1// SPDX-License-Identifier: GPL-2.0-only
   2/*
   3 * Copyright (C) 2017-2018 HUAWEI, Inc.
   4 *             https://www.huawei.com/
   5 * Copyright (C) 2021, Alibaba Cloud
   6 */
   7#include <linux/statfs.h>
   8#include <linux/seq_file.h>
   9#include <linux/crc32c.h>
  10#include <linux/fs_context.h>
  11#include <linux/fs_parser.h>
  12#include <linux/exportfs.h>
  13#include "xattr.h"
  14
  15#define CREATE_TRACE_POINTS
  16#include <trace/events/erofs.h>
  17
  18static struct kmem_cache *erofs_inode_cachep __read_mostly;
  19
  20void _erofs_err(struct super_block *sb, const char *func, const char *fmt, ...)
  21{
  22	struct va_format vaf;
  23	va_list args;
  24
  25	va_start(args, fmt);
  26
  27	vaf.fmt = fmt;
  28	vaf.va = &args;
  29
  30	if (sb)
  31		pr_err("(device %s): %s: %pV", sb->s_id, func, &vaf);
  32	else
  33		pr_err("%s: %pV", func, &vaf);
  34	va_end(args);
  35}
  36
  37void _erofs_info(struct super_block *sb, const char *func, const char *fmt, ...)
  38{
  39	struct va_format vaf;
  40	va_list args;
  41
  42	va_start(args, fmt);
  43
  44	vaf.fmt = fmt;
  45	vaf.va = &args;
  46
  47	if (sb)
  48		pr_info("(device %s): %pV", sb->s_id, &vaf);
  49	else
  50		pr_info("%pV", &vaf);
  51	va_end(args);
  52}
  53
  54static int erofs_superblock_csum_verify(struct super_block *sb, void *sbdata)
  55{
  56	size_t len = 1 << EROFS_SB(sb)->blkszbits;
  57	struct erofs_super_block *dsb;
  58	u32 expected_crc, crc;
  59
  60	if (len > EROFS_SUPER_OFFSET)
  61		len -= EROFS_SUPER_OFFSET;
  62
  63	dsb = kmemdup(sbdata + EROFS_SUPER_OFFSET, len, GFP_KERNEL);
  64	if (!dsb)
  65		return -ENOMEM;
  66
  67	expected_crc = le32_to_cpu(dsb->checksum);
  68	dsb->checksum = 0;
  69	/* to allow for x86 boot sectors and other oddities. */
  70	crc = crc32c(~0, dsb, len);
  71	kfree(dsb);
  72
  73	if (crc != expected_crc) {
  74		erofs_err(sb, "invalid checksum 0x%08x, 0x%08x expected",
  75			  crc, expected_crc);
  76		return -EBADMSG;
  77	}
  78	return 0;
  79}
  80
  81static void erofs_inode_init_once(void *ptr)
  82{
  83	struct erofs_inode *vi = ptr;
  84
  85	inode_init_once(&vi->vfs_inode);
  86}
  87
  88static struct inode *erofs_alloc_inode(struct super_block *sb)
  89{
  90	struct erofs_inode *vi =
  91		alloc_inode_sb(sb, erofs_inode_cachep, GFP_KERNEL);
  92
  93	if (!vi)
  94		return NULL;
  95
  96	/* zero out everything except vfs_inode */
  97	memset(vi, 0, offsetof(struct erofs_inode, vfs_inode));
  98	return &vi->vfs_inode;
  99}
 100
 101static void erofs_free_inode(struct inode *inode)
 102{
 103	struct erofs_inode *vi = EROFS_I(inode);
 104
 105	if (inode->i_op == &erofs_fast_symlink_iops)
 106		kfree(inode->i_link);
 107	kfree(vi->xattr_shared_xattrs);
 108	kmem_cache_free(erofs_inode_cachep, vi);
 109}
 110
 111static bool check_layout_compatibility(struct super_block *sb,
 112				       struct erofs_super_block *dsb)
 113{
 114	const unsigned int feature = le32_to_cpu(dsb->feature_incompat);
 115
 116	EROFS_SB(sb)->feature_incompat = feature;
 117
 118	/* check if current kernel meets all mandatory requirements */
 119	if (feature & (~EROFS_ALL_FEATURE_INCOMPAT)) {
 120		erofs_err(sb, "unidentified incompatible feature %x, please upgrade kernel",
 121			   feature & ~EROFS_ALL_FEATURE_INCOMPAT);
 122		return false;
 123	}
 124	return true;
 125}
 126
 127/* read variable-sized metadata, offset will be aligned by 4-byte */
 128void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
 129			  erofs_off_t *offset, int *lengthp)
 130{
 131	u8 *buffer, *ptr;
 132	int len, i, cnt;
 133
