1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 *  linux/fs/ext4/super.c
   4 *
   5 * Copyright (C) 1992, 1993, 1994, 1995
   6 * Remy Card (card@masi.ibp.fr)
   7 * Laboratoire MASI - Institut Blaise Pascal
   8 * Universite Pierre et Marie Curie (Paris VI)
   9 *
  10 *  from
  11 *
  12 *  linux/fs/minix/inode.c
  13 *
  14 *  Copyright (C) 1991, 1992  Linus Torvalds
  15 *
  16 *  Big-endian to little-endian byte-swapping/bitmaps by
  17 *        David S. Miller (davem@caip.rutgers.edu), 1995
  18 */
  19
  20#include <linux/module.h>
  21#include <linux/string.h>
  22#include <linux/fs.h>
  23#include <linux/time.h>
  24#include <linux/vmalloc.h>
  25#include <linux/slab.h>
  26#include <linux/init.h>
  27#include <linux/blkdev.h>
  28#include <linux/backing-dev.h>
  29#include <linux/parser.h>
  30#include <linux/buffer_head.h>
  31#include <linux/exportfs.h>
  32#include <linux/vfs.h>
  33#include <linux/random.h>
  34#include <linux/mount.h>
  35#include <linux/namei.h>
  36#include <linux/quotaops.h>
  37#include <linux/seq_file.h>
  38#include <linux/ctype.h>
  39#include <linux/log2.h>
  40#include <linux/crc16.h>
  41#include <linux/dax.h>
 
  42#include <linux/uaccess.h>
  43#include <linux/iversion.h>
  44#include <linux/unicode.h>
  45#include <linux/part_stat.h>
  46#include <linux/kthread.h>
  47#include <linux/freezer.h>
  48#include <linux/fsnotify.h>
  49#include <linux/fs_context.h>
  50#include <linux/fs_parser.h>
  51
  52#include "ext4.h"
  53#include "ext4_extents.h"	/* Needed for trace points definition */
  54#include "ext4_jbd2.h"
  55#include "xattr.h"
  56#include "acl.h"
  57#include "mballoc.h"
  58#include "fsmap.h"
  59
  60#define CREATE_TRACE_POINTS
  61#include <trace/events/ext4.h>
  62
  63static struct ext4_lazy_init *ext4_li_info;
  64static DEFINE_MUTEX(ext4_li_mtx);
  65static struct ratelimit_state ext4_mount_msg_ratelimit;
  66
  67static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
  68			     unsigned long journal_devnum);
  69static int ext4_show_options(struct seq_file *seq, struct dentry *root);
  70static void ext4_update_super(struct super_block *sb);
  71static int ext4_commit_super(struct super_block *sb);
  72static int ext4_mark_recovery_complete(struct super_block *sb,
  73					struct ext4_super_block *es);
  74static int ext4_clear_journal_err(struct super_block *sb,
  75				  struct ext4_super_block *es);
  76static int ext4_sync_fs(struct super_block *sb, int wait);
 
  77static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
  78static int ext4_unfreeze(struct super_block *sb);
  79static int ext4_freeze(struct super_block *sb);
 
 
  80static inline int ext2_feature_set_ok(struct super_block *sb);
  81static inline int ext3_feature_set_ok(struct super_block *sb);
 
  82static void ext4_destroy_lazyinit_thread(void);
  83static void ext4_unregister_li_request(struct super_block *sb);
  84static void ext4_clear_request_list(void);
  85static struct inode *ext4_get_journal_inode(struct super_block *sb,
  86					    unsigned int journal_inum);
  87static int ext4_validate_options(struct fs_context *fc);
  88static int ext4_check_opt_consistency(struct fs_context *fc,
  89				      struct super_block *sb);
  90static void ext4_apply_options(struct fs_context *fc, struct super_block *sb);
  91static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param);
  92static int ext4_get_tree(struct fs_context *fc);
  93static int ext4_reconfigure(struct fs_context *fc);
  94static void ext4_fc_free(struct fs_context *fc);
  95static int ext4_init_fs_context(struct fs_context *fc);
  96static const struct fs_parameter_spec ext4_param_specs[];
  97
  98/*
  99 * Lock ordering
 100 *
 
 
 
 101 * page fault path:
 102 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start
 103 *   -> page lock -> i_data_sem (rw)
 104 *
 105 * buffered write path:
 106 * sb_start_write -> i_mutex -> mmap_lock
 107 * sb_start_write -> i_mutex -> transaction start -> page lock ->
 108 *   i_data_sem (rw)
 109 *
 110 * truncate:
 111 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) ->
 112 *   page lock
 113 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start ->
 114 *   i_data_sem (rw)
 115 *
 116 * direct IO:
 117 * sb_start_write -> i_mutex -> mmap_lock
 118 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
 119 *
 120 * writepages:
 121 * transaction start -> page lock(s) -> i_data_sem (rw)
 122 */
 123
 124static const struct fs_context_operations ext4_context_ops = {
 125	.parse_param	= ext4_parse_param,
 126	.get_tree	= ext4_get_tree,
 127	.reconfigure	= ext4_reconfigure,
 128	.free		= ext4_fc_free,
 129};
 130
 131
 132#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
 133static struct file_system_type ext2_fs_type = {
 134	.owner			= THIS_MODULE,
 135	.name			= "ext2",
 136	.init_fs_context	= ext4_init_fs_context,
 137	.parameters		= ext4_param_specs,
 138	.kill_sb		= kill_block_super,
 139	.fs_flags		= FS_REQUIRES_DEV,
 140};
 141MODULE_ALIAS_FS("ext2");
 142MODULE_ALIAS("ext2");
 143#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
 144#else
 145#define IS_EXT2_SB(sb) (0)
 146#endif
 147
 148
 149static struct file_system_type ext3_fs_type = {
 150	.owner			= THIS_MODULE,
 151	.name			= "ext3",
 152	.init_fs_context	= ext4_init_fs_context,
 153	.parameters		= ext4_param_specs,
 154	.kill_sb		= kill_block_super,
 155	.fs_flags		= FS_REQUIRES_DEV,
 156};
 157MODULE_ALIAS_FS("ext3");
 158MODULE_ALIAS("ext3");
 159#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
 160
 161
 162static inline void __ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags,
 163				  bh_end_io_t *end_io)
 164{
 165	/*
 166	 * buffer's verified bit is no longer valid after reading from
 167	 * disk again due to write out error, clear it to make sure we
 168	 * recheck the buffer contents.
 169	 */
 170	clear_buffer_verified(bh);
 171
 172	bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
 173	get_bh(bh);
 174	submit_bh(REQ_OP_READ | op_flags, bh);
 175}
 176
 177void ext4_read_bh_nowait(struct buffer_head *bh, blk_opf_t op_flags,
 178			 bh_end_io_t *end_io)
 179{
 180	BUG_ON(!buffer_locked(bh));
 181
 182	if (ext4_buffer_uptodate(bh)) {
 183		unlock_buffer(bh);
 184		return;
 185	}
 186	__ext4_read_bh(bh, op_flags, end_io);
 187}
 188
 189int ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags, bh_end_io_t *end_io)
 190{
 191	BUG_ON(!buffer_locked(bh));
 192
 193	if (ext4_buffer_uptodate(bh)) {
 194		unlock_buffer(bh);
 195		return 0;
 196	}
 197
 198	__ext4_read_bh(bh, op_flags, end_io);
 199
 200	wait_on_buffer(bh);
 201	if (buffer_uptodate(bh))
 202		return 0;
 203	return -EIO;
 204}
 205
 206int ext4_read_bh_lock(struct buffer_head *bh, blk_opf_t op_flags, bool wait)
 207{
 208	lock_buffer(bh);
 209	if (!wait) {
 
 210		ext4_read_bh_nowait(bh, op_flags, NULL);
 211		return 0;
 212	}
 213	return ext4_read_bh(bh, op_flags, NULL);
 
 
 
 
 
 
 214}
 215
 216/*
 217 * This works like __bread_gfp() except it uses ERR_PTR for error
 218 * returns.  Currently with sb_bread it's impossible to distinguish
 219 * between ENOMEM and EIO situations (since both result in a NULL
 220 * return.
 221 */
 222static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
 223					       sector_t block,
 224					       blk_opf_t op_flags, gfp_t gfp)
 225{
 226	struct buffer_head *bh;
 227	int ret;
 228
 229	bh = sb_getblk_gfp(sb, block, gfp);
 230	if (bh == NULL)
 231		return ERR_PTR(-ENOMEM);
 232	if (ext4_buffer_uptodate(bh))
 233		return bh;
 234
 235	ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
 236	if (ret) {
 237		put_bh(bh);
 238		return ERR_PTR(ret);
 239	}
 240	return bh;
 241}
 242
 243struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
 244				   blk_opf_t op_flags)
 245{
 246	return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
 247}
 248
 249struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
 250					    sector_t block)
 251{
 252	return __ext4_sb_bread_gfp(sb, block, 0, 0);
 253}
 254
 255void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
 256{
 257	struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
 258
 259	if (likely(bh)) {
 260		if (trylock_buffer(bh))
 261			ext4_read_bh_nowait(bh, REQ_RAHEAD, NULL);
 262		brelse(bh);
 263	}
 264}
 265
 266static int ext4_verify_csum_type(struct super_block *sb,
 267				 struct ext4_super_block *es)
 268{
 269	if (!ext4_has_feature_metadata_csum(sb))
 270		return 1;
 271
 272	return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
 273}
 274
 275__le32 ext4_superblock_csum(struct super_block *sb,
 276			    struct ext4_super_block *es)
 277{
 278	struct ext4_sb_info *sbi = EXT4_SB(sb);
 279	int offset = offsetof(struct ext4_super_block, s_checksum);
 280	__u32 csum;
 281
 282	csum = ext4_chksum(sbi, ~0, (char *)es, offset);
 283
 284	return cpu_to_le32(csum);
 285}
 286
 287static int ext4_superblock_csum_verify(struct super_block *sb,
 288				       struct ext4_super_block *es)
 289{
 290	if (!ext4_has_metadata_csum(sb))
 291		return 1;
 292
 293	return es->s_checksum == ext4_superblock_csum(sb, es);
 294}
 295
 296void ext4_superblock_csum_set(struct super_block *sb)
 297{
 298	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 299
 300	if (!ext4_has_metadata_csum(sb))
 301		return;
 302
 303	es->s_checksum = ext4_superblock_csum(sb, es);
 304}
 305
 306ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
 307			       struct ext4_group_desc *bg)
 308{
 309	return le32_to_cpu(bg->bg_block_bitmap_lo) |
 310		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 311		 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
 312}
 313
 314ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
 315			       struct ext4_group_desc *bg)
 316{
 317	return le32_to_cpu(bg->bg_inode_bitmap_lo) |
 318		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 319		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
 320}
 321
 322ext4_fsblk_t ext4_inode_table(struct super_block *sb,
 323			      struct ext4_group_desc *bg)
 324{
 325	return le32_to_cpu(bg->bg_inode_table_lo) |
 326		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 327		 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
 328}
 329
 330__u32 ext4_free_group_clusters(struct super_block *sb,
 331			       struct ext4_group_desc *bg)
 332{
 333	return le16_to_cpu(bg->bg_free_blocks_count_lo) |
 334		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 335		 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
 336}
 337
 338__u32 ext4_free_inodes_count(struct super_block *sb,
 339			      struct ext4_group_desc *bg)
 340{
 341	return le16_to_cpu(bg->bg_free_inodes_count_lo) |
 342		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 343		 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
 344}
 345
 346__u32 ext4_used_dirs_count(struct super_block *sb,
 347			      struct ext4_group_desc *bg)
 348{
 349	return le16_to_cpu(bg->bg_used_dirs_count_lo) |
 350		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 351		 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
 352}
 353
 354__u32 ext4_itable_unused_count(struct super_block *sb,
 355			      struct ext4_group_desc *bg)
 356{
 357	return le16_to_cpu(bg->bg_itable_unused_lo) |
 358		(EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
 359		 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
 360}
 361
 362void ext4_block_bitmap_set(struct super_block *sb,
 363			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
 364{
 365	bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
 366	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 367		bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
 368}
 369
 370void ext4_inode_bitmap_set(struct super_block *sb,
 371			   struct ext4_group_desc *bg, ext4_fsblk_t blk)
 372{
 373	bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
 374	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 375		bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
 376}
 377
 378void ext4_inode_table_set(struct super_block *sb,
 379			  struct ext4_group_desc *bg, ext4_fsblk_t blk)
 380{
 381	bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
 382	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 383		bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
 384}
 385
 386void ext4_free_group_clusters_set(struct super_block *sb,
 387				  struct ext4_group_desc *bg, __u32 count)
 388{
 389	bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
 390	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 391		bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
 392}
 393
 394void ext4_free_inodes_set(struct super_block *sb,
 395			  struct ext4_group_desc *bg, __u32 count)
 396{
 397	bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
 398	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 399		bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
 400}
 401
 402void ext4_used_dirs_set(struct super_block *sb,
 403			  struct ext4_group_desc *bg, __u32 count)
 404{
 405	bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
 406	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 407		bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
 408}
 409
 410void ext4_itable_unused_set(struct super_block *sb,
 411			  struct ext4_group_desc *bg, __u32 count)
 412{
 413	bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
 414	if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
 415		bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
 416}
 417
 418static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
 419{
 420	now = clamp_val(now, 0, (1ull << 40) - 1);
 421
 422	*lo = cpu_to_le32(lower_32_bits(now));
 423	*hi = upper_32_bits(now);
 424}
 425
 426static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
 427{
 428	return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
 429}
 430#define ext4_update_tstamp(es, tstamp) \
 431	__ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
 432			     ktime_get_real_seconds())
 433#define ext4_get_tstamp(es, tstamp) \
 434	__ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
 435
 436/*
 437 * The del_gendisk() function uninitializes the disk-specific data
 438 * structures, including the bdi structure, without telling anyone
 439 * else.  Once this happens, any attempt to call mark_buffer_dirty()
 440 * (for example, by ext4_commit_super), will cause a kernel OOPS.
 441 * This is a kludge to prevent these oops until we can put in a proper
 442 * hook in del_gendisk() to inform the VFS and file system layers.
 443 */
 444static int block_device_ejected(struct super_block *sb)
 445{
 446	struct inode *bd_inode = sb->s_bdev->bd_inode;
 447	struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
 448
 449	return bdi->dev == NULL;
 450}
 451
 452static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
 453{
 454	struct super_block		*sb = journal->j_private;
 455	struct ext4_sb_info		*sbi = EXT4_SB(sb);
 456	int				error = is_journal_aborted(journal);
 457	struct ext4_journal_cb_entry	*jce;
 458
 459	BUG_ON(txn->t_state == T_FINISHED);
 460
 461	ext4_process_freed_data(sb, txn->t_tid);
 462
 463	spin_lock(&sbi->s_md_lock);
 464	while (!list_empty(&txn->t_private_list)) {
 465		jce = list_entry(txn->t_private_list.next,
 466				 struct ext4_journal_cb_entry, jce_list);
 467		list_del_init(&jce->jce_list);
 468		spin_unlock(&sbi->s_md_lock);
 469		jce->jce_func(sb, jce, error);
 470		spin_lock(&sbi->s_md_lock);
 471	}
 472	spin_unlock(&sbi->s_md_lock);
 473}
 474
 475/*
 476 * This writepage callback for write_cache_pages()
 477 * takes care of a few cases after page cleaning.
 478 *
 479 * write_cache_pages() already checks for dirty pages
 480 * and calls clear_page_dirty_for_io(), which we want,
 481 * to write protect the pages.
 482 *
 483 * However, we may have to redirty a page (see below.)
 484 */
 485static int ext4_journalled_writepage_callback(struct page *page,
 486					      struct writeback_control *wbc,
 487					      void *data)
 488{
 489	transaction_t *transaction = (transaction_t *) data;
 490	struct buffer_head *bh, *head;
 491	struct journal_head *jh;
 492
 493	bh = head = page_buffers(page);
 494	do {
 495		/*
 496		 * We have to redirty a page in these cases:
 497		 * 1) If buffer is dirty, it means the page was dirty because it
 498		 * contains a buffer that needs checkpointing. So the dirty bit
 499		 * needs to be preserved so that checkpointing writes the buffer
 500		 * properly.
 501		 * 2) If buffer is not part of the committing transaction
 502		 * (we may have just accidentally come across this buffer because
 503		 * inode range tracking is not exact) or if the currently running
 504		 * transaction already contains this buffer as well, dirty bit
 505		 * needs to be preserved so that the buffer gets writeprotected
 506		 * properly on running transaction's commit.
 507		 */
 508		jh = bh2jh(bh);
 509		if (buffer_dirty(bh) ||
 510		    (jh && (jh->b_transaction != transaction ||
 511			    jh->b_next_transaction))) {
 512			redirty_page_for_writepage(wbc, page);
 513			goto out;
 514		}
 515	} while ((bh = bh->b_this_page) != head);
 516
 517out:
 518	return AOP_WRITEPAGE_ACTIVATE;
 519}
 520
 521static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
 522{
 523	struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
 524	struct writeback_control wbc = {
 525		.sync_mode =  WB_SYNC_ALL,
 526		.nr_to_write = LONG_MAX,
 527		.range_start = jinode->i_dirty_start,
 528		.range_end = jinode->i_dirty_end,
 529        };
 530
 531	return write_cache_pages(mapping, &wbc,
 532				 ext4_journalled_writepage_callback,
 533				 jinode->i_transaction);
 534}
 535
 536static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
 537{
 538	int ret;
 539
 540	if (ext4_should_journal_data(jinode->i_vfs_inode))
 541		ret = ext4_journalled_submit_inode_data_buffers(jinode);
 542	else
 543		ret = ext4_normal_submit_inode_data_buffers(jinode);
 
