblock_csum_verify(struct super_block *sb, void *sbdata)
  53 {
  54         struct erofs_super_block *dsb;
  55         u32 expected_crc, crc;
  56
  57         dsb = kmemdup(sbdata + EROFS_SUPER_OFFSET,
  58                       EROFS_BLKSIZ - EROFS_SUPER_OFFSET, GFP_KERNEL);
  59         if (!dsb)
  60                 return -ENOMEM;
  61
  62         expected_crc = le32_to_cpu(dsb->checksum);
  63         dsb->checksum = 0;
  64         /* to allow for x86 boot sectors and other oddities. */
  65         crc = crc32c(~0, dsb, EROFS_BLKSIZ - EROFS_SUPER_OFFSET);
  66         kfree(dsb);
  67
  68         if (crc != expected_crc) {
  69                 erofs_err(sb, "invalid checksum 0x%08x, 0x%08x expected",
  70                           crc, expected_crc);
  71                 return -EBADMSG;
  72         }
  73         return 0;
  74 }
  75
  76 static void erofs_inode_init_once(void *ptr)
  77 {
  78         struct erofs_inode *vi = ptr;
  79
  80         inode_init_once(&vi->vfs_inode);
  81 }
  82
  83 static struct inode *erofs_alloc_inode(struct super_block *sb)
  84 {
  85         struct erofs_inode *vi =
  86                 kmem_cache_alloc(erofs_inode_cachep, GFP_KERNEL);
  87
  88         if (!vi)
  89                 return NULL;
  90
  91         /* zero out everything except vfs_inode */
  92         memset(vi, 0, offsetof(struct erofs_inode, vfs_inode));
  93         return &vi->vfs_inode;
  94 }
  95
  96 static void erofs_free_inode(struct inode *inode)
  97 {
  98         struct erofs_inode *vi = EROFS_I(inode);
  99
 100         /* be careful of RCU symlink path */
 101         if (inode->i_op == &erofs_fast_symlink_iops)
 102                 kfree(inode->i_link);
 103         kfree(vi->xattr_shared_xattrs);
 104
 105         kmem_cache_free(erofs_inode_cachep, vi);
 106 }
 107
 108 static bool check_layout_compatibility(struct super_block *sb,
 109                                        struct erofs_super_block *dsb)
 110 {
 111         const unsigned int feature = le32_to_cpu(dsb->feature_incompat);
 112
 113         EROFS_SB(sb)->feature_incompat = feature;
 114
 115         /* check if current kernel meets all mandatory requirements */
 116         if (feature & (~EROFS_ALL_FEATURE_INCOMPAT)) {
 117                 erofs_err(sb,
 118                           "unidentified incompatible feature %x, please upgrade kernel version",
 119                            feature & ~EROFS_ALL_FEATURE_INCOMPAT);
 120                 return false;
 121         }
 122         return true;
 123 }
 124
 125 static int erofs_read_superblock(struct super_block *sb)
 126 {
 127         struct erofs_sb_info *sbi;
 128         struct page *page;
 129         struct erofs_super_block *dsb;
 130         unsigned int blkszbits;
 131         void *data;
 132         int ret;
 133
 134         page = read_mapping_page(sb->s_bdev->bd_inode->i_mapping, 0, NULL);
 135         if (IS_ERR(page)) {
 136                 erofs_err(sb, "cannot read erofs superblock");
 137                 return PTR_ERR(page);
 138         }
 139
 140         sbi = EROFS_SB(sb);
 141
 142         data = kmap(page);
 143         dsb = (struct erofs_super_block *)(data + EROFS_SUPER_OFFSET);
 144
 145         ret = -EINVAL;
 146         if (le32_to_cpu(dsb->magic) != EROFS_SUPER_MAGIC_V1) {
 147                 erofs_err(sb, "cannot find valid erofs superblock");
 148                 goto out;
 149         }
 150
 151         sbi->feature_compat = le32_to_cpu(dsb->feature_compat);
 152         if (sbi->feature_compat & EROFS_FEATURE_COMPAT_SB_CHKSUM) {
 153                 ret = erofs_superblock_csum_verify(sb, data);
 154                 if (ret)
 155                         goto out;
 156         }
 157
 158         blkszbits = dsb->blkszbits;