 134	*offset = round_up(*offset, 4);
 135	ptr = erofs_bread(buf, erofs_blknr(sb, *offset), EROFS_KMAP);
 136	if (IS_ERR(ptr))
 137		return ptr;
 138
 139	len = le16_to_cpu(*(__le16 *)&ptr[erofs_blkoff(sb, *offset)]);
 140	if (!len)
 141		len = U16_MAX + 1;
 142	buffer = kmalloc(len, GFP_KERNEL);
 143	if (!buffer)
 144		return ERR_PTR(-ENOMEM);
 145	*offset += sizeof(__le16);
 146	*lengthp = len;
 147
 148	for (i = 0; i < len; i += cnt) {
 149		cnt = min_t(int, sb->s_blocksize - erofs_blkoff(sb, *offset),
 150			    len - i);
 151		ptr = erofs_bread(buf, erofs_blknr(sb, *offset), EROFS_KMAP);
 152		if (IS_ERR(ptr)) {
 153			kfree(buffer);
 154			return ptr;
 155		}
 156		memcpy(buffer + i, ptr + erofs_blkoff(sb, *offset), cnt);
 157		*offset += cnt;
 158	}
 159	return buffer;
 160}
 161
 162#ifndef CONFIG_EROFS_FS_ZIP
 163static int z_erofs_parse_cfgs(struct super_block *sb,
 164			      struct erofs_super_block *dsb)
 165{
 166	if (!dsb->u1.available_compr_algs)
 167		return 0;
 168
 169	erofs_err(sb, "compression disabled, unable to mount compressed EROFS");
 170	return -EOPNOTSUPP;
 171}
 172#endif
 173
 174static int erofs_init_device(struct erofs_buf *buf, struct super_block *sb,
 175			     struct erofs_device_info *dif, erofs_off_t *pos)
 176{
 177	struct erofs_sb_info *sbi = EROFS_SB(sb);
 178	struct erofs_fscache *fscache;
 179	struct erofs_deviceslot *dis;
 180	struct bdev_handle *bdev_handle;
 181	void *ptr;
 182
 183	ptr = erofs_read_metabuf(buf, sb, erofs_blknr(sb, *pos), EROFS_KMAP);
 184	if (IS_ERR(ptr))
 185		return PTR_ERR(ptr);
 186	dis = ptr + erofs_blkoff(sb, *pos);
 187
 188	if (!sbi->devs->flatdev && !dif->path) {
 189		if (!dis->tag[0]) {
 190			erofs_err(sb, "empty device tag @ pos %llu", *pos);
 191			return -EINVAL;
 192		}
 193		dif->path = kmemdup_nul(dis->tag, sizeof(dis->tag), GFP_KERNEL);
 194		if (!dif->path)
 195			return -ENOMEM;
 196	}
 197
 198	if (erofs_is_fscache_mode(sb)) {
 199		fscache = erofs_fscache_register_cookie(sb, dif->path, 0);
 200		if (IS_ERR(fscache))
 201			return PTR_ERR(fscache);
 202		dif->fscache = fscache;
 203	} else if (!sbi->devs->flatdev) {
 204		bdev_handle = bdev_open_by_path(dif->path, BLK_OPEN_READ,
 205						sb->s_type, NULL);
 206		if (IS_ERR(bdev_handle))
 207			return PTR_ERR(bdev_handle);
 208		dif->bdev_handle = bdev_handle;
 209		dif->dax_dev = fs_dax_get_by_bdev(bdev_handle->bdev,
 210				&dif->dax_part_off, NULL, NULL);
 211	}
 212
 213	dif->blocks = le32_to_cpu(dis->blocks);
 214	dif->mapped_blkaddr = le32_to_cpu(dis->mapped_blkaddr);
 215	sbi->total_blocks += dif->blocks;
 216	*pos += EROFS_DEVT_SLOT_SIZE;
 217	return 0;
 218}
 219
 220static int erofs_scan_devices(struct super_block *sb,
 221			      struct erofs_super_block *dsb)
 222{
 223	struct erofs_sb_info *sbi = EROFS_SB(sb);
 224	unsigned int ondisk_extradevs;
 225	erofs_off_t pos;
 226	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
 227	struct erofs_device_info *dif;
 228	int id, err = 0;
 229
 230	sbi->total_blocks = sbi->primarydevice_blocks;
 231	if (!erofs_sb_has_device_table(sbi))
 232		ondisk_extradevs = 0;
 233	else
 234		ondisk_extradevs = le16_to_cpu(dsb->extra_devices);
 235
 236	if (sbi->devs->extra_devices &&
 237	    ondisk_extradevs != sbi->devs->extra_devices) {
 238		erofs_err(sb, "extra devices don't match (ondisk %u, given %u)",
 239			  ondisk_extradevs, sbi->devs->extra_devices);
 240		return -EINVAL;
 241	}
 242	if (!ondisk_extradevs)
 243		return 0;
 244
 245	if (!sbi->devs->extra_devices && !erofs_is_fscache_mode(sb))
 246		sbi->devs->flatdev = true;
 247