 544	return ret;
 545}
 546
 547static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
 548{
 549	int ret = 0;
 550
 551	if (!ext4_should_journal_data(jinode->i_vfs_inode))
 552		ret = jbd2_journal_finish_inode_data_buffers(jinode);
 553
 554	return ret;
 555}
 556
 557static bool system_going_down(void)
 558{
 559	return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
 560		|| system_state == SYSTEM_RESTART;
 561}
 562
 563struct ext4_err_translation {
 564	int code;
 565	int errno;
 566};
 567
 568#define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
 569
 570static struct ext4_err_translation err_translation[] = {
 571	EXT4_ERR_TRANSLATE(EIO),
 572	EXT4_ERR_TRANSLATE(ENOMEM),
 573	EXT4_ERR_TRANSLATE(EFSBADCRC),
 574	EXT4_ERR_TRANSLATE(EFSCORRUPTED),
 575	EXT4_ERR_TRANSLATE(ENOSPC),
 576	EXT4_ERR_TRANSLATE(ENOKEY),
 577	EXT4_ERR_TRANSLATE(EROFS),
 578	EXT4_ERR_TRANSLATE(EFBIG),
 579	EXT4_ERR_TRANSLATE(EEXIST),
 580	EXT4_ERR_TRANSLATE(ERANGE),
 581	EXT4_ERR_TRANSLATE(EOVERFLOW),
 582	EXT4_ERR_TRANSLATE(EBUSY),
 583	EXT4_ERR_TRANSLATE(ENOTDIR),
 584	EXT4_ERR_TRANSLATE(ENOTEMPTY),
 585	EXT4_ERR_TRANSLATE(ESHUTDOWN),
 586	EXT4_ERR_TRANSLATE(EFAULT),
 587};
 588
 589static int ext4_errno_to_code(int errno)
 590{
 591	int i;
 592
 593	for (i = 0; i < ARRAY_SIZE(err_translation); i++)
 594		if (err_translation[i].errno == errno)
 595			return err_translation[i].code;
 596	return EXT4_ERR_UNKNOWN;
 597}
 598
 599static void save_error_info(struct super_block *sb, int error,
 600			    __u32 ino, __u64 block,
 601			    const char *func, unsigned int line)
 602{
 603	struct ext4_sb_info *sbi = EXT4_SB(sb);
 604
 605	/* We default to EFSCORRUPTED error... */
 606	if (error == 0)
 607		error = EFSCORRUPTED;
 608
 609	spin_lock(&sbi->s_error_lock);
 610	sbi->s_add_error_count++;
 611	sbi->s_last_error_code = error;
 612	sbi->s_last_error_line = line;
 613	sbi->s_last_error_ino = ino;
 614	sbi->s_last_error_block = block;
 615	sbi->s_last_error_func = func;
 616	sbi->s_last_error_time = ktime_get_real_seconds();
 617	if (!sbi->s_first_error_time) {
 618		sbi->s_first_error_code = error;
 619		sbi->s_first_error_line = line;
 620		sbi->s_first_error_ino = ino;
 621		sbi->s_first_error_block = block;
 622		sbi->s_first_error_func = func;
 623		sbi->s_first_error_time = sbi->s_last_error_time;
 624	}
 625	spin_unlock(&sbi->s_error_lock);
 626}
 627
 628/* Deal with the reporting of failure conditions on a filesystem such as
 629 * inconsistencies detected or read IO failures.
 630 *
 631 * On ext2, we can store the error state of the filesystem in the
 632 * superblock.  That is not possible on ext4, because we may have other
 633 * write ordering constraints on the superblock which prevent us from
 634 * writing it out straight away; and given that the journal is about to
 635 * be aborted, we can't rely on the current, or future, transactions to
 636 * write out the superblock safely.
 637 *
 638 * We'll just use the jbd2_journal_abort() error code to record an error in
 639 * the journal instead.  On recovery, the journal will complain about
 640 * that error until we've noted it down and cleared it.
 641 *
 642 * If force_ro is set, we unconditionally force the filesystem into an
 643 * ABORT|READONLY state, unless the error response on the fs has been set to
 644 * panic in which case we take the easy way out and panic immediately. This is
 645 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
 646 * at a critical moment in log management.
 647 */
 648static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
 649			      __u32 ino, __u64 block,
 650			      const char *func, unsigned int line)
 651{
 652	journal_t *journal = EXT4_SB(sb)->s_journal;
 653	bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
 654
 655	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
 656	if (test_opt(sb, WARN_ON_ERROR))
 657		WARN_ON_ONCE(1);
 658
 659	if (!continue_fs && !sb_rdonly(sb)) {
 660		ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
 661		if (journal)
 662			jbd2_journal_abort(journal, -EIO);
 663	}
 664
 665	if (!bdev_read_only(sb->s_bdev)) {
 666		save_error_info(sb, error, ino, block, func, line);
 667		/*
 668		 * In case the fs should keep running, we need to writeout
 669		 * superblock through the journal. Due to lock ordering
 670		 * constraints, it may not be safe to do it right here so we
 671		 * defer superblock flushing to a workqueue.
 672		 */
 673		if (continue_fs && journal)
 674			schedule_work(&EXT4_SB(sb)->s_error_work);
 675		else
 676			ext4_commit_super(sb);
 677	}
 678
 679	/*
 680	 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
 681	 * could panic during 'reboot -f' as the underlying device got already
 682	 * disabled.
 683	 */
 684	if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
 685		panic("EXT4-fs (device %s): panic forced after error\n",
 686			sb->s_id);
 687	}
 688
 689	if (sb_rdonly(sb) || continue_fs)
 690		return;
 691
 692	ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
 693	/*
 694	 * Make sure updated value of ->s_mount_flags will be visible before
 695	 * ->s_flags update
 696	 */
 697	smp_wmb();
 698	sb->s_flags |= SB_RDONLY;
 699}
 700
 701static void flush_stashed_error_work(struct work_struct *work)
 702{
 703	struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
 704						s_error_work);
 705	journal_t *journal = sbi->s_journal;
 706	handle_t *handle;
 707
 708	/*
 709	 * If the journal is still running, we have to write out superblock
 710	 * through the journal to avoid collisions of other journalled sb
 711	 * updates.
 712	 *
 713	 * We use directly jbd2 functions here to avoid recursing back into
 714	 * ext4 error handling code during handling of previous errors.
 715	 */
 716	if (!sb_rdonly(sbi->s_sb) && journal) {
 717		struct buffer_head *sbh = sbi->s_sbh;
 718		handle = jbd2_journal_start(journal, 1);
 719		if (IS_ERR(handle))
 720			goto write_directly;
 721		if (jbd2_journal_get_write_access(handle, sbh)) {
 722			jbd2_journal_stop(handle);
 723			goto write_directly;
 724		}
 725		ext4_update_super(sbi->s_sb);
 726		if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
 727			ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
 728				 "superblock detected");
 729			clear_buffer_write_io_error(sbh);
 730			set_buffer_uptodate(sbh);
 731		}
 732
 733		if (jbd2_journal_dirty_metadata(handle, sbh)) {
 734			jbd2_journal_stop(handle);
 735			goto write_directly;
 736		}
 737		jbd2_journal_stop(handle);
 738		ext4_notify_error_sysfs(sbi);
 739		return;
 740	}
 741write_directly:
 742	/*
 743	 * Write through journal failed. Write sb directly to get error info
 744	 * out and hope for the best.
 745	 */
 746	ext4_commit_super(sbi->s_sb);
 747	ext4_notify_error_sysfs(sbi);
 748}
 749
 750#define ext4_error_ratelimit(sb)					\
 751		___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),	\
 752			     "EXT4-fs error")
 753
 754void __ext4_error(struct super_block *sb, const char *function,
 755		  unsigned int line, bool force_ro, int error, __u64 block,
 756		  const char *fmt, ...)
 757{
 758	struct va_format vaf;
 759	va_list args;
 760
 761	if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
 762		return;
 763
 764	trace_ext4_error(sb, function, line);
 765	if (ext4_error_ratelimit(sb)) {
 766		va_start(args, fmt);
 767		vaf.fmt = fmt;
 768		vaf.va = &args;
 769		printk(KERN_CRIT
 770		       "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
 771		       sb->s_id, function, line, current->comm, &vaf);
 772		va_end(args);
 773	}
 774	fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED);
 775
 776	ext4_handle_error(sb, force_ro, error, 0, block, function, line);
 777}
 778
 779void __ext4_error_inode(struct inode *inode, const char *function,
 780			unsigned int line, ext4_fsblk_t block, int error,
 781			const char *fmt, ...)
 782{
 783	va_list args;
 784	struct va_format vaf;
 785
 786	if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
 787		return;
 788
 789	trace_ext4_error(inode->i_sb, function, line);
 790	if (ext4_error_ratelimit(inode->i_sb)) {
 791		va_start(args, fmt);
 792		vaf.fmt = fmt;
 793		vaf.va = &args;
 794		if (block)
 795			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
 796			       "inode #%lu: block %llu: comm %s: %pV\n",
 797			       inode->i_sb->s_id, function, line, inode->i_ino,
 798			       block, current->comm, &vaf);
 799		else
 800			printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
 801			       "inode #%lu: comm %s: %pV\n",
 802			       inode->i_sb->s_id, function, line, inode->i_ino,
 803			       current->comm, &vaf);
 804		va_end(args);
 805	}
 806	fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED);
 807
 808	ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
 809			  function, line);
 810}
 811
 812void __ext4_error_file(struct file *file, const char *function,
 813		       unsigned int line, ext4_fsblk_t block,
 814		       const char *fmt, ...)
 815{
 816	va_list args;
 817	struct va_format vaf;
 818	struct inode *inode = file_inode(file);
 819	char pathname[80], *path;
 820
 821	if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
 822		return;
 823
 824	trace_ext4_error(inode->i_sb, function, line);
 825	if (ext4_error_ratelimit(inode->i_sb)) {
 826		path = file_path(file, pathname, sizeof(pathname));
 827		if (IS_ERR(path))
 828			path = "(unknown)";
 829		va_start(args, fmt);
 830		vaf.fmt = fmt;
 831		vaf.va = &args;
 832		if (block)
 833			printk(KERN_CRIT
 834			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
 835			       "block %llu: comm %s: path %s: %pV\n",
 836			       inode->i_sb->s_id, function, line, inode->i_ino,
 837			       block, current->comm, path, &vaf);
 838		else
 839			printk(KERN_CRIT
 840			       "EXT4-fs error (device %s): %s:%d: inode #%lu: "
 841			       "comm %s: path %s: %pV\n",
 842			       inode->i_sb->s_id, function, line, inode->i_ino,
 843			       current->comm, path, &vaf);
 844		va_end(args);
 845	}
 846	fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED);
 847
 848	ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
 849			  function, line);
 850}
 851
 852const char *ext4_decode_error(struct super_block *sb, int errno,
 853			      char nbuf[16])
 854{
 855	char *errstr = NULL;
 856
 857	switch (errno) {
 858	case -EFSCORRUPTED:
 859		errstr = "Corrupt filesystem";
 860		break;
 861	case -EFSBADCRC:
 862		errstr = "Filesystem failed CRC";
 863		break;
 864	case -EIO:
 865		errstr = "IO failure";
 866		break;
 867	case -ENOMEM:
 868		errstr = "Out of memory";
 869		break;
 870	case -EROFS:
 871		if (!sb || (EXT4_SB(sb)->s_journal &&
 872			    EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
 873			errstr = "Journal has aborted";
 874		else
 875			errstr = "Readonly filesystem";
 876		break;
 877	default:
 878		/* If the caller passed in an extra buffer for unknown
 879		 * errors, textualise them now.  Else we just return
 880		 * NULL. */
 881		if (nbuf) {
 882			/* Check for truncated error codes... */
 883			if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
 884				errstr = nbuf;
 885		}
 886		break;
 887	}
 888
 889	return errstr;
 890}
 891
 892/* __ext4_std_error decodes expected errors from journaling functions
 893 * automatically and invokes the appropriate error response.  */
 894
 895void __ext4_std_error(struct super_block *sb, const char *function,
 896		      unsigned int line, int errno)
 897{
 898	char nbuf[16];
 899	const char *errstr;
 900
 901	if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
 902		return;
 903
 904	/* Special case: if the error is EROFS, and we're not already
 905	 * inside a transaction, then there's really no point in logging
 906	 * an error. */
 907	if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
 908		return;
 909
 910	if (ext4_error_ratelimit(sb)) {
 911		errstr = ext4_decode_error(sb, errno, nbuf);
 912		printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
 913		       sb->s_id, function, line, errstr);
 914	}
 915	fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED);
 916
 917	ext4_handle_error(sb, false, -errno, 0, 0, function, line);
 918}
 919
 920void __ext4_msg(struct super_block *sb,
 921		const char *prefix, const char *fmt, ...)
 922{
 923	struct va_format vaf;
 924	va_list args;
 925
 926	if (sb) {
 927		atomic_inc(&EXT4_SB(sb)->s_msg_count);
 928		if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state),
 929				  "EXT4-fs"))
 930			return;
 931	}
 932
 933	va_start(args, fmt);
 934	vaf.fmt = fmt;
 935	vaf.va = &args;
 936	if (sb)
 937		printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
 938	else
 939		printk("%sEXT4-fs: %pV\n", prefix, &vaf);
 940	va_end(args);
 941}
 942
 943static int ext4_warning_ratelimit(struct super_block *sb)
 944{
 945	atomic_inc(&EXT4_SB(sb)->s_warning_count);
 946	return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
 947			    "EXT4-fs warning");
 948}
 949
 950void __ext4_warning(struct super_block *sb, const char *function,
 951		    unsigned int line, const char *fmt, ...)
 952{
 953	struct va_format vaf;
 954	va_list args;
 955
 956	if (!ext4_warning_ratelimit(sb))
 957		return;
 958
 959	va_start(args, fmt);
 960	vaf.fmt = fmt;
 961	vaf.va = &args;
 962	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
 963	       sb->s_id, function, line, &vaf);
 964	va_end(args);
 965}
 966
 967void __ext4_warning_inode(const struct inode *inode, const char *function,
 968			  unsigned int line, const char *fmt, ...)
 969{
 970	struct va_format vaf;
 971	va_list args;
 972
 973	if (!ext4_warning_ratelimit(inode->i_sb))
 974		return;
 975
 976	va_start(args, fmt);
 977	vaf.fmt = fmt;
 978	vaf.va = &args;
 979	printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
 980	       "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
 981	       function, line, inode->i_ino, current->comm, &vaf);
 982	va_end(args);
 983}
 984
 985void __ext4_grp_locked_error(const char *function, unsigned int line,
 986			     struct super_block *sb, ext4_group_t grp,
 987			     unsigned long ino, ext4_fsblk_t block,
 988			     const char *fmt, ...)
 989__releases(bitlock)
 990__acquires(bitlock)
 991{
 992	struct va_format vaf;
 993	va_list args;
 994
 995	if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
 996		return;
 997
 998	trace_ext4_error(sb, function, line);
 999	if (ext4_error_ratelimit(sb)) {
1000		va_start(args, fmt);
1001		vaf.fmt = fmt;
1002		vaf.va = &args;
1003		printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1004		       sb->s_id, function, line, grp);
1005		if (ino)
1006			printk(KERN_CONT "inode %lu: ", ino);
1007		if (block)
1008			printk(KERN_CONT "block %llu:",
1009			       (unsigned long long) block);
1010		printk(KERN_CONT "%pV\n", &vaf);
1011		va_end(args);
1012	}
1013
1014	if (test_opt(sb, ERRORS_CONT)) {
1015		if (test_opt(sb, WARN_ON_ERROR))
1016			WARN_ON_ONCE(1);
1017		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1018		if (!bdev_read_only(sb->s_bdev)) {
1019			save_error_info(sb, EFSCORRUPTED, ino, block, function,
1020					line);
1021			schedule_work(&EXT4_SB(sb)->s_error_work);
1022		}
1023		return;
1024	}
1025	ext4_unlock_group(sb, grp);
1026	ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1027	/*
1028	 * We only get here in the ERRORS_RO case; relocking the group
1029	 * may be dangerous, but nothing bad will happen since the
1030	 * filesystem will have already been marked read/only and the
1031	 * journal has been aborted.  We return 1 as a hint to callers
1032	 * who might what to use the return value from
1033	 * ext4_grp_locked_error() to distinguish between the
1034	 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1035	 * aggressively from the ext4 function in question, with a
1036	 * more appropriate error code.
1037	 */
1038	ext4_lock_group(sb, grp);
1039	return;
1040}
1041
1042void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1043				     ext4_group_t group,
1044				     unsigned int flags)
1045{
1046	struct ext4_sb_info *sbi = EXT4_SB(sb);
1047	struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1048	struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1049	int ret;
1050
1051	if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1052		ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1053					    &grp->bb_state);
1054		if (!ret)
1055			percpu_counter_sub(&sbi->s_freeclusters_counter,
1056					   grp->bb_free);
1057	}
1058
1059	if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1060		ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1061					    &grp->bb_state);
1062		if (!ret && gdp) {
1063			int count;
1064
1065			count = ext4_free_inodes_count(sb, gdp);
1066			percpu_counter_sub(&sbi->s_freeinodes_counter,
1067					   count);
1068		}
1069	}
1070}
1071
1072void ext4_update_dynamic_rev(struct super_block *sb)
1073{
1074	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1075
1076	if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1077		return;
1078
1079	ext4_warning(sb,
1080		     "updating to rev %d because of new feature flag, "
1081		     "running e2fsck is recommended",
1082		     EXT4_DYNAMIC_REV);
1083
1084	es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1085	es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1086	es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1087	/* leave es->s_feature_*compat flags alone */
1088	/* es->s_uuid will be set by e2fsck if empty */
1089
1090	/*
1091	 * The rest of the superblock fields should be zero, and if not it
1092	 * means they are likely already in use, so leave them alone.  We
1093	 * can leave it up to e2fsck to clean up any inconsistencies there.
1094	 */
1095}
1096
1097/*
1098 * Open the external journal device
1099 */
1100static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1101{
1102	struct block_device *bdev;
1103
1104	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1105	if (IS_ERR(bdev))
1106		goto fail;
1107	return bdev;
1108
1109fail:
1110	ext4_msg(sb, KERN_ERR,
1111		 "failed to open journal device unknown-block(%u,%u) %ld",
1112		 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1113	return NULL;
1114}
1115
1116/*
1117 * Release the journal device
1118 */
1119static void ext4_blkdev_put(struct block_device *bdev)
1120{
1121	blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1122}
1123
1124static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1125{
1126	struct block_device *bdev;
1127	bdev = sbi->s_journal_bdev;
1128	if (bdev) {
1129		ext4_blkdev_put(bdev);
1130		sbi->s_journal_bdev = NULL;
1131	}
1132}
1133
1134static inline struct inode *orphan_list_entry(struct list_head *l)
1135{
1136	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1137}
1138
1139static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1140{
1141	struct list_head *l;
1142
1143	ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1144		 le32_to_cpu(sbi->s_es->s_last_orphan));
1145
1146	printk(KERN_ERR "sb_info orphan list:\n");
1147	list_for_each(l, &sbi->s_orphan) {
1148		struct inode *inode = orphan_list_entry(l);
1149		printk(KERN_ERR "  "
1150		       "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1151		       inode->i_sb->s_id, inode->i_ino, inode,
1152		       inode->i_mode, inode->i_nlink,
1153		       NEXT_ORPHAN(inode));
1154	}
1155}
1156
1157#ifdef CONFIG_QUOTA
1158static int ext4_quota_off(struct super_block *sb, int type);
1159
1160static inline void ext4_quota_off_umount(struct super_block *sb)
1161{
1162	int type;
1163
1164	/* Use our quota_off function to clear inode flags etc. */
1165	for (type = 0; type < EXT4_MAXQUOTAS; type++)
1166		ext4_quota_off(sb, type);
1167}
1168
1169/*
1170 * This is a helper function which is used in the mount/remount
1171 * codepaths (which holds s_umount) to fetch the quota file name.
1172 */
1173static inline char *get_qf_name(struct super_block *sb,
1174				struct ext4_sb_info *sbi,
1175				int type)
1176{
1177	return rcu_dereference_protected(sbi->s_qf_names[type],
1178					 lockdep_is_held(&sb->s_umount));
1179}
1180#else
1181static inline void ext4_quota_off_umount(struct super_block *sb)
1182{
1183}
1184#endif
1185
1186static void ext4_put_super(struct super_block *sb)
1187{
1188	struct ext4_sb_info *sbi = EXT4_SB(sb);
1189	struct ext4_super_block *es = sbi->s_es;
1190	struct buffer_head **group_desc;
1191	struct flex_groups **flex_groups;
1192	int aborted = 0;
1193	int i, err;
1194
 
