fs/erofs/super.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  * Copyright (C) 2017-2018 HUAWEI, Inc.
   4  *             http://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_nouser_xattr,
 215         Opt_acl,
 216         Opt_noacl,
 217         Opt_cache_strategy,
 218         Opt_err
 219 };
 220
 221 static const struct constant_table erofs_param_cache_strategy[] = {
 222         {"disabled",    EROFS_ZIP_CACHE_DISABLED},
 223         {"readahead",   EROFS_ZIP_CACHE_READAHEAD},
 224         {"readaround",  EROFS_ZIP_CACHE_READAROUND},
 225         {}
 226 };
 227
 228 static const struct fs_parameter_spec erofs_fs_parameters[] = {
 229         fsparam_flag_no("user_xattr",   Opt_user_xattr),
 230         fsparam_flag_no("acl",          Opt_acl),
 231         fsparam_enum("cache_strategy",  Opt_cache_strategy,
 232                      erofs_param_cache_strategy),
 233         {}
 234 };
 235
 236 static int erofs_fc_parse_param(struct fs_context *fc,
 237                                 struct fs_parameter *param)
 238 {
 239         struct erofs_fs_context *ctx __maybe_unused = fc->fs_private;
 240         struct fs_parse_result result;
 241         int opt;
 242
 243         opt = fs_parse(fc, erofs_fs_parameters, param, &result);
 244         if (opt < 0)
 245                 return opt;
 246
 247         switch (opt) {
 248         case Opt_user_xattr:
 249 #ifdef CONFIG_EROFS_FS_XATTR
 250                 if (result.boolean)
 251                         set_opt(ctx, XATTR_USER);
 252                 else
 253                         clear_opt(ctx, XATTR_USER);
 254 #else
 255                 errorfc(fc, "{,no}user_xattr options not supported");
 256 #endif
 257                 break;
 258         case Opt_acl:
 259 #ifdef CONFIG_EROFS_FS_POSIX_ACL
 260                 if (result.boolean)
 261                         set_opt(ctx, POSIX_ACL);
 262                 else
 263                         clear_opt(ctx, POSIX_ACL);
 264 #else
 265                 errorfc(fc, "{,no}acl options not supported");
 266 #endif
 267                 break;
 268         case Opt_cache_strategy:
 269 #ifdef CONFIG_EROFS_FS_ZIP
 270                 ctx->cache_strategy = result.uint_32;
 271 #else
 272                 errorfc(fc, "compression not supported, cache_strategy ignored");
 273 #endif
 274                 break;
 275         default:
 276                 return -ENOPARAM;
 277         }
 278         return 0;
 279 }
 280
 281 #ifdef CONFIG_EROFS_FS_ZIP
 282 static const struct address_space_operations managed_cache_aops;
 283
 284 static int erofs_managed_cache_releasepage(struct page *page, gfp_t gfp_mask)
 285 {
 286         int ret = 1;    /* 0 - busy */
 287         struct address_space *const mapping = page->mapping;
 288
 289         DBG_BUGON(!PageLocked(page));
 290         DBG_BUGON(mapping->a_ops != &managed_cache_aops);
 291
 292         if (PagePrivate(page))
 293                 ret = erofs_try_to_free_cached_page(mapping, page);
 294
 295         return ret;
 296 }
 297
 298 static void erofs_managed_cache_invalidatepage(struct page *page,
 299                                                unsigned int offset,
 300                                                unsigned int length)
 301 {
 302         const unsigned int stop = length + offset;
 303
 304         DBG_BUGON(!PageLocked(page));
 305
 306         /* Check for potential overflow in debug mode */
 307         DBG_BUGON(stop > PAGE_SIZE || stop < length);
 308
 309         if (offset == 0 && stop == PAGE_SIZE)
 310                 while (!erofs_managed_cache_releasepage(page, GFP_NOFS))
 311                         cond_resched();
 312 }
 313
 314 static const struct address_space_operations managed_cache_aops = {
 315         .releasepage = erofs_managed_cache_releasepage,
 316         .invalidatepage = erofs_managed_cache_invalidatepage,
 317 };
 318
 319 static int erofs_init_managed_cache(struct super_block *sb)
 320 {
 321         struct erofs_sb_info *const sbi = EROFS_SB(sb);
 322         struct inode *const inode = new_inode(sb);
 323
 324         if (!inode)
 325                 return -ENOMEM;
 326
 327         set_nlink(inode, 1);