 248	sbi->device_id_mask = roundup_pow_of_two(ondisk_extradevs + 1) - 1;
 249	pos = le16_to_cpu(dsb->devt_slotoff) * EROFS_DEVT_SLOT_SIZE;
 250	down_read(&sbi->devs->rwsem);
 251	if (sbi->devs->extra_devices) {
 252		idr_for_each_entry(&sbi->devs->tree, dif, id) {
 253			err = erofs_init_device(&buf, sb, dif, &pos);
 254			if (err)
 255				break;
 256		}
 257	} else {
 258		for (id = 0; id < ondisk_extradevs; id++) {
 259			dif = kzalloc(sizeof(*dif), GFP_KERNEL);
 260			if (!dif) {
 261				err = -ENOMEM;
 262				break;
 263			}
 264
 265			err = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
 266			if (err < 0) {
 267				kfree(dif);
 268				break;
 269			}
 270			++sbi->devs->extra_devices;
 271
 272			err = erofs_init_device(&buf, sb, dif, &pos);
 273			if (err)
 274				break;
 275		}
 276	}
 277	up_read(&sbi->devs->rwsem);
 278	erofs_put_metabuf(&buf);
 279	return err;
 280}
 281
 282static int erofs_read_superblock(struct super_block *sb)
 283{
 284	struct erofs_sb_info *sbi;
 285	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
 286	struct erofs_super_block *dsb;
 287	void *data;
 288	int ret;
 289
 290	data = erofs_read_metabuf(&buf, sb, 0, EROFS_KMAP);
 291	if (IS_ERR(data)) {
 292		erofs_err(sb, "cannot read erofs superblock");
 293		return PTR_ERR(data);
 294	}
 295
 296	sbi = EROFS_SB(sb);
 297	dsb = (struct erofs_super_block *)(data + EROFS_SUPER_OFFSET);
 298
 299	ret = -EINVAL;
 300	if (le32_to_cpu(dsb->magic) != EROFS_SUPER_MAGIC_V1) {
 301		erofs_err(sb, "cannot find valid erofs superblock");
 302		goto out;
 303	}
 304
 305	sbi->blkszbits  = dsb->blkszbits;
 306	if (sbi->blkszbits < 9 || sbi->blkszbits > PAGE_SHIFT) {
 307		erofs_err(sb, "blkszbits %u isn't supported", sbi->blkszbits);
 308		goto out;
 309	}
 310	if (dsb->dirblkbits) {
 311		erofs_err(sb, "dirblkbits %u isn't supported", dsb->dirblkbits);
 312		goto out;
 313	}
 314
 315	sbi->feature_compat = le32_to_cpu(dsb->feature_compat);
 316	if (erofs_sb_has_sb_chksum(sbi)) {
 317		ret = erofs_superblock_csum_verify(sb, data);
 318		if (ret)
 319			goto out;
 320	}
 321
 322	ret = -EINVAL;
 323	if (!check_layout_compatibility(sb, dsb))
 324		goto out;
 325
 326	sbi->sb_size = 128 + dsb->sb_extslots * EROFS_SB_EXTSLOT_SIZE;
 327	if (sbi->sb_size > PAGE_SIZE - EROFS_SUPER_OFFSET) {
 328		erofs_err(sb, "invalid sb_extslots %u (more than a fs block)",
 329			  sbi->sb_size);
 330		goto out;
 331	}
 332	sbi->primarydevice_blocks = le32_to_cpu(dsb->blocks);
 333	sbi->meta_blkaddr = le32_to_cpu(dsb->meta_blkaddr);
 334#ifdef CONFIG_EROFS_FS_XATTR
 335	sbi->xattr_blkaddr = le32_to_cpu(dsb->xattr_blkaddr);
 336	sbi->xattr_prefix_start = le32_to_cpu(dsb->xattr_prefix_start);
 337	sbi->xattr_prefix_count = dsb->xattr_prefix_count;
 338	sbi->xattr_filter_reserved = dsb->xattr_filter_reserved;
 339#endif
 340	sbi->islotbits = ilog2(sizeof(struct erofs_inode_compact));
 341	sbi->root_nid = le16_to_cpu(dsb->root_nid);
 342	sbi->packed_nid = le64_to_cpu(dsb->packed_nid);
 343	sbi->inos = le64_to_cpu(dsb->inos);
 344
 345	sbi->build_time = le64_to_cpu(dsb->build_time);
 346	sbi->build_time_nsec = le32_to_cpu(dsb->build_time_nsec);
 347
 348	memcpy(&sb->s_uuid, dsb->uuid, sizeof(dsb->uuid));
 349
 350	ret = strscpy(sbi->volume_name, dsb->volume_name,
 351		      sizeof(dsb->volume_name));
 352	if (ret < 0) {	/* -E2BIG */
 353		erofs_err(sb, "bad volume name without NIL terminator");
 354		ret = -EFSCORRUPTED;
 355		goto out;
 356	}
 357
 358	/* parse on-disk compression configurations */
 359	ret = z_erofs_parse_cfgs(sb, dsb);
 360	if (ret < 0)
 361		goto out;
 362
 363	/* handle multiple devices */