 
 
 
 
 
1195	/*
1196	 * Unregister sysfs before destroying jbd2 journal.
1197	 * Since we could still access attr_journal_task attribute via sysfs
1198	 * path which could have sbi->s_journal->j_task as NULL
1199	 * Unregister sysfs before flush sbi->s_error_work.
1200	 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1201	 * read metadata verify failed then will queue error work.
1202	 * flush_stashed_error_work will call start_this_handle may trigger
1203	 * BUG_ON.
1204	 */
1205	ext4_unregister_sysfs(sb);
1206
1207	if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount"))
1208		ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.",
1209			 &sb->s_uuid);
1210
1211	ext4_unregister_li_request(sb);
1212	ext4_quota_off_umount(sb);
1213
1214	flush_work(&sbi->s_error_work);
1215	destroy_workqueue(sbi->rsv_conversion_wq);
1216	ext4_release_orphan_info(sb);
1217
1218	if (sbi->s_journal) {
1219		aborted = is_journal_aborted(sbi->s_journal);
1220		err = jbd2_journal_destroy(sbi->s_journal);
1221		sbi->s_journal = NULL;
1222		if ((err < 0) && !aborted) {
1223			ext4_abort(sb, -err, "Couldn't clean up the journal");
1224		}
1225	}
1226
1227	ext4_es_unregister_shrinker(sbi);
1228	timer_shutdown_sync(&sbi->s_err_report);
1229	ext4_release_system_zone(sb);
1230	ext4_mb_release(sb);
1231	ext4_ext_release(sb);
1232
1233	if (!sb_rdonly(sb) && !aborted) {
1234		ext4_clear_feature_journal_needs_recovery(sb);
1235		ext4_clear_feature_orphan_present(sb);
1236		es->s_state = cpu_to_le16(sbi->s_mount_state);
1237	}
1238	if (!sb_rdonly(sb))
1239		ext4_commit_super(sb);
1240
1241	rcu_read_lock();
1242	group_desc = rcu_dereference(sbi->s_group_desc);
1243	for (i = 0; i < sbi->s_gdb_count; i++)
1244		brelse(group_desc[i]);
1245	kvfree(group_desc);
1246	flex_groups = rcu_dereference(sbi->s_flex_groups);
1247	if (flex_groups) {
1248		for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1249			kvfree(flex_groups[i]);
1250		kvfree(flex_groups);
1251	}
1252	rcu_read_unlock();
1253	percpu_counter_destroy(&sbi->s_freeclusters_counter);
1254	percpu_counter_destroy(&sbi->s_freeinodes_counter);
1255	percpu_counter_destroy(&sbi->s_dirs_counter);
1256	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1257	percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1258	percpu_free_rwsem(&sbi->s_writepages_rwsem);
1259#ifdef CONFIG_QUOTA
1260	for (i = 0; i < EXT4_MAXQUOTAS; i++)
1261		kfree(get_qf_name(sb, sbi, i));
1262#endif
1263
1264	/* Debugging code just in case the in-memory inode orphan list
1265	 * isn't empty.  The on-disk one can be non-empty if we've
1266	 * detected an error and taken the fs readonly, but the
1267	 * in-memory list had better be clean by this point. */
1268	if (!list_empty(&sbi->s_orphan))
1269		dump_orphan_list(sb, sbi);
1270	ASSERT(list_empty(&sbi->s_orphan));
1271
1272	sync_blockdev(sb->s_bdev);
1273	invalidate_bdev(sb->s_bdev);
1274	if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1275		/*
1276		 * Invalidate the journal device's buffers.  We don't want them
1277		 * floating about in memory - the physical journal device may
1278		 * hotswapped, and it breaks the `ro-after' testing code.
1279		 */
1280		sync_blockdev(sbi->s_journal_bdev);
1281		invalidate_bdev(sbi->s_journal_bdev);
1282		ext4_blkdev_remove(sbi);
1283	}
1284
1285	ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1286	sbi->s_ea_inode_cache = NULL;
1287
1288	ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1289	sbi->s_ea_block_cache = NULL;
1290
1291	ext4_stop_mmpd(sbi);
1292
1293	brelse(sbi->s_sbh);
1294	sb->s_fs_info = NULL;
1295	/*
1296	 * Now that we are completely done shutting down the
1297	 * superblock, we need to actually destroy the kobject.
1298	 */
1299	kobject_put(&sbi->s_kobj);
1300	wait_for_completion(&sbi->s_kobj_unregister);
1301	if (sbi->s_chksum_driver)
1302		crypto_free_shash(sbi->s_chksum_driver);
1303	kfree(sbi->s_blockgroup_lock);
1304	fs_put_dax(sbi->s_daxdev, NULL);
1305	fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1306#if IS_ENABLED(CONFIG_UNICODE)
1307	utf8_unload(sb->s_encoding);
1308#endif
1309	kfree(sbi);
1310}
1311
1312static struct kmem_cache *ext4_inode_cachep;
1313
1314/*
1315 * Called inside transaction, so use GFP_NOFS
1316 */
1317static struct inode *ext4_alloc_inode(struct super_block *sb)
1318{
1319	struct ext4_inode_info *ei;
1320
1321	ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
1322	if (!ei)
1323		return NULL;
1324
1325	inode_set_iversion(&ei->vfs_inode, 1);
1326	ei->i_flags = 0;
1327	spin_lock_init(&ei->i_raw_lock);
1328	INIT_LIST_HEAD(&ei->i_prealloc_list);
1329	atomic_set(&ei->i_prealloc_active, 0);
1330	spin_lock_init(&ei->i_prealloc_lock);
1331	ext4_es_init_tree(&ei->i_es_tree);
1332	rwlock_init(&ei->i_es_lock);
1333	INIT_LIST_HEAD(&ei->i_es_list);
1334	ei->i_es_all_nr = 0;
1335	ei->i_es_shk_nr = 0;
1336	ei->i_es_shrink_lblk = 0;
1337	ei->i_reserved_data_blocks = 0;
1338	spin_lock_init(&(ei->i_block_reservation_lock));
1339	ext4_init_pending_tree(&ei->i_pending_tree);
1340#ifdef CONFIG_QUOTA
1341	ei->i_reserved_quota = 0;
1342	memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1343#endif
1344	ei->jinode = NULL;
1345	INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1346	spin_lock_init(&ei->i_completed_io_lock);
1347	ei->i_sync_tid = 0;
1348	ei->i_datasync_tid = 0;
1349	atomic_set(&ei->i_unwritten, 0);
1350	INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1351	ext4_fc_init_inode(&ei->vfs_inode);
1352	mutex_init(&ei->i_fc_lock);
1353	return &ei->vfs_inode;
1354}
1355
1356static int ext4_drop_inode(struct inode *inode)
1357{
1358	int drop = generic_drop_inode(inode);
1359
1360	if (!drop)
1361		drop = fscrypt_drop_inode(inode);
1362
1363	trace_ext4_drop_inode(inode, drop);
1364	return drop;
1365}
1366
1367static void ext4_free_in_core_inode(struct inode *inode)
1368{
1369	fscrypt_free_inode(inode);
1370	if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1371		pr_warn("%s: inode %ld still in fc list",
1372			__func__, inode->i_ino);
1373	}
1374	kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1375}
1376
1377static void ext4_destroy_inode(struct inode *inode)
1378{
1379	if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1380		ext4_msg(inode->i_sb, KERN_ERR,
1381			 "Inode %lu (%p): orphan list check failed!",
1382			 inode->i_ino, EXT4_I(inode));
1383		print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1384				EXT4_I(inode), sizeof(struct ext4_inode_info),
1385				true);
1386		dump_stack();
1387	}
1388
1389	if (EXT4_I(inode)->i_reserved_data_blocks)
1390		ext4_msg(inode->i_sb, KERN_ERR,
1391			 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1392			 inode->i_ino, EXT4_I(inode),
1393			 EXT4_I(inode)->i_reserved_data_blocks);
1394}
1395
1396static void init_once(void *foo)
1397{
1398	struct ext4_inode_info *ei = foo;
1399
1400	INIT_LIST_HEAD(&ei->i_orphan);
1401	init_rwsem(&ei->xattr_sem);
1402	init_rwsem(&ei->i_data_sem);
 