 328         inode->i_size = OFFSET_MAX;
 329
 330         inode->i_mapping->a_ops = &managed_cache_aops;
 331         mapping_set_gfp_mask(inode->i_mapping,
 332                              GFP_NOFS | __GFP_HIGHMEM | __GFP_MOVABLE);
 333         sbi->managed_cache = inode;
 334         return 0;
 335 }
 336 #else
 337 static int erofs_init_managed_cache(struct super_block *sb) { return 0; }
 338 #endif
 339
 340 static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
 341 {
 342         struct inode *inode;
 343         struct erofs_sb_info *sbi;
 344         struct erofs_fs_context *ctx = fc->fs_private;
 345         int err;
 346
 347         sb->s_magic = EROFS_SUPER_MAGIC;
 348
 349         if (!sb_set_blocksize(sb, EROFS_BLKSIZ)) {
 350                 erofs_err(sb, "failed to set erofs blksize");
 351                 return -EINVAL;
 352         }
 353
 354         sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 355         if (!sbi)
 356                 return -ENOMEM;
 357
 358         sb->s_fs_info = sbi;
 359         err = erofs_read_superblock(sb);
 360         if (err)
 361                 return err;
 362
 363         sb->s_flags |= SB_RDONLY | SB_NOATIME;
 364         sb->s_maxbytes = MAX_LFS_FILESIZE;
 365         sb->s_time_gran = 1;
 366
 367         sb->s_op = &erofs_sops;
 368         sb->s_xattr = erofs_xattr_handlers;
 369
 370         if (test_opt(ctx, POSIX_ACL))
 371                 sb->s_flags |= SB_POSIXACL;
 372         else
 373                 sb->s_flags &= ~SB_POSIXACL;
 374
 375         sbi->ctx = *ctx;
 376
 377 #ifdef CONFIG_EROFS_FS_ZIP
 378         xa_init(&sbi->managed_pslots);
 379 #endif
 380
 381         /* get the root inode */
 382         inode = erofs_iget(sb, ROOT_NID(sbi), true);
 383         if (IS_ERR(inode))
 384                 return PTR_ERR(inode);
 385
 386         if (!S_ISDIR(inode->i_mode)) {
 387                 erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
 388                           ROOT_NID(sbi), inode->i_mode);
 389                 iput(inode);
 390                 return -EINVAL;
 391         }
 392
 393         sb->s_root = d_make_root(inode);
 394         if (!sb->s_root)
 395                 return -ENOMEM;
 396
 397         erofs_shrinker_register(sb);
 398         /* sb->s_umount is already locked, SB_ACTIVE and SB_BORN are not set */
 399         err = erofs_init_managed_cache(sb);
 400         if (err)
 401                 return err;
 402
 403         erofs_info(sb, "mounted with root inode @ nid %llu.", ROOT_NID(sbi));
 404         return 0;
 405 }
 406
 407 static int erofs_fc_get_tree(struct fs_context *fc)
 408 {
 409         return get_tree_bdev(fc, erofs_fc_fill_super);
 410 }
 411
 412 static int erofs_fc_reconfigure(struct fs_context *fc)
 413 {
 414         struct super_block *sb = fc->root->d_sb;
 415         struct erofs_sb_info *sbi = EROFS_SB(sb);
 416         struct erofs_fs_context *ctx = fc->fs_private;
 417
 418         DBG_BUGON(!sb_rdonly(sb));
 419
 420         if (test_opt(ctx, POSIX_ACL))
 421                 fc->sb_flags |= SB_POSIXACL;
 422         else
 423                 fc->sb_flags &= ~SB_POSIXACL;
 424
 425         sbi->ctx = *ctx;
 426
 427         fc->sb_flags |= SB_RDONLY;
 428         return 0;
 429 }
 430
 431 static void erofs_fc_free(struct fs_context *fc)
 432 {
 433         kfree(fc->fs_private);
 434 }
 435
 436 static const struct fs_context_operations erofs_context_ops = {
 437         .parse_param    = erofs_fc_parse_param,
 438         .get_tree       = erofs_fc_get_tree,
 439         .reconfigure    = erofs_fc_reconfigure,
 440         .free           = erofs_fc_free,
 441 };
 442
 443 static int erofs_init_fs_context(struct fs_context *fc)
 444 {
 445         fc->fs_private = kzalloc(sizeof(struct erofs_fs_context), GFP_KERNEL);
 446         if (!fc->fs_private)
 447                 return -ENOMEM;
 448
 449         /* set default mount options */
 450         erofs_default_options(fc->fs_private);
 451
 452         fc->ops = &erofs_context_ops;
 453
 454         return 0;
 455 }
 456
 457 /*
 458  * could be triggered after deactivate_locked_super()
 459  * is called, thus including umount and failed to initialize.