 364	ret = erofs_scan_devices(sb, dsb);
 365
 366	if (erofs_is_fscache_mode(sb))
 367		erofs_info(sb, "EXPERIMENTAL fscache-based on-demand read feature in use. Use at your own risk!");
 368out:
 369	erofs_put_metabuf(&buf);
 370	return ret;
 371}
 372
 373static void erofs_default_options(struct erofs_fs_context *ctx)
 374{
 375#ifdef CONFIG_EROFS_FS_ZIP
 376	ctx->opt.cache_strategy = EROFS_ZIP_CACHE_READAROUND;
 377	ctx->opt.max_sync_decompress_pages = 3;
 378	ctx->opt.sync_decompress = EROFS_SYNC_DECOMPRESS_AUTO;
 379#endif
 380#ifdef CONFIG_EROFS_FS_XATTR
 381	set_opt(&ctx->opt, XATTR_USER);
 382#endif
 383#ifdef CONFIG_EROFS_FS_POSIX_ACL
 384	set_opt(&ctx->opt, POSIX_ACL);
 385#endif
 386}
 387
 388enum {
 389	Opt_user_xattr,
 390	Opt_acl,
 391	Opt_cache_strategy,
 392	Opt_dax,
 393	Opt_dax_enum,
 394	Opt_device,
 395	Opt_fsid,
 396	Opt_domain_id,
 397	Opt_err
 398};
 399
 400static const struct constant_table erofs_param_cache_strategy[] = {
 401	{"disabled",	EROFS_ZIP_CACHE_DISABLED},
 402	{"readahead",	EROFS_ZIP_CACHE_READAHEAD},
 403	{"readaround",	EROFS_ZIP_CACHE_READAROUND},
 404	{}
 405};
 406
 407static const struct constant_table erofs_dax_param_enums[] = {
 408	{"always",	EROFS_MOUNT_DAX_ALWAYS},
 409	{"never",	EROFS_MOUNT_DAX_NEVER},
 410	{}
 411};
 412
 413static const struct fs_parameter_spec erofs_fs_parameters[] = {
 414	fsparam_flag_no("user_xattr",	Opt_user_xattr),
 415	fsparam_flag_no("acl",		Opt_acl),
 416	fsparam_enum("cache_strategy",	Opt_cache_strategy,
 417		     erofs_param_cache_strategy),
 418	fsparam_flag("dax",             Opt_dax),
 419	fsparam_enum("dax",		Opt_dax_enum, erofs_dax_param_enums),
 420	fsparam_string("device",	Opt_device),
 421	fsparam_string("fsid",		Opt_fsid),
 422	fsparam_string("domain_id",	Opt_domain_id),
 423	{}
 424};
 425
 426static bool erofs_fc_set_dax_mode(struct fs_context *fc, unsigned int mode)
 427{
 428#ifdef CONFIG_FS_DAX
 429	struct erofs_fs_context *ctx = fc->fs_private;
 430
 431	switch (mode) {
 432	case EROFS_MOUNT_DAX_ALWAYS:
 433		warnfc(fc, "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
 434		set_opt(&ctx->opt, DAX_ALWAYS);
 435		clear_opt(&ctx->opt, DAX_NEVER);
 436		return true;
 437	case EROFS_MOUNT_DAX_NEVER:
 438		set_opt(&ctx->opt, DAX_NEVER);
 439		clear_opt(&ctx->opt, DAX_ALWAYS);
 440		return true;
 441	default:
 442		DBG_BUGON(1);
 443		return false;
 444	}
 445#else
 446	errorfc(fc, "dax options not supported");
 447	return false;
 448#endif
 449}
 450
 451static int erofs_fc_parse_param(struct fs_context *fc,
 452				struct fs_parameter *param)
 453{
 454	struct erofs_fs_context *ctx = fc->fs_private;
 455	struct fs_parse_result result;
 456	struct erofs_device_info *dif;
 457	int opt, ret;
 458
 459	opt = fs_parse(fc, erofs_fs_parameters, param, &result);
 460	if (opt < 0)
 461		return opt;
 462
 463	switch (opt) {
 464	case Opt_user_xattr:
 465#ifdef CONFIG_EROFS_FS_XATTR
 466		if (result.boolean)
 467			set_opt(&ctx->opt, XATTR_USER);
 468		else
 469			clear_opt(&ctx->opt, XATTR_USER);
 470#else
 471		errorfc(fc, "{,no}user_xattr options not supported");
 472#endif
 473		break;
 474	case Opt_acl:
 475#ifdef CONFIG_EROFS_FS_POSIX_ACL
 476		if (result.boolean)
 477			set_opt(&ctx->opt, POSIX_ACL);
 478		else
 479			clear_opt(&ctx->opt, POSIX_ACL);
 480#else
 481		errorfc(fc, "{,no}acl options not supported");
 482#endif
 483		break;
 484	case Opt_cache_strategy:
 485#ifdef CONFIG_EROFS_FS_ZIP
 486		ctx->opt.cache_strategy = result.uint_32;
 487#else
 488		errorfc(fc, "compression not supported, cache_strategy ignored");
 489#endif
 490		break;