1403	inode_init_once(&ei->vfs_inode);
1404	ext4_fc_init_inode(&ei->vfs_inode);
1405}
1406
1407static int __init init_inodecache(void)
1408{
1409	ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1410				sizeof(struct ext4_inode_info), 0,
1411				(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1412					SLAB_ACCOUNT),
1413				offsetof(struct ext4_inode_info, i_data),
1414				sizeof_field(struct ext4_inode_info, i_data),
1415				init_once);
1416	if (ext4_inode_cachep == NULL)
1417		return -ENOMEM;
1418	return 0;
1419}
1420
1421static void destroy_inodecache(void)
1422{
1423	/*
1424	 * Make sure all delayed rcu free inodes are flushed before we
1425	 * destroy cache.
1426	 */
1427	rcu_barrier();
1428	kmem_cache_destroy(ext4_inode_cachep);
1429}
1430
1431void ext4_clear_inode(struct inode *inode)
1432{
1433	ext4_fc_del(inode);
1434	invalidate_inode_buffers(inode);
1435	clear_inode(inode);
1436	ext4_discard_preallocations(inode, 0);
1437	ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1438	dquot_drop(inode);
1439	if (EXT4_I(inode)->jinode) {
1440		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1441					       EXT4_I(inode)->jinode);
1442		jbd2_free_inode(EXT4_I(inode)->jinode);
1443		EXT4_I(inode)->jinode = NULL;
1444	}
1445	fscrypt_put_encryption_info(inode);
1446	fsverity_cleanup_inode(inode);
1447}
1448
1449static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1450					u64 ino, u32 generation)
1451{
1452	struct inode *inode;
1453
1454	/*
1455	 * Currently we don't know the generation for parent directory, so
1456	 * a generation of 0 means "accept any"
1457	 */
1458	inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1459	if (IS_ERR(inode))
1460		return ERR_CAST(inode);
1461	if (generation && inode->i_generation != generation) {
1462		iput(inode);
1463		return ERR_PTR(-ESTALE);
1464	}
1465
1466	return inode;
1467}
1468
1469static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1470					int fh_len, int fh_type)
1471{
1472	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1473				    ext4_nfs_get_inode);
1474}
1475
1476static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1477					int fh_len, int fh_type)
1478{
1479	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1480				    ext4_nfs_get_inode);
1481}
1482
1483static int ext4_nfs_commit_metadata(struct inode *inode)
1484{
1485	struct writeback_control wbc = {
1486		.sync_mode = WB_SYNC_ALL
1487	};
1488
1489	trace_ext4_nfs_commit_metadata(inode);
1490	return ext4_write_inode(inode, &wbc);
1491}
1492
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1493#ifdef CONFIG_QUOTA
1494static const char * const quotatypes[] = INITQFNAMES;
1495#define QTYPE2NAME(t) (quotatypes[t])
1496
1497static int ext4_write_dquot(struct dquot *dquot);
1498static int ext4_acquire_dquot(struct dquot *dquot);
1499static int ext4_release_dquot(struct dquot *dquot);
1500static int ext4_mark_dquot_dirty(struct dquot *dquot);
1501static int ext4_write_info(struct super_block *sb, int type);
1502static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1503			 const struct path *path);
 
1504static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1505			       size_t len, loff_t off);
1506static ssize_t ext4_quota_write(struct super_block *sb, int type,
1507				const char *data, size_t len, loff_t off);
1508static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1509			     unsigned int flags);
 
1510
1511static struct dquot **ext4_get_dquots(struct inode *inode)
1512{
1513	return EXT4_I(inode)->i_dquot;
1514}
1515
1516static const struct dquot_operations ext4_quota_operations = {
1517	.get_reserved_space	= ext4_get_reserved_space,
1518	.write_dquot		= ext4_write_dquot,
1519	.acquire_dquot		= ext4_acquire_dquot,
1520	.release_dquot		= ext4_release_dquot,
1521	.mark_dirty		= ext4_mark_dquot_dirty,
1522	.write_info		= ext4_write_info,
1523	.alloc_dquot		= dquot_alloc,
1524	.destroy_dquot		= dquot_destroy,
1525	.get_projid		= ext4_get_projid,
1526	.get_inode_usage	= ext4_get_inode_usage,
1527	.get_next_id		= dquot_get_next_id,
1528};
1529
1530static const struct quotactl_ops ext4_qctl_operations = {
1531	.quota_on	= ext4_quota_on,
1532	.quota_off	= ext4_quota_off,
1533	.quota_sync	= dquot_quota_sync,
1534	.get_state	= dquot_get_state,
1535	.set_info	= dquot_set_dqinfo,
1536	.get_dqblk	= dquot_get_dqblk,
1537	.set_dqblk	= dquot_set_dqblk,
1538	.get_nextdqblk	= dquot_get_next_dqblk,
1539};
1540#endif
1541
1542static const struct super_operations ext4_sops = {
1543	.alloc_inode	= ext4_alloc_inode,
1544	.free_inode	= ext4_free_in_core_inode,
1545	.destroy_inode	= ext4_destroy_inode,
1546	.write_inode	= ext4_write_inode,
1547	.dirty_inode	= ext4_dirty_inode,
1548	.drop_inode	= ext4_drop_inode,
1549	.evict_inode	= ext4_evict_inode,
1550	.put_super	= ext4_put_super,
1551	.sync_fs	= ext4_sync_fs,
1552	.freeze_fs	= ext4_freeze,
1553	.unfreeze_fs	= ext4_unfreeze,
1554	.statfs		= ext4_statfs,
 
1555	.show_options	= ext4_show_options,
1556#ifdef CONFIG_QUOTA
1557	.quota_read	= ext4_quota_read,
1558	.quota_write	= ext4_quota_write,
1559	.get_dquots	= ext4_get_dquots,
1560#endif
1561};
1562
1563static const struct export_operations ext4_export_ops = {
1564	.fh_to_dentry = ext4_fh_to_dentry,
1565	.fh_to_parent = ext4_fh_to_parent,
1566	.get_parent = ext4_get_parent,
1567	.commit_metadata = ext4_nfs_commit_metadata,
1568};
1569
1570enum {
1571	Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1572	Opt_resgid, Opt_resuid, Opt_sb,
1573	Opt_nouid32, Opt_debug, Opt_removed,
1574	Opt_user_xattr, Opt_acl,
1575	Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1576	Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1577	Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1578	Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1579	Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1580	Opt_inlinecrypt,
1581	Opt_usrjquota, Opt_grpjquota, Opt_quota,
 
1582	Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1583	Opt_usrquota, Opt_grpquota, Opt_prjquota,
1584	Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1585	Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1586	Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
 
1587	Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1588	Opt_inode_readahead_blks, Opt_journal_ioprio,
1589	Opt_dioread_nolock, Opt_dioread_lock,
1590	Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1591	Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1592	Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1593	Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
1594#ifdef CONFIG_EXT4_DEBUG
1595	Opt_fc_debug_max_replay, Opt_fc_debug_force
1596#endif
1597};
1598
1599static const struct constant_table ext4_param_errors[] = {
1600	{"continue",	EXT4_MOUNT_ERRORS_CONT},
1601	{"panic",	EXT4_MOUNT_ERRORS_PANIC},
1602	{"remount-ro",	EXT4_MOUNT_ERRORS_RO},
1603	{}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1604};
1605
1606static const struct constant_table ext4_param_data[] = {
1607	{"journal",	EXT4_MOUNT_JOURNAL_DATA},
1608	{"ordered",	EXT4_MOUNT_ORDERED_DATA},
1609	{"writeback",	EXT4_MOUNT_WRITEBACK_DATA},
1610	{}
1611};
1612
1613static const struct constant_table ext4_param_data_err[] = {
1614	{"abort",	Opt_data_err_abort},
1615	{"ignore",	Opt_data_err_ignore},
1616	{}
1617};
1618
1619static const struct constant_table ext4_param_jqfmt[] = {
1620	{"vfsold",	QFMT_VFS_OLD},
1621	{"vfsv0",	QFMT_VFS_V0},
1622	{"vfsv1",	QFMT_VFS_V1},
1623	{}
1624};
 
 
 
 
 
1625
1626static const struct constant_table ext4_param_dax[] = {
1627	{"always",	Opt_dax_always},
1628	{"inode",	Opt_dax_inode},
1629	{"never",	Opt_dax_never},
1630	{}
1631};
1632
1633/* String parameter that allows empty argument */
1634#define fsparam_string_empty(NAME, OPT) \
1635	__fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL)
1636
1637/*
1638 * Mount option specification
1639 * We don't use fsparam_flag_no because of the way we set the
1640 * options and the way we show them in _ext4_show_options(). To
1641 * keep the changes to a minimum, let's keep the negative options
1642 * separate for now.
1643 */
1644static const struct fs_parameter_spec ext4_param_specs[] = {
1645	fsparam_flag	("bsddf",		Opt_bsd_df),
1646	fsparam_flag	("minixdf",		Opt_minix_df),
1647	fsparam_flag	("grpid",		Opt_grpid),
1648	fsparam_flag	("bsdgroups",		Opt_grpid),
1649	fsparam_flag	("nogrpid",		Opt_nogrpid),
1650	fsparam_flag	("sysvgroups",		Opt_nogrpid),
1651	fsparam_u32	("resgid",		Opt_resgid),
1652	fsparam_u32	("resuid",		Opt_resuid),
1653	fsparam_u32	("sb",			Opt_sb),
1654	fsparam_enum	("errors",		Opt_errors, ext4_param_errors),
1655	fsparam_flag	("nouid32",		Opt_nouid32),
1656	fsparam_flag	("debug",		Opt_debug),
1657	fsparam_flag	("oldalloc",		Opt_removed),
1658	fsparam_flag	("orlov",		Opt_removed),
1659	fsparam_flag	("user_xattr",		Opt_user_xattr),
1660	fsparam_flag	("acl",			Opt_acl),
1661	fsparam_flag	("norecovery",		Opt_noload),
1662	fsparam_flag	("noload",		Opt_noload),
1663	fsparam_flag	("bh",			Opt_removed),
1664	fsparam_flag	("nobh",		Opt_removed),
1665	fsparam_u32	("commit",		Opt_commit),
1666	fsparam_u32	("min_batch_time",	Opt_min_batch_time),
1667	fsparam_u32	("max_batch_time",	Opt_max_batch_time),
1668	fsparam_u32	("journal_dev",		Opt_journal_dev),
1669	fsparam_bdev	("journal_path",	Opt_journal_path),
1670	fsparam_flag	("journal_checksum",	Opt_journal_checksum),
1671	fsparam_flag	("nojournal_checksum",	Opt_nojournal_checksum),
1672	fsparam_flag	("journal_async_commit",Opt_journal_async_commit),
1673	fsparam_flag	("abort",		Opt_abort),
1674	fsparam_enum	("data",		Opt_data, ext4_param_data),
1675	fsparam_enum	("data_err",		Opt_data_err,
1676						ext4_param_data_err),
1677	fsparam_string_empty
1678			("usrjquota",		Opt_usrjquota),
1679	fsparam_string_empty
1680			("grpjquota",		Opt_grpjquota),
1681	fsparam_enum	("jqfmt",		Opt_jqfmt, ext4_param_jqfmt),
1682	fsparam_flag	("grpquota",		Opt_grpquota),
1683	fsparam_flag	("quota",		Opt_quota),
1684	fsparam_flag	("noquota",		Opt_noquota),
1685	fsparam_flag	("usrquota",		Opt_usrquota),
1686	fsparam_flag	("prjquota",		Opt_prjquota),
1687	fsparam_flag	("barrier",		Opt_barrier),
1688	fsparam_u32	("barrier",		Opt_barrier),
1689	fsparam_flag	("nobarrier",		Opt_nobarrier),
1690	fsparam_flag	("i_version",		Opt_removed),
1691	fsparam_flag	("dax",			Opt_dax),
1692	fsparam_enum	("dax",			Opt_dax_type, ext4_param_dax),
1693	fsparam_u32	("stripe",		Opt_stripe),
1694	fsparam_flag	("delalloc",		Opt_delalloc),
1695	fsparam_flag	("nodelalloc",		Opt_nodelalloc),
1696	fsparam_flag	("warn_on_error",	Opt_warn_on_error),
1697	fsparam_flag	("nowarn_on_error",	Opt_nowarn_on_error),
1698	fsparam_u32	("debug_want_extra_isize",
1699						Opt_debug_want_extra_isize),
1700	fsparam_flag	("mblk_io_submit",	Opt_removed),
1701	fsparam_flag	("nomblk_io_submit",	Opt_removed),
1702	fsparam_flag	("block_validity",	Opt_block_validity),
1703	fsparam_flag	("noblock_validity",	Opt_noblock_validity),
1704	fsparam_u32	("inode_readahead_blks",
1705						Opt_inode_readahead_blks),
1706	fsparam_u32	("journal_ioprio",	Opt_journal_ioprio),
1707	fsparam_u32	("auto_da_alloc",	Opt_auto_da_alloc),
1708	fsparam_flag	("auto_da_alloc",	Opt_auto_da_alloc),
1709	fsparam_flag	("noauto_da_alloc",	Opt_noauto_da_alloc),
1710	fsparam_flag	("dioread_nolock",	Opt_dioread_nolock),
1711	fsparam_flag	("nodioread_nolock",	Opt_dioread_lock),
1712	fsparam_flag	("dioread_lock",	Opt_dioread_lock),
1713	fsparam_flag	("discard",		Opt_discard),
1714	fsparam_flag	("nodiscard",		Opt_nodiscard),
1715	fsparam_u32	("init_itable",		Opt_init_itable),
1716	fsparam_flag	("init_itable",		Opt_init_itable),
1717	fsparam_flag	("noinit_itable",	Opt_noinit_itable),
1718#ifdef CONFIG_EXT4_DEBUG
1719	fsparam_flag	("fc_debug_force",	Opt_fc_debug_force),
1720	fsparam_u32	("fc_debug_max_replay",	Opt_fc_debug_max_replay),
1721#endif
1722	fsparam_u32	("max_dir_size_kb",	Opt_max_dir_size_kb),
1723	fsparam_flag	("test_dummy_encryption",
1724						Opt_test_dummy_encryption),
1725	fsparam_string	("test_dummy_encryption",
1726						Opt_test_dummy_encryption),
1727	fsparam_flag	("inlinecrypt",		Opt_inlinecrypt),
1728	fsparam_flag	("nombcache",		Opt_nombcache),
1729	fsparam_flag	("no_mbcache",		Opt_nombcache),	/* for backward compatibility */
1730	fsparam_flag	("prefetch_block_bitmaps",
1731						Opt_removed),
1732	fsparam_flag	("no_prefetch_block_bitmaps",
1733						Opt_no_prefetch_block_bitmaps),
1734	fsparam_s32	("mb_optimize_scan",	Opt_mb_optimize_scan),
1735	fsparam_string	("check",		Opt_removed),	/* mount option from ext2/3 */
1736	fsparam_flag	("nocheck",		Opt_removed),	/* mount option from ext2/3 */
1737	fsparam_flag	("reservation",		Opt_removed),	/* mount option from ext2/3 */
1738	fsparam_flag	("noreservation",	Opt_removed),	/* mount option from ext2/3 */
1739	fsparam_u32	("journal",		Opt_removed),	/* mount option from ext2/3 */
1740	{}
1741};
1742
1743#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1744
1745#define MOPT_SET	0x0001
1746#define MOPT_CLEAR	0x0002
1747#define MOPT_NOSUPPORT	0x0004
1748#define MOPT_EXPLICIT	0x0008
 
 
1749#ifdef CONFIG_QUOTA
1750#define MOPT_Q		0
1751#define MOPT_QFMT	0x0010
1752#else
1753#define MOPT_Q		MOPT_NOSUPPORT
1754#define MOPT_QFMT	MOPT_NOSUPPORT
1755#endif
1756#define MOPT_NO_EXT2	0x0020
1757#define MOPT_NO_EXT3	0x0040
 