 159         /* 9(512 bytes) + LOG_SECTORS_PER_BLOCK == LOG_BLOCK_SIZE */
 160         if (blkszbits != LOG_BLOCK_SIZE) {
 161                 erofs_err(sb, "blksize %u isn't supported on this platform",
 162                           1 << blkszbits);
 163                 goto out;
 164         }
 165
 166         if (!check_layout_compatibility(sb, dsb))
 167                 goto out;
 168
 169         sbi->blocks = le32_to_cpu(dsb->blocks);
 170         sbi->meta_blkaddr = le32_to_cpu(dsb->meta_blkaddr);
 171 #ifdef CONFIG_EROFS_FS_XATTR
 172         sbi->xattr_blkaddr = le32_to_cpu(dsb->xattr_blkaddr);
 173 #endif
 174         sbi->islotbits = ilog2(sizeof(struct erofs_inode_compact));
 175         sbi->root_nid = le16_to_cpu(dsb->root_nid);
 176         sbi->inos = le64_to_cpu(dsb->inos);
 177
 178         sbi->build_time = le64_to_cpu(dsb->build_time);
 179         sbi->build_time_nsec = le32_to_cpu(dsb->build_time_nsec);
 180
 181         memcpy(&sb->s_uuid, dsb->uuid, sizeof(dsb->uuid));
 182
 183         ret = strscpy(sbi->volume_name, dsb->volume_name,
 184                       sizeof(dsb->volume_name));
 185         if (ret < 0) {  /* -E2BIG */
 186                 erofs_err(sb, "bad volume name without NIL terminator");
 187                 ret = -EFSCORRUPTED;
 188                 goto out;
 189         }
 190         ret = 0;
 191 out:
 192         kunmap(page);
 193         put_page(page);
 194         return ret;
 195 }
 196
 197 /* set up default EROFS parameters */
 198 static void erofs_default_options(struct erofs_fs_context *ctx)
 199 {
 200 #ifdef CONFIG_EROFS_FS_ZIP
 201         ctx->cache_strategy = EROFS_ZIP_CACHE_READAROUND;
 202         ctx->max_sync_decompress_pages = 3;
 203 #endif
 204 #ifdef CONFIG_EROFS_FS_XATTR
 205         set_opt(ctx, XATTR_USER);
 206 #endif
 207 #ifdef CONFIG_EROFS_FS_POSIX_ACL
 208         set_opt(ctx, POSIX_ACL);
 209 #endif
 210 }
 211
 212 enum {
 213         Opt_user_xattr,
 214         Opt_acl,
 215         Opt_cache_strategy,
 216         Opt_err
 217 };
 218
 219 static const struct constant_table erofs_param_cache_strategy[] = {
 220         {"disabled",    EROFS_ZIP_CACHE_DISABLED},
 221         {"readahead",   EROFS_ZIP_CACHE_READAHEAD},
 222         {"readaround",  EROFS_ZIP_CACHE_READAROUND},
 223         {}
 224 };
 225
 226 static const struct fs_parameter_spec erofs_fs_parameters[] = {
 227         fsparam_flag_no("user_xattr",   Opt_user_xattr),
 228         fsparam_flag_no("acl",          Opt_acl),
 229         fsparam_enum("cache_strategy",  Opt_cache_strategy,
 230                      erofs_param_cache_strategy),
 231         {}
 232 };
 233
 234 static int erofs_fc_parse_param(struct fs_context *fc,
 235                                 struct fs_parameter *param)
 236 {
 237         struct erofs_fs_context *ctx __maybe_unused = fc->fs_private;
 238         struct fs_parse_result result;
 239         int opt;
 240
 241         opt = fs_parse(fc, erofs_fs_parameters, param, &result);
 242         if (opt < 0)
 243                 return opt;
 244
 245         switch (opt) {
 246         case Opt_user_xattr:
 247 #ifdef CONFIG_EROFS_FS_XATTR
 248                 if (result.boolean)
 249                         set_opt(ctx, XATTR_USER);
 250                 else
 251                         clear_opt(ctx, XATTR_USER);
 252 #else
 253                 errorfc(fc, "{,no}user_xattr options not supported");
 254 #endif
 255                 break;
 256         case Opt_acl:
 257 #ifdef CONFIG_EROFS_FS_POSIX_ACL
 258                 if (result.boolean)
 259                         set_opt(ctx, POSIX_ACL);
 260                 else
 261                         clear_opt(ctx, POSIX_ACL);
 262 #else
 263                 errorfc(fc, "{,no}acl options not supported");