 491	case Opt_dax:
 492		if (!erofs_fc_set_dax_mode(fc, EROFS_MOUNT_DAX_ALWAYS))
 493			return -EINVAL;
 494		break;
 495	case Opt_dax_enum:
 496		if (!erofs_fc_set_dax_mode(fc, result.uint_32))
 497			return -EINVAL;
 498		break;
 499	case Opt_device:
 500		dif = kzalloc(sizeof(*dif), GFP_KERNEL);
 501		if (!dif)
 502			return -ENOMEM;
 503		dif->path = kstrdup(param->string, GFP_KERNEL);
 504		if (!dif->path) {
 505			kfree(dif);
 506			return -ENOMEM;
 507		}
 508		down_write(&ctx->devs->rwsem);
 509		ret = idr_alloc(&ctx->devs->tree, dif, 0, 0, GFP_KERNEL);
 510		up_write(&ctx->devs->rwsem);
 511		if (ret < 0) {
 512			kfree(dif->path);
 513			kfree(dif);
 514			return ret;
 515		}
 516		++ctx->devs->extra_devices;
 517		break;
 518#ifdef CONFIG_EROFS_FS_ONDEMAND
 519	case Opt_fsid:
 520		kfree(ctx->fsid);
 521		ctx->fsid = kstrdup(param->string, GFP_KERNEL);
 522		if (!ctx->fsid)
 523			return -ENOMEM;
 524		break;
 525	case Opt_domain_id:
 526		kfree(ctx->domain_id);
 527		ctx->domain_id = kstrdup(param->string, GFP_KERNEL);
 528		if (!ctx->domain_id)
 529			return -ENOMEM;
 530		break;
 531#else
 532	case Opt_fsid:
 533	case Opt_domain_id:
 534		errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
 535		break;
 536#endif
 537	default:
 538		return -ENOPARAM;
 539	}
 540	return 0;
 541}
 542
 543static struct inode *erofs_nfs_get_inode(struct super_block *sb,
 544					 u64 ino, u32 generation)
 545{
 546	return erofs_iget(sb, ino);
 547}
 548
 549static struct dentry *erofs_fh_to_dentry(struct super_block *sb,
 550		struct fid *fid, int fh_len, int fh_type)
 551{
 552	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
 553				    erofs_nfs_get_inode);
 554}
 555
 556static struct dentry *erofs_fh_to_parent(struct super_block *sb,
 557		struct fid *fid, int fh_len, int fh_type)
 558{
 559	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
 560				    erofs_nfs_get_inode);
 561}
 562
 563static struct dentry *erofs_get_parent(struct dentry *child)
 564{
 565	erofs_nid_t nid;
 566	unsigned int d_type;
 567	int err;
 568
 569	err = erofs_namei(d_inode(child), &dotdot_name, &nid, &d_type);
 570	if (err)
 571		return ERR_PTR(err);
 572	return d_obtain_alias(erofs_iget(child->d_sb, nid));
 573}
 574
 575static const struct export_operations erofs_export_ops = {
 576	.encode_fh = generic_encode_ino32_fh,
 577	.fh_to_dentry = erofs_fh_to_dentry,
 578	.fh_to_parent = erofs_fh_to_parent,
 579	.get_parent = erofs_get_parent,
 580};
 581
 582static int erofs_fc_fill_pseudo_super(struct super_block *sb, struct fs_context *fc)
 583{
 584	static const struct tree_descr empty_descr = {""};
 585
 586	return simple_fill_super(sb, EROFS_SUPER_MAGIC, &empty_descr);
 587}
 588
 589static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
 590{
 591	struct inode *inode;
 592	struct erofs_sb_info *sbi;
 593	struct erofs_fs_context *ctx = fc->fs_private;
 594	int err;
 595
 596	sb->s_magic = EROFS_SUPER_MAGIC;
 597	sb->s_flags |= SB_RDONLY | SB_NOATIME;
 598	sb->s_maxbytes = MAX_LFS_FILESIZE;
 599	sb->s_op = &erofs_sops;
 600
 601	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
 602	if (!sbi)
 603		return -ENOMEM;
 604
 605	sb->s_fs_info = sbi;
 606	sbi->opt = ctx->opt;
 607	sbi->devs = ctx->devs;
 608	ctx->devs = NULL;
 609	sbi->fsid = ctx->fsid;
 610	ctx->fsid = NULL;
 611	sbi->domain_id = ctx->domain_id;
 612	ctx->domain_id = NULL;
 613
 614	sbi->blkszbits = PAGE_SHIFT;
 615	if (erofs_is_fscache_mode(sb)) {
 616		sb->s_blocksize = PAGE_SIZE;
 617		sb->s_blocksize_bits = PAGE_SHIFT;
 618
 619		err = erofs_fscache_register_fs(sb);