1758#define MOPT_EXT4_ONLY	(MOPT_NO_EXT2 | MOPT_NO_EXT3)
1759#define MOPT_SKIP	0x0080
1760#define	MOPT_2		0x0100
 
1761
1762static const struct mount_opts {
1763	int	token;
1764	int	mount_opt;
1765	int	flags;
1766} ext4_mount_opts[] = {
1767	{Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1768	{Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1769	{Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1770	{Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1771	{Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1772	{Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1773	{Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1774	 MOPT_EXT4_ONLY | MOPT_SET},
1775	{Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1776	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1777	{Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1778	{Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1779	{Opt_delalloc, EXT4_MOUNT_DELALLOC,
1780	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1781	{Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1782	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1783	{Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1784	{Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1785	{Opt_commit, 0, MOPT_NO_EXT2},
1786	{Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1787	 MOPT_EXT4_ONLY | MOPT_CLEAR},
1788	{Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1789	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1790	{Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1791				    EXT4_MOUNT_JOURNAL_CHECKSUM),
1792	 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1793	{Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1794	{Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
 
 
 
 
 
 
1795	{Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1796	{Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1797	{Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1798	{Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1799	{Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1800	{Opt_dax_type, 0, MOPT_EXT4_ONLY},
1801	{Opt_journal_dev, 0, MOPT_NO_EXT2},
1802	{Opt_journal_path, 0, MOPT_NO_EXT2},
1803	{Opt_journal_ioprio, 0, MOPT_NO_EXT2},
1804	{Opt_data, 0, MOPT_NO_EXT2},
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1805	{Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
 
1806#ifdef CONFIG_EXT4_FS_POSIX_ACL
1807	{Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
 
1808#else
1809	{Opt_acl, 0, MOPT_NOSUPPORT},
 
1810#endif
1811	{Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1812	{Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
 
1813	{Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1814	{Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1815							MOPT_SET | MOPT_Q},
1816	{Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1817							MOPT_SET | MOPT_Q},
1818	{Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1819							MOPT_SET | MOPT_Q},
1820	{Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1821		       EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1822							MOPT_CLEAR | MOPT_Q},
1823	{Opt_usrjquota, 0, MOPT_Q},
1824	{Opt_grpjquota, 0, MOPT_Q},
1825	{Opt_jqfmt, 0, MOPT_QFMT},
 
 
 
 
 
 
1826	{Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1827	{Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
1828	 MOPT_SET},
 
1829#ifdef CONFIG_EXT4_DEBUG
1830	{Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1831	 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
 
1832#endif
1833	{Opt_err, 0, 0}
1834};
1835
1836#if IS_ENABLED(CONFIG_UNICODE)
1837static const struct ext4_sb_encodings {
1838	__u16 magic;
1839	char *name;
1840	unsigned int version;
1841} ext4_sb_encoding_map[] = {
1842	{EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
1843};
1844
1845static const struct ext4_sb_encodings *
1846ext4_sb_read_encoding(const struct ext4_super_block *es)
 
1847{
1848	__u16 magic = le16_to_cpu(es->s_encoding);
1849	int i;
1850
1851	for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1852		if (magic == ext4_sb_encoding_map[i].magic)
1853			return &ext4_sb_encoding_map[i];
1854
1855	return NULL;
1856}
1857#endif
1858
1859#define EXT4_SPEC_JQUOTA			(1 <<  0)
1860#define EXT4_SPEC_JQFMT				(1 <<  1)
1861#define EXT4_SPEC_DATAJ				(1 <<  2)
1862#define EXT4_SPEC_SB_BLOCK			(1 <<  3)
1863#define EXT4_SPEC_JOURNAL_DEV			(1 <<  4)
1864#define EXT4_SPEC_JOURNAL_IOPRIO		(1 <<  5)
1865#define EXT4_SPEC_s_want_extra_isize		(1 <<  7)
1866#define EXT4_SPEC_s_max_batch_time		(1 <<  8)
1867#define EXT4_SPEC_s_min_batch_time		(1 <<  9)
1868#define EXT4_SPEC_s_inode_readahead_blks	(1 << 10)
1869#define EXT4_SPEC_s_li_wait_mult		(1 << 11)
1870#define EXT4_SPEC_s_max_dir_size_kb		(1 << 12)
1871#define EXT4_SPEC_s_stripe			(1 << 13)
1872#define EXT4_SPEC_s_resuid			(1 << 14)
1873#define EXT4_SPEC_s_resgid			(1 << 15)
1874#define EXT4_SPEC_s_commit_interval		(1 << 16)
1875#define EXT4_SPEC_s_fc_debug_max_replay		(1 << 17)
1876#define EXT4_SPEC_s_sb_block			(1 << 18)
1877#define EXT4_SPEC_mb_optimize_scan		(1 << 19)
1878
1879struct ext4_fs_context {
1880	char		*s_qf_names[EXT4_MAXQUOTAS];
1881	struct fscrypt_dummy_policy dummy_enc_policy;
1882	int		s_jquota_fmt;	/* Format of quota to use */
1883#ifdef CONFIG_EXT4_DEBUG
1884	int s_fc_debug_max_replay;
1885#endif
1886	unsigned short	qname_spec;
1887	unsigned long	vals_s_flags;	/* Bits to set in s_flags */
1888	unsigned long	mask_s_flags;	/* Bits changed in s_flags */
1889	unsigned long	journal_devnum;
1890	unsigned long	s_commit_interval;
1891	unsigned long	s_stripe;
1892	unsigned int	s_inode_readahead_blks;
1893	unsigned int	s_want_extra_isize;
1894	unsigned int	s_li_wait_mult;
1895	unsigned int	s_max_dir_size_kb;
1896	unsigned int	journal_ioprio;
1897	unsigned int	vals_s_mount_opt;
1898	unsigned int	mask_s_mount_opt;
1899	unsigned int	vals_s_mount_opt2;
1900	unsigned int	mask_s_mount_opt2;
1901	unsigned long	vals_s_mount_flags;
1902	unsigned long	mask_s_mount_flags;
1903	unsigned int	opt_flags;	/* MOPT flags */
1904	unsigned int	spec;
1905	u32		s_max_batch_time;
1906	u32		s_min_batch_time;
1907	kuid_t		s_resuid;
1908	kgid_t		s_resgid;
1909	ext4_fsblk_t	s_sb_block;
1910};
1911
1912static void ext4_fc_free(struct fs_context *fc)
1913{
1914	struct ext4_fs_context *ctx = fc->fs_private;
1915	int i;
1916
1917	if (!ctx)
1918		return;
1919
1920	for (i = 0; i < EXT4_MAXQUOTAS; i++)
1921		kfree(ctx->s_qf_names[i]);
1922
1923	fscrypt_free_dummy_policy(&ctx->dummy_enc_policy);
1924	kfree(ctx);
1925}
1926
1927int ext4_init_fs_context(struct fs_context *fc)
1928{
1929	struct ext4_fs_context *ctx;
1930
1931	ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
1932	if (!ctx)
1933		return -ENOMEM;
1934
1935	fc->fs_private = ctx;
1936	fc->ops = &ext4_context_ops;
1937
1938	return 0;
1939}
1940
1941#ifdef CONFIG_QUOTA
1942/*
1943 * Note the name of the specified quota file.
1944 */
1945static int note_qf_name(struct fs_context *fc, int qtype,
1946		       struct fs_parameter *param)
1947{
1948	struct ext4_fs_context *ctx = fc->fs_private;
1949	char *qname;
1950
1951	if (param->size < 1) {
1952		ext4_msg(NULL, KERN_ERR, "Missing quota name");
1953		return -EINVAL;
1954	}
1955	if (strchr(param->string, '/')) {
1956		ext4_msg(NULL, KERN_ERR,
1957			 "quotafile must be on filesystem root");
1958		return -EINVAL;
1959	}
1960	if (ctx->s_qf_names[qtype]) {
1961		if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
1962			ext4_msg(NULL, KERN_ERR,
1963				 "%s quota file already specified",
1964				 QTYPE2NAME(qtype));
1965			return -EINVAL;
1966		}
1967		return 0;
1968	}
1969
1970	qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
1971	if (!qname) {
1972		ext4_msg(NULL, KERN_ERR,
1973			 "Not enough memory for storing quotafile name");
1974		return -ENOMEM;
1975	}
1976	ctx->s_qf_names[qtype] = qname;
1977	ctx->qname_spec |= 1 << qtype;
1978	ctx->spec |= EXT4_SPEC_JQUOTA;
1979	return 0;
1980}
1981
1982/*
1983 * Clear the name of the specified quota file.
1984 */
1985static int unnote_qf_name(struct fs_context *fc, int qtype)
1986{
1987	struct ext4_fs_context *ctx = fc->fs_private;
1988
1989	if (ctx->s_qf_names[qtype])
1990		kfree(ctx->s_qf_names[qtype]);
1991
1992	ctx->s_qf_names[qtype] = NULL;
1993	ctx->qname_spec |= 1 << qtype;
1994	ctx->spec |= EXT4_SPEC_JQUOTA;
1995	return 0;
1996}
1997#endif
1998
1999static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param,
2000					    struct ext4_fs_context *ctx)
 
 
2001{
 
 
2002	int err;
2003
2004	if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
2005		ext4_msg(NULL, KERN_WARNING,
2006			 "test_dummy_encryption option not supported");
2007		return -EINVAL;
 
 
 
 
 
 
2008	}
2009	err = fscrypt_parse_test_dummy_encryption(param,
2010						  &ctx->dummy_enc_policy);
2011	if (err == -EINVAL) {
2012		ext4_msg(NULL, KERN_WARNING,
2013			 "Value of option \"%s\" is unrecognized", param->key);
2014	} else if (err == -EEXIST) {
2015		ext4_msg(NULL, KERN_WARNING,
2016			 "Conflicting test_dummy_encryption options");
2017		return -EINVAL;
 
 
 
 
 
2018	}
2019	return err;
2020}
2021
2022#define EXT4_SET_CTX(name)						\
2023static inline void ctx_set_##name(struct ext4_fs_context *ctx,		\
2024				  unsigned long flag)			\
2025{									\
2026	ctx->mask_s_##name |= flag;					\
2027	ctx->vals_s_##name |= flag;					\
2028}
2029
2030#define EXT4_CLEAR_CTX(name)						\
2031static inline void ctx_clear_##name(struct ext4_fs_context *ctx,	\
2032				    unsigned long flag)			\
2033{									\
2034	ctx->mask_s_##name |= flag;					\
2035	ctx->vals_s_##name &= ~flag;					\
2036}
2037
2038#define EXT4_TEST_CTX(name)						\
2039static inline unsigned long						\
2040ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag)	\
2041{									\
2042	return (ctx->vals_s_##name & flag);				\
2043}
2044
2045EXT4_SET_CTX(flags); /* set only */
2046EXT4_SET_CTX(mount_opt);
2047EXT4_CLEAR_CTX(mount_opt);
2048EXT4_TEST_CTX(mount_opt);
2049EXT4_SET_CTX(mount_opt2);
2050EXT4_CLEAR_CTX(mount_opt2);
2051EXT4_TEST_CTX(mount_opt2);
2052
2053static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit)
2054{
2055	set_bit(bit, &ctx->mask_s_mount_flags);
2056	set_bit(bit, &ctx->vals_s_mount_flags);
2057}
2058
2059static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
 
 
2060{
2061	struct ext4_fs_context *ctx = fc->fs_private;
2062	struct fs_parse_result result;
2063	const struct mount_opts *m;
2064	int is_remount;
2065	kuid_t uid;
2066	kgid_t gid;
2067	int token;
2068
2069	token = fs_parse(fc, ext4_param_specs, param, &result);
2070	if (token < 0)
2071		return token;
2072	is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2073
2074	for (m = ext4_mount_opts; m->token != Opt_err; m++)
2075		if (token == m->token)
2076			break;
2077
2078	ctx->opt_flags |= m->flags;
 
 
 
 
2079
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2080	if (m->flags & MOPT_EXPLICIT) {
2081		if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2082			ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2083		} else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2084			ctx_set_mount_opt2(ctx,
2085				       EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2086		} else
2087			return -EINVAL;
2088	}
2089
2090	if (m->flags & MOPT_NOSUPPORT) {
2091		ext4_msg(NULL, KERN_ERR, "%s option not supported",
2092			 param->key);
2093		return 0;
 
2094	}
2095
2096	switch (token) {
2097#ifdef CONFIG_QUOTA
2098	case Opt_usrjquota:
2099		if (!*param->string)
2100			return unnote_qf_name(fc, USRQUOTA);
2101		else
2102			return note_qf_name(fc, USRQUOTA, param);
2103	case Opt_grpjquota:
2104		if (!*param->string)
2105			return unnote_qf_name(fc, GRPQUOTA);
2106		else
2107			return note_qf_name(fc, GRPQUOTA, param);
2108#endif
2109	case Opt_sb:
2110		if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2111			ext4_msg(NULL, KERN_WARNING,
2112				 "Ignoring %s option on remount", param->key);
2113		} else {
2114			ctx->s_sb_block = result.uint_32;
2115			ctx->spec |= EXT4_SPEC_s_sb_block;
2116		}
2117		return 0;
2118	case Opt_removed:
2119		ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
2120			 param->key);
2121		return 0;
2122	case Opt_abort:
2123		ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED);
2124		return 0;
2125	case Opt_inlinecrypt:
2126#ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2127		ctx_set_flags(ctx, SB_INLINECRYPT);
2128#else
2129		ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
2130#endif
2131		return 0;
2132	case Opt_errors:
2133		ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
2134		ctx_set_mount_opt(ctx, result.uint_32);
2135		return 0;
2136#ifdef CONFIG_QUOTA
2137	case Opt_jqfmt:
2138		ctx->s_jquota_fmt = result.uint_32;
2139		ctx->spec |= EXT4_SPEC_JQFMT;
2140		return 0;
2141#endif
2142	case Opt_data:
2143		ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2144		ctx_set_mount_opt(ctx, result.uint_32);
2145		ctx->spec |= EXT4_SPEC_DATAJ;
2146		return 0;
2147	case Opt_commit:
2148		if (result.uint_32 == 0)
2149			ctx->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE;
2150		else if (result.uint_32 > INT_MAX / HZ) {
2151			ext4_msg(NULL, KERN_ERR,
2152				 "Invalid commit interval %d, "
2153				 "must be smaller than %d",
2154				 result.uint_32, INT_MAX / HZ);
2155			return -EINVAL;
2156		}
2157		ctx->s_commit_interval = HZ * result.uint_32;
2158		ctx->spec |= EXT4_SPEC_s_commit_interval;
2159		return 0;
2160	case Opt_debug_want_extra_isize:
2161		if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
2162			ext4_msg(NULL, KERN_ERR,
2163				 "Invalid want_extra_isize %d", result.uint_32);
2164			return -EINVAL;
2165		}
2166		ctx->s_want_extra_isize = result.uint_32;
2167		ctx->spec |= EXT4_SPEC_s_want_extra_isize;
2168		return 0;
2169	case Opt_max_batch_time:
2170		ctx->s_max_batch_time = result.uint_32;
2171		ctx->spec |= EXT4_SPEC_s_max_batch_time;
2172		return 0;
2173	case Opt_min_batch_time:
2174		ctx->s_min_batch_time = result.uint_32;
2175		ctx->spec |= EXT4_SPEC_s_min_batch_time;
2176		return 0;
2177	case Opt_inode_readahead_blks:
2178		if (result.uint_32 &&
2179		    (result.uint_32 > (1 << 30) ||
2180		     !is_power_of_2(result.uint_32))) {
2181			ext4_msg(NULL, KERN_ERR,
2182				 "EXT4-fs: inode_readahead_blks must be "
2183				 "0 or a power of 2 smaller than 2^31");
2184			return -EINVAL;
2185		}
2186		ctx->s_inode_readahead_blks = result.uint_32;
2187		ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
2188		return 0;
2189	case Opt_init_itable:
2190		ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2191		ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
2192		if (param->type == fs_value_is_string)
2193			ctx->s_li_wait_mult = result.uint_32;
2194		ctx->spec |= EXT4_SPEC_s_li_wait_mult;
2195		return 0;
2196	case Opt_max_dir_size_kb:
2197		ctx->s_max_dir_size_kb = result.uint_32;
2198		ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
2199		return 0;
2200#ifdef CONFIG_EXT4_DEBUG
2201	case Opt_fc_debug_max_replay:
2202		ctx->s_fc_debug_max_replay = result.uint_32;
2203		ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
2204		return 0;
2205#endif
2206	case Opt_stripe:
2207		ctx->s_stripe = result.uint_32;
2208		ctx->spec |= EXT4_SPEC_s_stripe;
2209		return 0;
2210	case Opt_resuid:
2211		uid = make_kuid(current_user_ns(), result.uint_32);
2212		if (!uid_valid(uid)) {
2213			ext4_msg(NULL, KERN_ERR, "Invalid uid value %d",
2214				 result.uint_32);
2215			return -EINVAL;
2216		}
2217		ctx->s_resuid = uid;
2218		ctx->spec |= EXT4_SPEC_s_resuid;
2219		return 0;
2220	case Opt_resgid:
2221		gid = make_kgid(current_user_ns(), result.uint_32);
2222		if (!gid_valid(gid)) {
2223			ext4_msg(NULL, KERN_ERR, "Invalid gid value %d",
2224				 result.uint_32);
2225			return -EINVAL;
2226		}
2227		ctx->s_resgid = gid;
2228		ctx->spec |= EXT4_SPEC_s_resgid;
2229		return 0;
2230	case Opt_journal_dev:
2231		if (is_remount) {
2232			ext4_msg(NULL, KERN_ERR,
2233				 "Cannot specify journal on remount");
2234			return -EINVAL;
2235		}
2236		ctx->journal_devnum = result.uint_32;
2237		ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2238		return 0;
2239	case Opt_journal_path:
2240	{
2241		struct inode *journal_inode;
2242		struct path path;
2243		int error;
2244
2245		if (is_remount) {
2246			ext4_msg(NULL, KERN_ERR,
2247				 "Cannot specify journal on remount");
2248			return -EINVAL;
 