 460  */
 461 static void erofs_kill_sb(struct super_block *sb)
 462 {
 463         struct erofs_sb_info *sbi;
 464
 465         WARN_ON(sb->s_magic != EROFS_SUPER_MAGIC);
 466
 467         kill_block_super(sb);
 468
 469         sbi = EROFS_SB(sb);
 470         if (!sbi)
 471                 return;
 472         kfree(sbi);
 473         sb->s_fs_info = NULL;
 474 }
 475
 476 /* called when ->s_root is non-NULL */
 477 static void erofs_put_super(struct super_block *sb)
 478 {
 479         struct erofs_sb_info *const sbi = EROFS_SB(sb);
 480
 481         DBG_BUGON(!sbi);
 482
 483         erofs_shrinker_unregister(sb);
 484 #ifdef CONFIG_EROFS_FS_ZIP
 485         iput(sbi->managed_cache);
 486         sbi->managed_cache = NULL;
 487 #endif
 488 }
 489
 490 static struct file_system_type erofs_fs_type = {
 491         .owner          = THIS_MODULE,
 492         .name           = "erofs",
 493         .init_fs_context = erofs_init_fs_context,
 494         .kill_sb        = erofs_kill_sb,
 495         .fs_flags       = FS_REQUIRES_DEV,
 496 };
 497 MODULE_ALIAS_FS("erofs");
 498
 499 static int __init erofs_module_init(void)
 500 {
 501         int err;
 502
 503         erofs_check_ondisk_layout_definitions();
 504
 505         erofs_inode_cachep = kmem_cache_create("erofs_inode",
 506                                                sizeof(struct erofs_inode), 0,
 507                                                SLAB_RECLAIM_ACCOUNT,
 508                                                erofs_inode_init_once);
 509         if (!erofs_inode_cachep) {
 510                 err = -ENOMEM;
 511                 goto icache_err;
 512         }
 513
 514         err = erofs_init_shrinker();
 515         if (err)
 516                 goto shrinker_err;
 517
 518         err = z_erofs_init_zip_subsystem();
 519         if (err)
 520                 goto zip_err;
 521
 522         err = register_filesystem(&erofs_fs_type);
 523         if (err)
 524                 goto fs_err;
 525
 526         return 0;
 527
 528 fs_err:
 529         z_erofs_exit_zip_subsystem();
 530 zip_err:
 531         erofs_exit_shrinker();
 532 shrinker_err:
 533         kmem_cache_destroy(erofs_inode_cachep);
 534 icache_err:
 535         return err;
 536 }
 537
 538 static void __exit erofs_module_exit(void)
 539 {
 540         unregister_filesystem(&erofs_fs_type);
 541         z_erofs_exit_zip_subsystem();
 542         erofs_exit_shrinker();
 543
 544         /* Ensure all RCU free inodes are safe before cache is destroyed. */
 545         rcu_barrier();
 546         kmem_cache_destroy(erofs_inode_cachep);
 547 }
 548
 549 /* get filesystem statistics */
 550 static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
 551 {
 552         struct super_block *sb = dentry->d_sb;
 553         struct erofs_sb_info *sbi = EROFS_SB(sb);
 554         u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
 555
 556         buf->f_type = sb->s_magic;
 557         buf->f_bsize = EROFS_BLKSIZ;
 558         buf->f_blocks = sbi->blocks;
 559         buf->f_bfree = buf->f_bavail = 0;
 560
 561         buf->f_files = ULLONG_MAX;
 562         buf->f_ffree = ULLONG_MAX - sbi->inos;
 563
 564         buf->f_namelen = EROFS_NAME_LEN;
 565
 566         buf->f_fsid.val[0] = (u32)id;
 567         buf->f_fsid.val[1] = (u32)(id >> 32);
 568         return 0;
 569 }
 570
 571 static int erofs_show_options(struct seq_file *seq, struct dentry *root)
 572 {
 573         struct erofs_sb_info *sbi __maybe_unused = EROFS_SB(root->d_sb);
 574         struct erofs_fs_context *ctx __maybe_unused = &sbi->ctx;
 575
 576 #ifdef CONFIG_EROFS_FS_XATTR
 577         if (test_opt(ctx, XATTR_USER))
 578                 seq_puts(seq, ",user_xattr");
 579         else
 580                 seq_puts(seq, ",nouser_xattr");
 581 #endif
 582 #ifdef CONFIG_EROFS_FS_POSIX_ACL
 583         if (test_opt(ctx, POSIX_ACL))
 584                 seq_puts(seq, ",acl");
 585         else
 586                 seq_puts(seq, ",noacl");
 587 #endif
 588 #ifdef CONFIG_EROFS_FS_ZIP
 589         if (ctx->cache_strategy == EROFS_ZIP_CACHE_DISABLED)
 590                 seq_puts(seq, ",cache_strategy=disabled");
 591         else if (ctx->cache_strategy == EROFS_ZIP_CACHE_READAHEAD)
 592                 seq_puts(seq, ",cache_strategy=readahead");
 593         else if (ctx->cache_strategy == EROFS_ZIP_CACHE_READAROUND)
 594                 seq_puts(seq, ",cache_strategy=readaround");
 595 #endif
 596         return 0;
 597 }
 598
 599 const struct super_operations erofs_sops = {
 600         .put_super = erofs_put_super,
 601         .alloc_inode = erofs_alloc_inode,
 602         .free_inode = erofs_free_inode,
 603         .statfs = erofs_statfs,
 604         .show_options = erofs_show_options,
 605 };
 606
 607 module_init(erofs_module_init);
 608 module_exit(erofs_module_exit);
 609
 610 MODULE_DESCRIPTION("Enhanced ROM File System");
 611 MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
 612 MODULE_LICENSE("GPL");
 613