 264 #endif
 265                 break;
 266         case Opt_cache_strategy:
 267 #ifdef CONFIG_EROFS_FS_ZIP
 268                 ctx->cache_strategy = result.uint_32;
 269 #else
 270                 errorfc(fc, "compression not supported, cache_strategy ignored");
 271 #endif
 272                 break;
 273         default:
 274                 return -ENOPARAM;
 275         }
 276         return 0;
 277 }
 278
 279 #ifdef CONFIG_EROFS_FS_ZIP
 280 static const struct address_space_operations managed_cache_aops;
 281
 282 static int erofs_managed_cache_releasepage(struct page *page, gfp_t gfp_mask)
 283 {
 284         int ret = 1;    /* 0 - busy */
 285         struct address_space *const mapping = page->mapping;
 286
 287         DBG_BUGON(!PageLocked(page));
 288         DBG_BUGON(mapping->a_ops != &managed_cache_aops);
 289
 290         if (PagePrivate(page))
 291                 ret = erofs_try_to_free_cached_page(mapping, page);
 292
 293         return ret;
 294 }
 295
 296 static void erofs_managed_cache_invalidatepage(struct page *page,
 297                                                unsigned int offset,
 298                                                unsigned int length)
 299 {
 300         const unsigned int stop = length + offset;
 301
 302         DBG_BUGON(!PageLocked(page));
 303
 304         /* Check for potential overflow in debug mode */
 305         DBG_BUGON(stop > PAGE_SIZE || stop < length);
 306
 307         if (offset == 0 && stop == PAGE_SIZE)
 308                 while (!erofs_managed_cache_releasepage(page, GFP_NOFS))
 309                         cond_resched();
 310 }
 311
 312 static const struct address_space_operations managed_cache_aops = {
 313         .releasepage = erofs_managed_cache_releasepage,
 314         .invalidatepage = erofs_managed_cache_invalidatepage,
 315 };
 316
 317 static int erofs_init_managed_cache(struct super_block *sb)
 318 {
 319         struct erofs_sb_info *const sbi = EROFS_SB(sb);
 320         struct inode *const inode = new_inode(sb);
 321
 322         if (!inode)
 323                 return -ENOMEM;
 324
 325         set_nlink(inode, 1);
 326         inode->i_size = OFFSET_MAX;
 327
 328         inode->i_mapping->a_ops = &managed_cache_aops;
 329         mapping_set_gfp_mask(inode->i_mapping,
 330                              GFP_NOFS | __GFP_HIGHMEM | __GFP_MOVABLE);
 331         sbi->managed_cache = inode;
 332         return 0;
 333 }
 334 #else
 335 static int erofs_init_managed_cache(struct super_block *sb) { return 0; }
 336 #endif
 337
 338 static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
 339 {
 340         struct inode *inode;
 341         struct erofs_sb_info *sbi;
 342         struct erofs_fs_context *ctx = fc->fs_private;
 343         int err;
 344
 345         sb->s_magic = EROFS_SUPER_MAGIC;
 346
 347         if (!sb_set_blocksize(sb, EROFS_BLKSIZ)) {
 348                 erofs_err(sb, "failed to set erofs blksize");
 349                 return -EINVAL;
 350         }
 351
 352         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 353         if (!sbi)
 354                 return -ENOMEM;
 355
 356         sb->s_fs_info = sbi;
 357         err = erofs_read_superblock(sb);
 358         if (err)
 359                 return err;
 360
 361         sb->s_flags |= SB_RDONLY | SB_NOATIME;
 362         sb->s_maxbytes = MAX_LFS_FILESIZE;
 363         sb->s_time_gran = 1;
 364
 365         sb->s_op = &erofs_sops;
 366         sb->s_xattr = erofs_xattr_handlers;
 367
 368         if (test_opt(ctx, POSIX_ACL))
 369                 sb->s_flags |= SB_POSIXACL;
 370         else
 371                 sb->s_flags &= ~SB_POSIXACL;
 372
 373         sbi->ctx = *ctx;
 374
 375 #ifdef CONFIG_EROFS_FS_ZIP