 620		if (err)
 621			return err;
 622
 623		err = super_setup_bdi(sb);
 624		if (err)
 625			return err;
 626	} else {
 627		if (!sb_set_blocksize(sb, PAGE_SIZE)) {
 628			errorfc(fc, "failed to set initial blksize");
 629			return -EINVAL;
 630		}
 631
 632		sbi->dax_dev = fs_dax_get_by_bdev(sb->s_bdev,
 633						  &sbi->dax_part_off,
 634						  NULL, NULL);
 635	}
 636
 637	err = erofs_read_superblock(sb);
 638	if (err)
 639		return err;
 640
 641	if (sb->s_blocksize_bits != sbi->blkszbits) {
 642		if (erofs_is_fscache_mode(sb)) {
 643			errorfc(fc, "unsupported blksize for fscache mode");
 644			return -EINVAL;
 645		}
 646		if (!sb_set_blocksize(sb, 1 << sbi->blkszbits)) {
 647			errorfc(fc, "failed to set erofs blksize");
 648			return -EINVAL;
 649		}
 650	}
 651
 652	if (test_opt(&sbi->opt, DAX_ALWAYS)) {
 653		if (!sbi->dax_dev) {
 654			errorfc(fc, "DAX unsupported by block device. Turning off DAX.");
 655			clear_opt(&sbi->opt, DAX_ALWAYS);
 656		} else if (sbi->blkszbits != PAGE_SHIFT) {
 657			errorfc(fc, "unsupported blocksize for DAX");
 658			clear_opt(&sbi->opt, DAX_ALWAYS);
 659		}
 660	}
 661
 662	sb->s_time_gran = 1;
 663	sb->s_xattr = erofs_xattr_handlers;
 664	sb->s_export_op = &erofs_export_ops;
 665
 666	if (test_opt(&sbi->opt, POSIX_ACL))
 667		sb->s_flags |= SB_POSIXACL;
 668	else
 669		sb->s_flags &= ~SB_POSIXACL;
 670
 671#ifdef CONFIG_EROFS_FS_ZIP
 672	xa_init(&sbi->managed_pslots);
 673#endif
 674
 675	inode = erofs_iget(sb, sbi->root_nid);
 676	if (IS_ERR(inode))
 677		return PTR_ERR(inode);
 678
 679	if (!S_ISDIR(inode->i_mode)) {
 680		erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
 681			  sbi->root_nid, inode->i_mode);
 682		iput(inode);
 683		return -EINVAL;
 684	}
 685
 686	sb->s_root = d_make_root(inode);
 687	if (!sb->s_root)
 688		return -ENOMEM;
 689
 690	erofs_shrinker_register(sb);
 691	if (erofs_sb_has_fragments(sbi) && sbi->packed_nid) {
 692		sbi->packed_inode = erofs_iget(sb, sbi->packed_nid);
 693		if (IS_ERR(sbi->packed_inode)) {
 694			err = PTR_ERR(sbi->packed_inode);
 695			sbi->packed_inode = NULL;
 696			return err;
 697		}
 698	}
 699	err = erofs_init_managed_cache(sb);
 700	if (err)
 701		return err;
 702
 703	err = erofs_xattr_prefixes_init(sb);
 704	if (err)
 705		return err;
 706
 707	err = erofs_register_sysfs(sb);
 708	if (err)
 709		return err;
 710
 711	erofs_info(sb, "mounted with root inode @ nid %llu.", sbi->root_nid);
 712	return 0;
 713}
 714
 715static int erofs_fc_anon_get_tree(struct fs_context *fc)
 716{
 717	return get_tree_nodev(fc, erofs_fc_fill_pseudo_super);
 718}
 719
 720static int erofs_fc_get_tree(struct fs_context *fc)
 721{
 722	struct erofs_fs_context *ctx = fc->fs_private;
 723
 724	if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && ctx->fsid)
 725		return get_tree_nodev(fc, erofs_fc_fill_super);
 726
 727	return get_tree_bdev(fc, erofs_fc_fill_super);
 728}
 729
 730static int erofs_fc_reconfigure(struct fs_context *fc)
 731{
 732	struct super_block *sb = fc->root->d_sb;
 733	struct erofs_sb_info *sbi = EROFS_SB(sb);
 734	struct erofs_fs_context *ctx = fc->fs_private;
 735
 736	DBG_BUGON(!sb_rdonly(sb));
 737
 738	if (ctx->fsid || ctx->domain_id)
 739		erofs_info(sb, "ignoring reconfiguration for fsid|domain_id.");
 740
 741	if (test_opt(&ctx->opt, POSIX_ACL))
 742		fc->sb_flags |= SB_POSIXACL;
 743	else
 744		fc->sb_flags &= ~SB_POSIXACL;
 745
 746	sbi->opt = ctx->opt;
 747
 748	fc->sb_flags |= SB_RDONLY;
 749	return 0;
 750}
 751
 752static int erofs_release_device_info(int id, void *ptr, void *data)
 753{
 754	struct erofs_device_info *dif = ptr;