 
 
 
 
 
2249		}
2250
2251		error = fs_lookup_param(fc, param, 1, LOOKUP_FOLLOW, &path);
2252		if (error) {
2253			ext4_msg(NULL, KERN_ERR, "error: could not find "
2254				 "journal device path");
2255			return -EINVAL;
 
2256		}
2257
2258		journal_inode = d_inode(path.dentry);
2259		ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2260		ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
 
 
 
 
 
 
 
2261		path_put(&path);
2262		return 0;
2263	}
2264	case Opt_journal_ioprio:
2265		if (result.uint_32 > 7) {
2266			ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
2267				 " (must be 0-7)");
2268			return -EINVAL;
2269		}
2270		ctx->journal_ioprio =
2271			IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
2272		ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
2273		return 0;
2274	case Opt_test_dummy_encryption:
2275		return ext4_parse_test_dummy_encryption(param, ctx);
2276	case Opt_dax:
2277	case Opt_dax_type:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2278#ifdef CONFIG_FS_DAX
2279	{
2280		int type = (token == Opt_dax) ?
2281			   Opt_dax : result.uint_32;
2282
2283		switch (type) {
2284		case Opt_dax:
2285		case Opt_dax_always:
2286			ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2287			ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2288			break;
2289		case Opt_dax_never:
2290			ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2291			ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
 
 
 
 
2292			break;
2293		case Opt_dax_inode:
2294			ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2295			ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
 
 
 
 
 
2296			/* Strictly for printing options */
2297			ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
2298			break;
2299		}
2300		return 0;
2301	}
2302#else
2303		ext4_msg(NULL, KERN_INFO, "dax option not supported");
2304		return -EINVAL;
2305#endif
2306	case Opt_data_err:
2307		if (result.uint_32 == Opt_data_err_abort)
2308			ctx_set_mount_opt(ctx, m->mount_opt);
2309		else if (result.uint_32 == Opt_data_err_ignore)
2310			ctx_clear_mount_opt(ctx, m->mount_opt);
2311		return 0;
2312	case Opt_mb_optimize_scan:
2313		if (result.int_32 == 1) {
2314			ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2315			ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2316		} else if (result.int_32 == 0) {
2317			ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2318			ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2319		} else {
2320			ext4_msg(NULL, KERN_WARNING,
2321				 "mb_optimize_scan should be set to 0 or 1.");
2322			return -EINVAL;
2323		}
2324		return 0;
2325	}
2326
2327	/*
2328	 * At this point we should only be getting options requiring MOPT_SET,
2329	 * or MOPT_CLEAR. Anything else is a bug
2330	 */
2331	if (m->token == Opt_err) {
2332		ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2333			 param->key);
2334		WARN_ON(1);
2335		return -EINVAL;
2336	}
2337
2338	else {
2339		unsigned int set = 0;
2340
2341		if ((param->type == fs_value_is_flag) ||
2342		    result.uint_32 > 0)
2343			set = 1;
2344
2345		if (m->flags & MOPT_CLEAR)
2346			set = !set;
2347		else if (unlikely(!(m->flags & MOPT_SET))) {
2348			ext4_msg(NULL, KERN_WARNING,
2349				 "buggy handling of option %s",
2350				 param->key);
2351			WARN_ON(1);
2352			return -EINVAL;
2353		}
2354		if (m->flags & MOPT_2) {
2355			if (set != 0)
2356				ctx_set_mount_opt2(ctx, m->mount_opt);
2357			else
2358				ctx_clear_mount_opt2(ctx, m->mount_opt);
2359		} else {
2360			if (set != 0)
2361				ctx_set_mount_opt(ctx, m->mount_opt);
2362			else
2363				ctx_clear_mount_opt(ctx, m->mount_opt);
2364		}
2365	}
2366
2367	return 0;
2368}
2369
2370static int parse_options(struct fs_context *fc, char *options)
2371{
2372	struct fs_parameter param;
2373	int ret;
2374	char *key;
 
 
 
2375
2376	if (!options)
2377		return 0;
2378
2379	while ((key = strsep(&options, ",")) != NULL) {
2380		if (*key) {
2381			size_t v_len = 0;
2382			char *value = strchr(key, '=');
2383
2384			param.type = fs_value_is_flag;
2385			param.string = NULL;
2386
2387			if (value) {
2388				if (value == key)
2389					continue;
2390
2391				*value++ = 0;
2392				v_len = strlen(value);
2393				param.string = kmemdup_nul(value, v_len,
2394							   GFP_KERNEL);
2395				if (!param.string)
2396					return -ENOMEM;
2397				param.type = fs_value_is_string;
2398			}
2399
2400			param.key = key;
2401			param.size = v_len;
2402
2403			ret = ext4_parse_param(fc, &param);
2404			if (param.string)
2405				kfree(param.string);
2406			if (ret < 0)
2407				return ret;
2408		}
2409	}
2410
2411	ret = ext4_validate_options(fc);
2412	if (ret < 0)
2413		return ret;
2414
2415	return 0;
2416}
2417
2418static int parse_apply_sb_mount_options(struct super_block *sb,
2419					struct ext4_fs_context *m_ctx)
2420{
2421	struct ext4_sb_info *sbi = EXT4_SB(sb);
2422	char *s_mount_opts = NULL;
2423	struct ext4_fs_context *s_ctx = NULL;
2424	struct fs_context *fc = NULL;
2425	int ret = -ENOMEM;
2426
2427	if (!sbi->s_es->s_mount_opts[0])
2428		return 0;
2429
2430	s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2431				sizeof(sbi->s_es->s_mount_opts),
2432				GFP_KERNEL);
2433	if (!s_mount_opts)
2434		return ret;
2435
2436	fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2437	if (!fc)
2438		goto out_free;
2439
2440	s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2441	if (!s_ctx)
2442		goto out_free;
2443
2444	fc->fs_private = s_ctx;
2445	fc->s_fs_info = sbi;
2446
2447	ret = parse_options(fc, s_mount_opts);
2448	if (ret < 0)
2449		goto parse_failed;
2450
2451	ret = ext4_check_opt_consistency(fc, sb);
2452	if (ret < 0) {
2453parse_failed:
2454		ext4_msg(sb, KERN_WARNING,
2455			 "failed to parse options in superblock: %s",
2456			 s_mount_opts);
2457		ret = 0;
2458		goto out_free;
2459	}
2460
2461	if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2462		m_ctx->journal_devnum = s_ctx->journal_devnum;
2463	if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2464		m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2465
2466	ext4_apply_options(fc, sb);
2467	ret = 0;
2468
2469out_free:
2470	if (fc) {
2471		ext4_fc_free(fc);
2472		kfree(fc);
2473	}
2474	kfree(s_mount_opts);
2475	return ret;
2476}
2477
2478static void ext4_apply_quota_options(struct fs_context *fc,
2479				     struct super_block *sb)
2480{
2481#ifdef CONFIG_QUOTA
2482	bool quota_feature = ext4_has_feature_quota(sb);
2483	struct ext4_fs_context *ctx = fc->fs_private;
2484	struct ext4_sb_info *sbi = EXT4_SB(sb);
2485	char *qname;
2486	int i;
2487
2488	if (quota_feature)
2489		return;
2490
2491	if (ctx->spec & EXT4_SPEC_JQUOTA) {
2492		for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2493			if (!(ctx->qname_spec & (1 << i)))
2494				continue;
2495
2496			qname = ctx->s_qf_names[i]; /* May be NULL */
2497			if (qname)
2498				set_opt(sb, QUOTA);
2499			ctx->s_qf_names[i] = NULL;
2500			qname = rcu_replace_pointer(sbi->s_qf_names[i], qname,
2501						lockdep_is_held(&sb->s_umount));
2502			if (qname)
2503				kfree_rcu(qname);
2504		}
 
 
 
2505	}
2506
2507	if (ctx->spec & EXT4_SPEC_JQFMT)
2508		sbi->s_jquota_fmt = ctx->s_jquota_fmt;
2509#endif
2510}
2511
2512/*
2513 * Check quota settings consistency.
2514 */
2515static int ext4_check_quota_consistency(struct fs_context *fc,
2516					struct super_block *sb)
2517{
2518#ifdef CONFIG_QUOTA
2519	struct ext4_fs_context *ctx = fc->fs_private;
2520	struct ext4_sb_info *sbi = EXT4_SB(sb);
2521	bool quota_feature = ext4_has_feature_quota(sb);
2522	bool quota_loaded = sb_any_quota_loaded(sb);
2523	bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2524	int quota_flags, i;
2525
2526	/*
2527	 * We do the test below only for project quotas. 'usrquota' and
2528	 * 'grpquota' mount options are allowed even without quota feature
2529	 * to support legacy quotas in quota files.
2530	 */
2531	if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2532	    !ext4_has_feature_project(sb)) {
2533		ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2534			 "Cannot enable project quota enforcement.");
2535		return -EINVAL;
2536	}
 
 
 
 
 