 376         xa_init(&sbi->managed_pslots);
 377 #endif
 378
 379         /* get the root inode */
 380         inode = erofs_iget(sb, ROOT_NID(sbi), true);
 381         if (IS_ERR(inode))
 382                 return PTR_ERR(inode);
 383
 384         if (!S_ISDIR(inode->i_mode)) {
 385                 erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
 386                           ROOT_NID(sbi), inode->i_mode);
 387                 iput(inode);
 388                 return -EINVAL;
 389         }
 390
 391         sb->s_root = d_make_root(inode);
 392         if (!sb->s_root)
 393                 return -ENOMEM;
 394
 395         erofs_shrinker_register(sb);
 396         /* sb->s_umount is already locked, SB_ACTIVE and SB_BORN are not set */
 397         err = erofs_init_managed_cache(sb);
 398         if (err)
 399                 return err;
 400
 401         erofs_info(sb, "mounted with root inode @ nid %llu.", ROOT_NID(sbi));
 402         return 0;
 403 }
 404
 405 static int erofs_fc_get_tree(struct fs_context *fc)
 406 {
 407         return get_tree_bdev(fc, erofs_fc_fill_super);
 408 }
 409
 410 static int erofs_fc_reconfigure(struct fs_context *fc)
 411 {
 412         struct super_block *sb = fc->root->d_sb;
 413         struct erofs_sb_info *sbi = EROFS_SB(sb);
 414         struct erofs_fs_context *ctx = fc->fs_private;
 415
 416         DBG_BUGON(!sb_rdonly(sb));
 417
 418         if (test_opt(ctx, POSIX_ACL))
 419                 fc->sb_flags |= SB_POSIXACL;
 420         else
 421                 fc->sb_flags &= ~SB_POSIXACL;
 422
 423         sbi->ctx = *ctx;
 424
 425         fc->sb_flags |= SB_RDONLY;
 426         return 0;
 427 }
 428
 429 static void erofs_fc_free(struct fs_context *fc)
 430 {
 431         kfree(fc->fs_private);
 432 }
 433
 434 static const struct fs_context_operations erofs_context_ops = {
 435         .parse_param    = erofs_fc_parse_param,
 436         .get_tree       = erofs_fc_get_tree,
 437         .reconfigure    = erofs_fc_reconfigure,
 438         .free           = erofs_fc_free,
 439 };
 440
 441 static int erofs_init_fs_context(struct fs_context *fc)
 442 {
 443         fc->fs_private = kzalloc(sizeof(struct erofs_fs_context), GFP_KERNEL);
 444         if (!fc->fs_private)
 445                 return -ENOMEM;
 446
 447         /* set default mount options */
 448         erofs_default_options(fc->fs_private);
 449
 450         fc->ops = &erofs_context_ops;
 451
 452         return 0;
 453 }
 454
 455 /*
 456  * could be triggered after deactivate_locked_super()
 457  * is called, thus including umount and failed to initialize.
 458  */
 459 static void erofs_kill_sb(struct super_block *sb)
 460 {
 461         struct erofs_sb_info *sbi;
 462
 463         WARN_ON(sb->s_magic != EROFS_SUPER_MAGIC);
 464
 465         kill_block_super(sb);
 466
 467         sbi = EROFS_SB(sb);
 468         if (!sbi)
 469                 return;
 470         kfree(sbi);
 471         sb->s_fs_info = NULL;
 472 }
 473
 474 /* called when ->s_root is non-NULL */
 475 static void erofs_put_super(struct super_block *sb)
 476 {
 477         struct erofs_sb_info *const sbi = EROFS_SB(sb);
 478
 479         DBG_BUGON(!sbi);
 480
 481         erofs_shrinker_unregister(sb);
 482 #ifdef CONFIG_EROFS_FS_ZIP
 483         iput(sbi->managed_cache);
 484         sbi->managed_cache = NULL;
 485 #endif
 486 }
 487
 488 static struct file_system_type erofs_fs_type = {
 489         .owner          = THIS_MODULE,
 490         .name           = "erofs",
 491         .init_fs_context = erofs_init_fs_context,
 492         .kill_sb        = erofs_kill_sb,
 493         .fs_flags       = FS_REQUIRES_DEV,
 494 };
 495 MODULE_ALIAS_FS("erofs");
 496
 497 static int __init erofs_module_init(void)
 498 {
 499         int err;
 500