 755
 756	fs_put_dax(dif->dax_dev, NULL);
 757	if (dif->bdev_handle)
 758		bdev_release(dif->bdev_handle);
 759	erofs_fscache_unregister_cookie(dif->fscache);
 760	dif->fscache = NULL;
 761	kfree(dif->path);
 762	kfree(dif);
 763	return 0;
 764}
 765
 766static void erofs_free_dev_context(struct erofs_dev_context *devs)
 767{
 768	if (!devs)
 769		return;
 770	idr_for_each(&devs->tree, &erofs_release_device_info, NULL);
 771	idr_destroy(&devs->tree);
 772	kfree(devs);
 773}
 774
 775static void erofs_fc_free(struct fs_context *fc)
 776{
 777	struct erofs_fs_context *ctx = fc->fs_private;
 
 
 
 778
 779	erofs_free_dev_context(ctx->devs);
 780	kfree(ctx->fsid);
 781	kfree(ctx->domain_id);
 782	kfree(ctx);
 783}
 784
 785static const struct fs_context_operations erofs_context_ops = {
 786	.parse_param	= erofs_fc_parse_param,
 787	.get_tree       = erofs_fc_get_tree,
 788	.reconfigure    = erofs_fc_reconfigure,
 789	.free		= erofs_fc_free,
 790};
 791
 792static const struct fs_context_operations erofs_anon_context_ops = {
 793	.get_tree       = erofs_fc_anon_get_tree,
 794};
 795
 796static int erofs_init_fs_context(struct fs_context *fc)
 797{
 798	struct erofs_fs_context *ctx;
 799
 800	/* pseudo mount for anon inodes */
 801	if (fc->sb_flags & SB_KERNMOUNT) {
 802		fc->ops = &erofs_anon_context_ops;
 803		return 0;
 804	}
 805
 806	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
 807	if (!ctx)
 808		return -ENOMEM;
 809	ctx->devs = kzalloc(sizeof(struct erofs_dev_context), GFP_KERNEL);
 810	if (!ctx->devs) {
 811		kfree(ctx);
 
 812		return -ENOMEM;
 813	}
 814	fc->fs_private = ctx;
 815
 816	idr_init(&ctx->devs->tree);
 817	init_rwsem(&ctx->devs->rwsem);
 818	erofs_default_options(ctx);
 819	fc->ops = &erofs_context_ops;
 820	return 0;
 821}
 822
 823static void erofs_kill_sb(struct super_block *sb)
 824{
 825	struct erofs_sb_info *sbi;
 826
 827	/* pseudo mount for anon inodes */
 828	if (sb->s_flags & SB_KERNMOUNT) {
 829		kill_anon_super(sb);
 830		return;
 831	}
 832
 833	if (erofs_is_fscache_mode(sb))
 834		kill_anon_super(sb);
 835	else
 836		kill_block_super(sb);
 837
 838	sbi = EROFS_SB(sb);
 839	if (!sbi)
 840		return;
 841
 842	erofs_free_dev_context(sbi->devs);
 843	fs_put_dax(sbi->dax_dev, NULL);
 844	erofs_fscache_unregister_fs(sb);
 845	kfree(sbi->fsid);
 846	kfree(sbi->domain_id);
 847	kfree(sbi);
 848	sb->s_fs_info = NULL;
 849}
 850
 851static void erofs_put_super(struct super_block *sb)
 852{
 853	struct erofs_sb_info *const sbi = EROFS_SB(sb);
 854
 855	DBG_BUGON(!sbi);
 856
 857	erofs_unregister_sysfs(sb);
 858	erofs_shrinker_unregister(sb);
 859	erofs_xattr_prefixes_cleanup(sb);
 860#ifdef CONFIG_EROFS_FS_ZIP
 861	iput(sbi->managed_cache);
 862	sbi->managed_cache = NULL;
 863#endif
 864	iput(sbi->packed_inode);
 865	sbi->packed_inode = NULL;
 866	erofs_free_dev_context(sbi->devs);
 867	sbi->devs = NULL;
 868	erofs_fscache_unregister_fs(sb);
 869}
 870
 871struct file_system_type erofs_fs_type = {
 872	.owner          = THIS_MODULE,
 873	.name           = "erofs",
 874	.init_fs_context = erofs_init_fs_context,
 875	.kill_sb        = erofs_kill_sb,
 876	.fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
 877};
 878MODULE_ALIAS_FS("erofs");
 879
 880static int __init erofs_module_init(void)
 881{
 882	int err;
 883
 884	erofs_check_ondisk_layout_definitions();
 885
 886	erofs_inode_cachep = kmem_cache_create("erofs_inode",
 887			sizeof(struct erofs_inode), 0,
 888			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD | SLAB_ACCOUNT,
 889			erofs_inode_init_once);
 890	if (!erofs_inode_cachep)