2537
2538	quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2539		      EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2540	if (quota_loaded &&
2541	    ctx->mask_s_mount_opt & quota_flags &&
2542	    !ctx_test_mount_opt(ctx, quota_flags))
2543		goto err_quota_change;
2544
2545	if (ctx->spec & EXT4_SPEC_JQUOTA) {
2546
2547		for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2548			if (!(ctx->qname_spec & (1 << i)))
2549				continue;
2550
2551			if (quota_loaded &&
2552			    !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
2553				goto err_jquota_change;
2554
2555			if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
2556			    strcmp(get_qf_name(sb, sbi, i),
2557				   ctx->s_qf_names[i]) != 0)
2558				goto err_jquota_specified;
2559		}
2560
2561		if (quota_feature) {
2562			ext4_msg(NULL, KERN_INFO,
2563				 "Journaled quota options ignored when "
2564				 "QUOTA feature is enabled");
2565			return 0;
2566		}
2567	}
2568
2569	if (ctx->spec & EXT4_SPEC_JQFMT) {
2570		if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
2571			goto err_jquota_change;
2572		if (quota_feature) {
2573			ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2574				 "ignored when QUOTA feature is enabled");
2575			return 0;
2576		}
2577	}
2578
2579	/* Make sure we don't mix old and new quota format */
2580	usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2581		       ctx->s_qf_names[USRQUOTA]);
2582	grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2583		       ctx->s_qf_names[GRPQUOTA]);
2584
2585	usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2586		    test_opt(sb, USRQUOTA));
2587
2588	grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2589		    test_opt(sb, GRPQUOTA));
2590
2591	if (usr_qf_name) {
2592		ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2593		usrquota = false;
2594	}
2595	if (grp_qf_name) {
2596		ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2597		grpquota = false;
2598	}
2599
2600	if (usr_qf_name || grp_qf_name) {
2601		if (usrquota || grpquota) {
2602			ext4_msg(NULL, KERN_ERR, "old and new quota "
2603				 "format mixing");
2604			return -EINVAL;
2605		}
2606
2607		if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2608			ext4_msg(NULL, KERN_ERR, "journaled quota format "
2609				 "not specified");
2610			return -EINVAL;
2611		}
2612	}
2613
2614	return 0;
2615
2616err_quota_change:
2617	ext4_msg(NULL, KERN_ERR,
2618		 "Cannot change quota options when quota turned on");
2619	return -EINVAL;
2620err_jquota_change:
2621	ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2622		 "options when quota turned on");
2623	return -EINVAL;
2624err_jquota_specified:
2625	ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2626		 QTYPE2NAME(i));
2627	return -EINVAL;
2628#else
2629	return 0;
2630#endif
2631}
2632
2633static int ext4_check_test_dummy_encryption(const struct fs_context *fc,
2634					    struct super_block *sb)
2635{
2636	const struct ext4_fs_context *ctx = fc->fs_private;
2637	const struct ext4_sb_info *sbi = EXT4_SB(sb);
2638	int err;
2639
2640	if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy))
2641		return 0;
2642
2643	if (!ext4_has_feature_encrypt(sb)) {
2644		ext4_msg(NULL, KERN_WARNING,
2645			 "test_dummy_encryption requires encrypt feature");
2646		return -EINVAL;
2647	}
2648	/*
2649	 * This mount option is just for testing, and it's not worthwhile to
2650	 * implement the extra complexity (e.g. RCU protection) that would be
2651	 * needed to allow it to be set or changed during remount.  We do allow
2652	 * it to be specified during remount, but only if there is no change.
2653	 */
2654	if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2655		if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2656						 &ctx->dummy_enc_policy))
2657			return 0;
2658		ext4_msg(NULL, KERN_WARNING,
2659			 "Can't set or change test_dummy_encryption on remount");
2660		return -EINVAL;
2661	}
2662	/* Also make sure s_mount_opts didn't contain a conflicting value. */
2663	if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) {
2664		if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2665						 &ctx->dummy_enc_policy))
2666			return 0;
2667		ext4_msg(NULL, KERN_WARNING,
2668			 "Conflicting test_dummy_encryption options");
2669		return -EINVAL;
2670	}
2671	/*
2672	 * fscrypt_add_test_dummy_key() technically changes the super_block, so
2673	 * technically it should be delayed until ext4_apply_options() like the
2674	 * other changes.  But since we never get here for remounts (see above),
2675	 * and this is the last chance to report errors, we do it here.
2676	 */
2677	err = fscrypt_add_test_dummy_key(sb, &ctx->dummy_enc_policy);
2678	if (err)
2679		ext4_msg(NULL, KERN_WARNING,
2680			 "Error adding test dummy encryption key [%d]", err);
2681	return err;
2682}
2683
2684static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx,
2685					     struct super_block *sb)
2686{
2687	if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) ||
2688	    /* if already set, it was already verified to be the same */
2689	    fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy))
2690		return;
2691	EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy;
2692	memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy));
2693	ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2694}
2695
2696static int ext4_check_opt_consistency(struct fs_context *fc,
2697				      struct super_block *sb)
2698{
2699	struct ext4_fs_context *ctx = fc->fs_private;
2700	struct ext4_sb_info *sbi = fc->s_fs_info;
2701	int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2702	int err;
2703
2704	if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2705		ext4_msg(NULL, KERN_ERR,
2706			 "Mount option(s) incompatible with ext2");
2707		return -EINVAL;
2708	}
2709	if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2710		ext4_msg(NULL, KERN_ERR,
2711			 "Mount option(s) incompatible with ext3");
2712		return -EINVAL;
2713	}
2714
2715	if (ctx->s_want_extra_isize >
2716	    (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2717		ext4_msg(NULL, KERN_ERR,
2718			 "Invalid want_extra_isize %d",
2719			 ctx->s_want_extra_isize);
2720		return -EINVAL;
2721	}
2722
2723	if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) {
2724		int blocksize =
2725			BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2726		if (blocksize < PAGE_SIZE)
2727			ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an "
2728				 "experimental mount option 'dioread_nolock' "
2729				 "for blocksize < PAGE_SIZE");
2730	}
2731
2732	err = ext4_check_test_dummy_encryption(fc, sb);
2733	if (err)
2734		return err;
2735
2736	if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2737		if (!sbi->s_journal) {
2738			ext4_msg(NULL, KERN_WARNING,
2739				 "Remounting file system with no journal "
2740				 "so ignoring journalled data option");
2741			ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2742		} else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2743			   test_opt(sb, DATA_FLAGS)) {
2744			ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2745				 "on remount");
2746			return -EINVAL;
2747		}
2748	}
2749
2750	if (is_remount) {
2751		if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2752		    (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2753			ext4_msg(NULL, KERN_ERR, "can't mount with "
2754				 "both data=journal and dax");
2755			return -EINVAL;
2756		}
2757
2758		if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2759		    (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2760		     (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2761fail_dax_change_remount:
2762			ext4_msg(NULL, KERN_ERR, "can't change "
2763				 "dax mount option while remounting");
2764			return -EINVAL;
2765		} else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2766			 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2767			  (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2768			goto fail_dax_change_remount;
2769		} else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2770			   ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2771			    (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2772			    !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2773			goto fail_dax_change_remount;
2774		}
2775	}
2776
2777	return ext4_check_quota_consistency(fc, sb);
2778}
2779
2780static void ext4_apply_options(struct fs_context *fc, struct super_block *sb)
2781{
2782	struct ext4_fs_context *ctx = fc->fs_private;
2783	struct ext4_sb_info *sbi = fc->s_fs_info;
2784
2785	sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2786	sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2787	sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2788	sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2789	sbi->s_mount_flags &= ~ctx->mask_s_mount_flags;
2790	sbi->s_mount_flags |= ctx->vals_s_mount_flags;
2791	sb->s_flags &= ~ctx->mask_s_flags;
2792	sb->s_flags |= ctx->vals_s_flags;
2793
2794#define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2795	APPLY(s_commit_interval);
2796	APPLY(s_stripe);
2797	APPLY(s_max_batch_time);
2798	APPLY(s_min_batch_time);
2799	APPLY(s_want_extra_isize);
2800	APPLY(s_inode_readahead_blks);
2801	APPLY(s_max_dir_size_kb);
2802	APPLY(s_li_wait_mult);
2803	APPLY(s_resgid);
2804	APPLY(s_resuid);
2805
2806#ifdef CONFIG_EXT4_DEBUG
2807	APPLY(s_fc_debug_max_replay);
2808#endif
2809
2810	ext4_apply_quota_options(fc, sb);
2811	ext4_apply_test_dummy_encryption(ctx, sb);
2812}
2813
2814
2815static int ext4_validate_options(struct fs_context *fc)
2816{
2817#ifdef CONFIG_QUOTA
2818	struct ext4_fs_context *ctx = fc->fs_private;
2819	char *usr_qf_name, *grp_qf_name;
2820
2821	usr_qf_name = ctx->s_qf_names[USRQUOTA];
2822	grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2823
2824	if (usr_qf_name || grp_qf_name) {
2825		if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2826			ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2827
2828		if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2829			ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2830
2831		if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2832		    ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
2833			ext4_msg(NULL, KERN_ERR, "old and new quota "
2834				 "format mixing");
2835			return -EINVAL;
2836		}
2837	}
2838#endif
2839	return 1;
2840}
2841
2842static inline void ext4_show_quota_options(struct seq_file *seq,
2843					   struct super_block *sb)
2844{
2845#if defined(CONFIG_QUOTA)
2846	struct ext4_sb_info *sbi = EXT4_SB(sb);
2847	char *usr_qf_name, *grp_qf_name;
2848
2849	if (sbi->s_jquota_fmt) {
2850		char *fmtname = "";
2851
2852		switch (sbi->s_jquota_fmt) {
2853		case QFMT_VFS_OLD:
2854			fmtname = "vfsold";
2855			break;
2856		case QFMT_VFS_V0:
2857			fmtname = "vfsv0";
2858			break;
2859		case QFMT_VFS_V1:
2860			fmtname = "vfsv1";
2861			break;
2862		}
2863		seq_printf(seq, ",jqfmt=%s", fmtname);
2864	}
2865
2866	rcu_read_lock();
2867	usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2868	grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2869	if (usr_qf_name)
2870		seq_show_option(seq, "usrjquota", usr_qf_name);
2871	if (grp_qf_name)
2872		seq_show_option(seq, "grpjquota", grp_qf_name);
2873	rcu_read_unlock();
2874#endif
2875}
2876
2877static const char *token2str(int token)
2878{
2879	const struct fs_parameter_spec *spec;
2880
2881	for (spec = ext4_param_specs; spec->name != NULL; spec++)
2882		if (spec->opt == token && !spec->type)
2883			break;
2884	return spec->name;
2885}
2886
2887/*
2888 * Show an option if
2889 *  - it's set to a non-default value OR
2890 *  - if the per-sb default is different from the global default
2891 */
2892static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2893			      int nodefs)
2894{
2895	struct ext4_sb_info *sbi = EXT4_SB(sb);
2896	struct ext4_super_block *es = sbi->s_es;
2897	int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2898	const struct mount_opts *m;
2899	char sep = nodefs ? '\n' : ',';
2900
2901#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2902#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2903
2904	if (sbi->s_sb_block != 1)
2905		SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2906
2907	for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2908		int want_set = m->flags & MOPT_SET;
2909		if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2910		    m->flags & MOPT_SKIP)
2911			continue;
2912		if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2913			continue; /* skip if same as the default */
2914		if ((want_set &&
2915		     (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2916		    (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2917			continue; /* select Opt_noFoo vs Opt_Foo */
2918		SEQ_OPTS_PRINT("%s", token2str(m->token));
2919	}
2920
2921	if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2922	    le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2923		SEQ_OPTS_PRINT("resuid=%u",
2924				from_kuid_munged(&init_user_ns, sbi->s_resuid));
2925	if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2926	    le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2927		SEQ_OPTS_PRINT("resgid=%u",
2928				from_kgid_munged(&init_user_ns, sbi->s_resgid));
2929	def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2930	if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2931		SEQ_OPTS_PUTS("errors=remount-ro");
2932	if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2933		SEQ_OPTS_PUTS("errors=continue");
2934	if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2935		SEQ_OPTS_PUTS("errors=panic");
2936	if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2937		SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2938	if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2939		SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2940	if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2941		SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
 
 
2942	if (nodefs || sbi->s_stripe)
2943		SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2944	if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2945			(sbi->s_mount_opt ^ def_mount_opt)) {
2946		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2947			SEQ_OPTS_PUTS("data=journal");
2948		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2949			SEQ_OPTS_PUTS("data=ordered");
2950		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2951			SEQ_OPTS_PUTS("data=writeback");
2952	}
2953	if (nodefs ||
2954	    sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2955		SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2956			       sbi->s_inode_readahead_blks);
2957
2958	if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2959		       (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2960		SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2961	if (nodefs || sbi->s_max_dir_size_kb)
2962		SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2963	if (test_opt(sb, DATA_ERR_ABORT))
2964		SEQ_OPTS_PUTS("data_err=abort");
2965
2966	fscrypt_show_test_dummy_encryption(seq, sep, sb);
2967
2968	if (sb->s_flags & SB_INLINECRYPT)
2969		SEQ_OPTS_PUTS("inlinecrypt");
2970
2971	if (test_opt(sb, DAX_ALWAYS)) {
2972		if (IS_EXT2_SB(sb))
2973			SEQ_OPTS_PUTS("dax");
2974		else
2975			SEQ_OPTS_PUTS("dax=always");
2976	} else if (test_opt2(sb, DAX_NEVER)) {
2977		SEQ_OPTS_PUTS("dax=never");
2978	} else if (test_opt2(sb, DAX_INODE)) {
2979		SEQ_OPTS_PUTS("dax=inode");
2980	}
2981
2982	if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
2983			!test_opt2(sb, MB_OPTIMIZE_SCAN)) {
2984		SEQ_OPTS_PUTS("mb_optimize_scan=0");
2985	} else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
2986			test_opt2(sb, MB_OPTIMIZE_SCAN)) {
2987		SEQ_OPTS_PUTS("mb_optimize_scan=1");
2988	}
2989
2990	ext4_show_quota_options(seq, sb);
2991	return 0;
2992}
2993
2994static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2995{
2996	return _ext4_show_options(seq, root->d_sb, 0);
2997}
2998
2999int ext4_seq_options_show(struct seq_file *seq, void *offset)
3000{
3001	struct super_block *sb = seq->private;
3002	int rc;
3003
3004	seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
3005	rc = _ext4_show_options(seq, sb, 1);
3006	seq_puts(seq, "\n");
3007	return rc;
3008}
3009
3010static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
3011			    int read_only)
3012{
3013	struct ext4_sb_info *sbi = EXT4_SB(sb);
3014	int err = 0;
3015
3016	if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
3017		ext4_msg(sb, KERN_ERR, "revision level too high, "
3018			 "forcing read-only mode");
3019		err = -EROFS;
3020		goto done;
3021	}
3022	if (read_only)
3023		goto done;
3024	if (!(sbi->s_mount_state & EXT4_VALID_FS))
3025		ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3026			 "running e2fsck is recommended");
3027	else if (sbi->s_mount_state & EXT4_ERROR_FS)
3028		ext4_msg(sb, KERN_WARNING,
3029			 "warning: mounting fs with errors, "
3030			 "running e2fsck is recommended");
3031	else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
3032		 le16_to_cpu(es->s_mnt_count) >=
3033		 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
3034		ext4_msg(sb, KERN_WARNING,
3035			 "warning: maximal mount count reached, "
3036			 "running e2fsck is recommended");
3037	else if (le32_to_cpu(es->s_checkinterval) &&
3038		 (ext4_get_tstamp(es, s_lastcheck) +
3039		  le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
3040		ext4_msg(sb, KERN_WARNING,
3041			 "warning: checktime reached, "
3042			 "running e2fsck is recommended");
3043	if (!sbi->s_journal)
3044		es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
3045	if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
3046		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
3047	le16_add_cpu(&es->s_mnt_count, 1);
3048	ext4_update_tstamp(es, s_mtime);
3049	if (sbi->s_journal) {
3050		ext4_set_feature_journal_needs_recovery(sb);
3051		if (ext4_has_feature_orphan_file(sb))
3052			ext4_set_feature_orphan_present(sb);
3053	}
3054
3055	err = ext4_commit_super(sb);
3056done:
3057	if (test_opt(sb, DEBUG))
3058		printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
3059				"bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
3060			sb->s_blocksize,
3061			sbi->s_groups_count,
3062			EXT4_BLOCKS_PER_GROUP(sb),
3063			EXT4_INODES_PER_GROUP(sb),
3064			sbi->s_mount_opt, sbi->s_mount_opt2);
 