 501         erofs_check_ondisk_layout_definitions();
 502
 503         erofs_inode_cachep = kmem_cache_create("erofs_inode",
 504                                                sizeof(struct erofs_inode), 0,
 505                                                SLAB_RECLAIM_ACCOUNT,
 506                                                erofs_inode_init_once);
 507         if (!erofs_inode_cachep) {
 508                 err = -ENOMEM;
 509                 goto icache_err;
 510         }
 511
 512         err = erofs_init_shrinker();
 513         if (err)
 514                 goto shrinker_err;
 515
 516         err = z_erofs_init_zip_subsystem();
 517         if (err)
 518                 goto zip_err;
 519
 520         err = register_filesystem(&erofs_fs_type);
 521         if (err)
 522                 goto fs_err;
 523
 524         return 0;
 525
 526 fs_err:
 527         z_erofs_exit_zip_subsystem();
 528 zip_err:
 529         erofs_exit_shrinker();
 530 shrinker_err:
 531         kmem_cache_destroy(erofs_inode_cachep);
 532 icache_err:
 533         return err;
 534 }
 535
 536 static void __exit erofs_module_exit(void)
 537 {
 538         unregister_filesystem(&erofs_fs_type);
 539         z_erofs_exit_zip_subsystem();
 540         erofs_exit_shrinker();
 541
 542         /* Ensure all RCU free inodes are safe before cache is destroyed. */
 543         rcu_barrier();
 544         kmem_cache_destroy(erofs_inode_cachep);
 545 }
 546
 547 /* get filesystem statistics */
 548 static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
 549 {
 550         struct super_block *sb = dentry->d_sb;
 551         struct erofs_sb_info *sbi = EROFS_SB(sb);
 552         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 553
 554         buf->f_type = sb->s_magic;
 555         buf->f_bsize = EROFS_BLKSIZ;
 556         buf->f_blocks = sbi->blocks;
 557         buf->f_bfree = buf->f_bavail = 0;
 558
 559         buf->f_files = ULLONG_MAX;
 560         buf->f_ffree = ULLONG_MAX - sbi->inos;
 561
 562         buf->f_namelen = EROFS_NAME_LEN;
 563
 564         buf->f_fsid    = u64_to_fsid(id);
 565         return 0;
 566 }
 567
 568 static int erofs_show_options(struct seq_file *seq, struct dentry *root)
 569 {
 570         struct erofs_sb_info *sbi __maybe_unused = EROFS_SB(root->d_sb);
 571         struct erofs_fs_context *ctx __maybe_unused = &sbi->ctx;
 572
 573 #ifdef CONFIG_EROFS_FS_XATTR
 574         if (test_opt(ctx, XATTR_USER))
 575                 seq_puts(seq, ",user_xattr");
 576         else
 577                 seq_puts(seq, ",nouser_xattr");
 578 #endif
 579 #ifdef CONFIG_EROFS_FS_POSIX_ACL
 580         if (test_opt(ctx, POSIX_ACL))
 581                 seq_puts(seq, ",acl");
 582         else
 583                 seq_puts(seq, ",noacl");
 584 #endif
 585 #ifdef CONFIG_EROFS_FS_ZIP
 586         if (ctx->cache_strategy == EROFS_ZIP_CACHE_DISABLED)
 587                 seq_puts(seq, ",cache_strategy=disabled");
 588         else if (ctx->cache_strategy == EROFS_ZIP_CACHE_READAHEAD)
 589                 seq_puts(seq, ",cache_strategy=readahead");
 590         else if (ctx->cache_strategy == EROFS_ZIP_CACHE_READAROUND)
 591                 seq_puts(seq, ",cache_strategy=readaround");
 592 #endif
 593         return 0;
 594 }
 595
 596 const struct super_operations erofs_sops = {
 597         .put_super = erofs_put_super,
 598         .alloc_inode = erofs_alloc_inode,
 599         .free_inode = erofs_free_inode,
 600         .statfs = erofs_statfs,
 601         .show_options = erofs_show_options,
 602 };
 603
 604 module_init(erofs_module_init);
 605 module_exit(erofs_module_exit);
 606
 607 MODULE_DESCRIPTION("Enhanced ROM File System");
 608 MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
 609 MODULE_LICENSE("GPL");
 610