 891		return -ENOMEM;
 892
 893	err = erofs_init_shrinker();
 894	if (err)
 895		goto shrinker_err;
 896
 897	err = z_erofs_lzma_init();
 898	if (err)
 899		goto lzma_err;
 900
 901	err = z_erofs_deflate_init();
 902	if (err)
 903		goto deflate_err;
 904
 905	erofs_pcpubuf_init();
 906	err = z_erofs_init_zip_subsystem();
 907	if (err)
 908		goto zip_err;
 909
 910	err = erofs_init_sysfs();
 911	if (err)
 912		goto sysfs_err;
 913
 914	err = register_filesystem(&erofs_fs_type);
 915	if (err)
 916		goto fs_err;
 917
 918	return 0;
 919
 920fs_err:
 921	erofs_exit_sysfs();
 922sysfs_err:
 923	z_erofs_exit_zip_subsystem();
 924zip_err:
 925	z_erofs_deflate_exit();
 926deflate_err:
 927	z_erofs_lzma_exit();
 928lzma_err:
 929	erofs_exit_shrinker();
 930shrinker_err:
 931	kmem_cache_destroy(erofs_inode_cachep);
 932	return err;
 933}
 934
 935static void __exit erofs_module_exit(void)
 936{
 937	unregister_filesystem(&erofs_fs_type);
 938
 939	/* Ensure all RCU free inodes / pclusters are safe to be destroyed. */
 940	rcu_barrier();
 941
 942	erofs_exit_sysfs();
 943	z_erofs_exit_zip_subsystem();
 944	z_erofs_deflate_exit();
 945	z_erofs_lzma_exit();
 946	erofs_exit_shrinker();
 947	kmem_cache_destroy(erofs_inode_cachep);
 948	erofs_pcpubuf_exit();
 949}
 950
 951static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
 952{
 953	struct super_block *sb = dentry->d_sb;
 954	struct erofs_sb_info *sbi = EROFS_SB(sb);
 955	u64 id = 0;
 956
 957	if (!erofs_is_fscache_mode(sb))
 958		id = huge_encode_dev(sb->s_bdev->bd_dev);
 959
 960	buf->f_type = sb->s_magic;
 961	buf->f_bsize = sb->s_blocksize;
 962	buf->f_blocks = sbi->total_blocks;
 963	buf->f_bfree = buf->f_bavail = 0;
 964
 965	buf->f_files = ULLONG_MAX;
 966	buf->f_ffree = ULLONG_MAX - sbi->inos;
 967
 968	buf->f_namelen = EROFS_NAME_LEN;
 969
 970	buf->f_fsid    = u64_to_fsid(id);
 971	return 0;
 972}
 973
 974static int erofs_show_options(struct seq_file *seq, struct dentry *root)
 975{
 976	struct erofs_sb_info *sbi = EROFS_SB(root->d_sb);
 977	struct erofs_mount_opts *opt = &sbi->opt;
 978
 979#ifdef CONFIG_EROFS_FS_XATTR
 980	if (test_opt(opt, XATTR_USER))
 981		seq_puts(seq, ",user_xattr");
 982	else
 983		seq_puts(seq, ",nouser_xattr");
 984#endif
 985#ifdef CONFIG_EROFS_FS_POSIX_ACL
 986	if (test_opt(opt, POSIX_ACL))
 987		seq_puts(seq, ",acl");
 988	else
 989		seq_puts(seq, ",noacl");
 990#endif
 991#ifdef CONFIG_EROFS_FS_ZIP
 992	if (opt->cache_strategy == EROFS_ZIP_CACHE_DISABLED)
 993		seq_puts(seq, ",cache_strategy=disabled");
 994	else if (opt->cache_strategy == EROFS_ZIP_CACHE_READAHEAD)
 995		seq_puts(seq, ",cache_strategy=readahead");
 996	else if (opt->cache_strategy == EROFS_ZIP_CACHE_READAROUND)
 997		seq_puts(seq, ",cache_strategy=readaround");
 998#endif
 999	if (test_opt(opt, DAX_ALWAYS))
1000		seq_puts(seq, ",dax=always");
1001	if (test_opt(opt, DAX_NEVER))
1002		seq_puts(seq, ",dax=never");
1003#ifdef CONFIG_EROFS_FS_ONDEMAND
1004	if (sbi->fsid)
1005		seq_printf(seq, ",fsid=%s", sbi->fsid);
1006	if (sbi->domain_id)
1007		seq_printf(seq, ",domain_id=%s", sbi->domain_id);
1008#endif
1009	return 0;
1010}
1011
1012const struct super_operations erofs_sops = {
1013	.put_super = erofs_put_super,
1014	.alloc_inode = erofs_alloc_inode,
1015	.free_inode = erofs_free_inode,
1016	.statfs = erofs_statfs,
1017	.show_options = erofs_show_options,
1018};
1019
1020module_init(erofs_module_init);
1021module_exit(erofs_module_exit);
1022
1023MODULE_DESCRIPTION("Enhanced ROM File System");
1024MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
1025MODULE_LICENSE("GPL");