 
3065	return err;
3066}
3067
3068int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3069{
3070	struct ext4_sb_info *sbi = EXT4_SB(sb);
3071	struct flex_groups **old_groups, **new_groups;
3072	int size, i, j;
3073
3074	if (!sbi->s_log_groups_per_flex)
3075		return 0;
3076
3077	size = ext4_flex_group(sbi, ngroup - 1) + 1;
3078	if (size <= sbi->s_flex_groups_allocated)
3079		return 0;
3080
3081	new_groups = kvzalloc(roundup_pow_of_two(size *
3082			      sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
3083	if (!new_groups) {
3084		ext4_msg(sb, KERN_ERR,
3085			 "not enough memory for %d flex group pointers", size);
3086		return -ENOMEM;
3087	}
3088	for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3089		new_groups[i] = kvzalloc(roundup_pow_of_two(
3090					 sizeof(struct flex_groups)),
3091					 GFP_KERNEL);
3092		if (!new_groups[i]) {
3093			for (j = sbi->s_flex_groups_allocated; j < i; j++)
3094				kvfree(new_groups[j]);
3095			kvfree(new_groups);
3096			ext4_msg(sb, KERN_ERR,
3097				 "not enough memory for %d flex groups", size);
3098			return -ENOMEM;
3099		}
3100	}
3101	rcu_read_lock();
3102	old_groups = rcu_dereference(sbi->s_flex_groups);
3103	if (old_groups)
3104		memcpy(new_groups, old_groups,
3105		       (sbi->s_flex_groups_allocated *
3106			sizeof(struct flex_groups *)));
3107	rcu_read_unlock();
3108	rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3109	sbi->s_flex_groups_allocated = size;
3110	if (old_groups)
3111		ext4_kvfree_array_rcu(old_groups);
3112	return 0;
3113}
3114
3115static int ext4_fill_flex_info(struct super_block *sb)
3116{
3117	struct ext4_sb_info *sbi = EXT4_SB(sb);
3118	struct ext4_group_desc *gdp = NULL;
3119	struct flex_groups *fg;
3120	ext4_group_t flex_group;
3121	int i, err;
3122
3123	sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
3124	if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
3125		sbi->s_log_groups_per_flex = 0;
3126		return 1;
3127	}
3128
3129	err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3130	if (err)
3131		goto failed;
3132
3133	for (i = 0; i < sbi->s_groups_count; i++) {
3134		gdp = ext4_get_group_desc(sb, i, NULL);
3135
3136		flex_group = ext4_flex_group(sbi, i);
3137		fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3138		atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
3139		atomic64_add(ext4_free_group_clusters(sb, gdp),
3140			     &fg->free_clusters);
3141		atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
3142	}
3143
3144	return 1;
3145failed:
3146	return 0;
3147}
3148
3149static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
3150				   struct ext4_group_desc *gdp)
3151{
3152	int offset = offsetof(struct ext4_group_desc, bg_checksum);
3153	__u16 crc = 0;
3154	__le32 le_group = cpu_to_le32(block_group);
3155	struct ext4_sb_info *sbi = EXT4_SB(sb);
3156
3157	if (ext4_has_metadata_csum(sbi->s_sb)) {
3158		/* Use new metadata_csum algorithm */
3159		__u32 csum32;
3160		__u16 dummy_csum = 0;
3161
3162		csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3163				     sizeof(le_group));
3164		csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3165		csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3166				     sizeof(dummy_csum));
3167		offset += sizeof(dummy_csum);
3168		if (offset < sbi->s_desc_size)
3169			csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3170					     sbi->s_desc_size - offset);
3171
3172		crc = csum32 & 0xFFFF;
3173		goto out;
3174	}
3175
3176	/* old crc16 code */
3177	if (!ext4_has_feature_gdt_csum(sb))
3178		return 0;
3179
3180	crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3181	crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3182	crc = crc16(crc, (__u8 *)gdp, offset);
3183	offset += sizeof(gdp->bg_checksum); /* skip checksum */
3184	/* for checksum of struct ext4_group_desc do the rest...*/
3185	if (ext4_has_feature_64bit(sb) &&
3186	    offset < le16_to_cpu(sbi->s_es->s_desc_size))
3187		crc = crc16(crc, (__u8 *)gdp + offset,
3188			    le16_to_cpu(sbi->s_es->s_desc_size) -
3189				offset);
3190
3191out:
3192	return cpu_to_le16(crc);
3193}
3194
3195int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
3196				struct ext4_group_desc *gdp)
3197{
3198	if (ext4_has_group_desc_csum(sb) &&
3199	    (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
3200		return 0;
3201
3202	return 1;
3203}
3204
3205void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3206			      struct ext4_group_desc *gdp)
3207{
3208	if (!ext4_has_group_desc_csum(sb))
3209		return;
3210	gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
3211}
3212
3213/* Called at mount-time, super-block is locked */
3214static int ext4_check_descriptors(struct super_block *sb,
3215				  ext4_fsblk_t sb_block,
3216				  ext4_group_t *first_not_zeroed)
3217{
3218	struct ext4_sb_info *sbi = EXT4_SB(sb);
3219	ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3220	ext4_fsblk_t last_block;
3221	ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
3222	ext4_fsblk_t block_bitmap;
3223	ext4_fsblk_t inode_bitmap;
3224	ext4_fsblk_t inode_table;
3225	int flexbg_flag = 0;
3226	ext4_group_t i, grp = sbi->s_groups_count;
3227
3228	if (ext4_has_feature_flex_bg(sb))
3229		flexbg_flag = 1;
3230
3231	ext4_debug("Checking group descriptors");
3232
3233	for (i = 0; i < sbi->s_groups_count; i++) {
3234		struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3235
3236		if (i == sbi->s_groups_count - 1 || flexbg_flag)
3237			last_block = ext4_blocks_count(sbi->s_es) - 1;
3238		else
3239			last_block = first_block +
3240				(EXT4_BLOCKS_PER_GROUP(sb) - 1);
3241
3242		if ((grp == sbi->s_groups_count) &&
3243		   !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3244			grp = i;
3245
3246		block_bitmap = ext4_block_bitmap(sb, gdp);
3247		if (block_bitmap == sb_block) {
3248			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3249				 "Block bitmap for group %u overlaps "
3250				 "superblock", i);
3251			if (!sb_rdonly(sb))
3252				return 0;
3253		}
3254		if (block_bitmap >= sb_block + 1 &&
3255		    block_bitmap <= last_bg_block) {
3256			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3257				 "Block bitmap for group %u overlaps "
3258				 "block group descriptors", i);
3259			if (!sb_rdonly(sb))
3260				return 0;
3261		}
3262		if (block_bitmap < first_block || block_bitmap > last_block) {
3263			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3264			       "Block bitmap for group %u not in group "
3265			       "(block %llu)!", i, block_bitmap);
3266			return 0;
3267		}
3268		inode_bitmap = ext4_inode_bitmap(sb, gdp);
3269		if (inode_bitmap == sb_block) {
3270			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3271				 "Inode bitmap for group %u overlaps "
3272				 "superblock", i);
3273			if (!sb_rdonly(sb))
3274				return 0;
3275		}
3276		if (inode_bitmap >= sb_block + 1 &&
3277		    inode_bitmap <= last_bg_block) {
3278			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3279				 "Inode bitmap for group %u overlaps "
3280				 "block group descriptors", i);
3281			if (!sb_rdonly(sb))
3282				return 0;
3283		}
3284		if (inode_bitmap < first_block || inode_bitmap > last_block) {
3285			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3286			       "Inode bitmap for group %u not in group "
3287			       "(block %llu)!", i, inode_bitmap);
3288			return 0;
3289		}
3290		inode_table = ext4_inode_table(sb, gdp);
3291		if (inode_table == sb_block) {
3292			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3293				 "Inode table for group %u overlaps "
3294				 "superblock", i);
3295			if (!sb_rdonly(sb))
3296				return 0;
3297		}
3298		if (inode_table >= sb_block + 1 &&
3299		    inode_table <= last_bg_block) {
3300			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3301				 "Inode table for group %u overlaps "
3302				 "block group descriptors", i);
3303			if (!sb_rdonly(sb))
3304				return 0;
3305		}
3306		if (inode_table < first_block ||
3307		    inode_table + sbi->s_itb_per_group - 1 > last_block) {
3308			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3309			       "Inode table for group %u not in group "
3310			       "(block %llu)!", i, inode_table);
3311			return 0;
3312		}
3313		ext4_lock_group(sb, i);
3314		if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
3315			ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3316				 "Checksum for group %u failed (%u!=%u)",
3317				 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
3318				     gdp)), le16_to_cpu(gdp->bg_checksum));
3319			if (!sb_rdonly(sb)) {
3320				ext4_unlock_group(sb, i);
3321				return 0;
3322			}
3323		}
3324		ext4_unlock_group(sb, i);
3325		if (!flexbg_flag)
3326			first_block += EXT4_BLOCKS_PER_GROUP(sb);
3327	}
3328	if (NULL != first_not_zeroed)
3329		*first_not_zeroed = grp;
3330	return 1;
3331}
3332
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3333/*
3334 * Maximal extent format file size.
3335 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3336 * extent format containers, within a sector_t, and within i_blocks
3337 * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
3338 * so that won't be a limiting factor.
3339 *
3340 * However there is other limiting factor. We do store extents in the form
3341 * of starting block and length, hence the resulting length of the extent
3342 * covering maximum file size must fit into on-disk format containers as
3343 * well. Given that length is always by 1 unit bigger than max unit (because
3344 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3345 *
3346 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3347 */
3348static loff_t ext4_max_size(int blkbits, int has_huge_files)
3349{
3350	loff_t res;
3351	loff_t upper_limit = MAX_LFS_FILESIZE;
3352
3353	BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3354
3355	if (!has_huge_files) {
3356		upper_limit = (1LL << 32) - 1;
3357
3358		/* total blocks in file system block size */
3359		upper_limit >>= (blkbits - 9);
3360		upper_limit <<= blkbits;
3361	}
3362
3363	/*
3364	 * 32-bit extent-start container, ee_block. We lower the maxbytes
3365	 * by one fs block, so ee_len can cover the extent of maximum file
3366	 * size
3367	 */
3368	res = (1LL << 32) - 1;
3369	res <<= blkbits;
3370
3371	/* Sanity check against vm- & vfs- imposed limits */
3372	if (res > upper_limit)
3373		res = upper_limit;
3374
3375	return res;
3376}
3377
3378/*
3379 * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
3380 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3381 * We need to be 1 filesystem block less than the 2^48 sector limit.
3382 */
3383static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3384{
3385	loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
3386	int meta_blocks;
3387	unsigned int ppb = 1 << (bits - 2);
3388
3389	/*
3390	 * This is calculated to be the largest file size for a dense, block
3391	 * mapped file such that the file's total number of 512-byte sectors,
3392	 * including data and all indirect blocks, does not exceed (2^48 - 1).
3393	 *
3394	 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3395	 * number of 512-byte sectors of the file.
3396	 */
3397	if (!has_huge_files) {
3398		/*
3399		 * !has_huge_files or implies that the inode i_block field
3400		 * represents total file blocks in 2^32 512-byte sectors ==
3401		 * size of vfs inode i_blocks * 8
3402		 */
3403		upper_limit = (1LL << 32) - 1;
3404
3405		/* total blocks in file system block size */
3406		upper_limit >>= (bits - 9);
3407
3408	} else {
3409		/*
3410		 * We use 48 bit ext4_inode i_blocks
3411		 * With EXT4_HUGE_FILE_FL set the i_blocks
3412		 * represent total number of blocks in
3413		 * file system block size
3414		 */
3415		upper_limit = (1LL << 48) - 1;
3416
3417	}
3418
3419	/* Compute how many blocks we can address by block tree */
3420	res += ppb;
3421	res += ppb * ppb;
3422	res += ((loff_t)ppb) * ppb * ppb;
3423	/* Compute how many metadata blocks are needed */
3424	meta_blocks = 1;
3425	meta_blocks += 1 + ppb;
3426	meta_blocks += 1 + ppb + ppb * ppb;
3427	/* Does block tree limit file size? */
3428	if (res + meta_blocks <= upper_limit)
3429		goto check_lfs;
3430
3431	res = upper_limit;
3432	/* How many metadata blocks are needed for addressing upper_limit? */
3433	upper_limit -= EXT4_NDIR_BLOCKS;
3434	/* indirect blocks */
3435	meta_blocks = 1;
3436	upper_limit -= ppb;
3437	/* double indirect blocks */
3438	if (upper_limit < ppb * ppb) {
3439		meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3440		res -= meta_blocks;
3441		goto check_lfs;
3442	}
3443	meta_blocks += 1 + ppb;
3444	upper_limit -= ppb * ppb;
3445	/* tripple indirect blocks for the rest */
3446	meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3447		DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3448	res -= meta_blocks;
3449check_lfs:
3450	res <<= bits;
 
 
 
3451	if (res > MAX_LFS_FILESIZE)
3452		res = MAX_LFS_FILESIZE;
3453
3454	return res;
3455}
3456
3457static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3458				   ext4_fsblk_t logical_sb_block, int nr)
3459{
3460	struct ext4_sb_info *sbi = EXT4_SB(sb);
3461	ext4_group_t bg, first_meta_bg;
3462	int has_super = 0;
3463
3464	first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3465
3466	if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3467		return logical_sb_block + nr + 1;
3468	bg = sbi->s_desc_per_block * nr;
3469	if (ext4_bg_has_super(sb, bg))
3470		has_super = 1;
3471
3472	/*
3473	 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3474	 * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
3475	 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3476	 * compensate.
3477	 */
3478	if (sb->s_blocksize == 1024 && nr == 0 &&
3479	    le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3480		has_super++;
3481
3482	return (has_super + ext4_group_first_block_no(sb, bg));
3483}
3484
3485/**
3486 * ext4_get_stripe_size: Get the stripe size.
3487 * @sbi: In memory super block info
3488 *
3489 * If we have specified it via mount option, then
3490 * use the mount option value. If the value specified at mount time is
3491 * greater than the blocks per group use the super block value.
3492 * If the super block value is greater than blocks per group return 0.
3493 * Allocator needs it be less than blocks per group.
3494 *
3495 */
3496static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3497{
3498	unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3499	unsigned long stripe_width =
3500			le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3501	int ret;
3502
3503	if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3504		ret = sbi->s_stripe;
3505	else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3506		ret = stripe_width;
3507	else if (stride && stride <= sbi->s_blocks_per_group)
3508		ret = stride;
3509	else
3510		ret = 0;
3511
3512	/*
3513	 * If the stripe width is 1, this makes no sense and
3514	 * we set it to 0 to turn off stripe handling code.
3515	 */
3516	if (ret <= 1)
3517		ret = 0;
3518
3519	return ret;
3520}
3521
3522/*
3523 * Check whether this filesystem can be mounted based on
3524 * the features present and the RDONLY/RDWR mount requested.
3525 * Returns 1 if this filesystem can be mounted as requested,
3526 * 0 if it cannot be.
3527 */
3528int ext4_feature_set_ok(struct super_block *sb, int readonly)
3529{
3530	if (ext4_has_unknown_ext4_incompat_features(sb)) {
3531		ext4_msg(sb, KERN_ERR,
3532			"Couldn't mount because of "
3533			"unsupported optional features (%x)",
3534			(le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3535			~EXT4_FEATURE_INCOMPAT_SUPP));
3536		return 0;
3537	}
3538
3539#if !IS_ENABLED(CONFIG_UNICODE)
3540	if (ext4_has_feature_casefold(sb)) {
3541		ext4_msg(sb, KERN_ERR,
3542			 "Filesystem with casefold feature cannot be "
3543			 "mounted without CONFIG_UNICODE");
3544		return 0;
3545	}
3546#endif
3547
3548	if (readonly)
3549		return 1;
3550
3551	if (ext4_has_feature_readonly(sb)) {
3552		ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3553		sb->s_flags |= SB_RDONLY;
3554		return 1;
3555	}
3556
3557	/* Check that feature set is OK for a read-write mount */
3558	if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3559		ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3560			 "unsupported optional features (%x)",
3561			 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3562				~EXT4_FEATURE_RO_COMPAT_SUPP));
3563		return 0;
3564	}
3565	if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3566		ext4_msg(sb, KERN_ERR,
3567			 "Can't support bigalloc feature without "
3568			 "extents feature\n");
3569		return 0;
3570	}
3571
3572#if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3573	if (!readonly && (ext4_has_feature_quota(sb) ||
3574			  ext4_has_feature_project(sb))) {
3575		ext4_msg(sb, KERN_ERR,
3576			 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3577		return 0;
3578	}
3579#endif  /* CONFIG_QUOTA */
3580	return 1;
3581}
3582
3583/*
3584 * This function is called once a day if we have errors logged
3585 * on the file system
3586 */
3587static void print_daily_error_info(struct timer_list *t)
3588{
3589	struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3590	struct super_block *sb = sbi->s_sb;
3591	struct ext4_super_block *es = sbi->s_es;
3592
3593	if (es->s_error_count)
3594		/* fsck newer than v1.41.13 is needed to clean this condition. */
3595		ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3596			 le32_to_cpu(es->s_error_count));
3597	if (es->s_first_error_time) {
3598		printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3599		       sb->s_id,
3600		       ext4_get_tstamp(es, s_first_error_time),
3601		       (int) sizeof(es->s_first_error_func),
3602		       es->s_first_error_func,
3603		       le32_to_cpu(es->s_first_error_line));
3604		if (es->s_first_error_ino)
3605			printk(KERN_CONT ": inode %u",
3606			       le32_to_cpu(es->s_first_error_ino));
3607		if (es->s_first_error_block)
3608			printk(KERN_CONT ": block %llu", (unsigned long long)
3609			       le64_to_cpu(es->s_first_error_block));
3610		printk(KERN_CONT "\n");
3611	}
3612	if (es->s_last_error_time) {
3613		printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3614		       sb->s_id,
3615		       ext4_get_tstamp(es, s_last_error_time),
3616		       (int) sizeof(es->s_last_error_func),
3617		       es->s_last_error_func,
3618		       le32_to_cpu(es->s_last_error_line));
3619		if (es->s_last_error_ino)
3620			printk(KERN_CONT ": inode %u",
3621			       le32_to_cpu(es->s_last_error_ino));
3622		if (es->s_last_error_block)
3623			printk(KERN_CONT ": block %llu", (unsigned long long)
3624			       le64_to_cpu(es->s_last_error_block));
3625		printk(KERN_CONT "\n");
3626	}
3627	mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
3628}
3629
3630/* Find next suitable group and run ext4_init_inode_table */
3631static int ext4_run_li_request(struct ext4_li_request *elr)
3632{
3633	struct ext4_group_desc *gdp = NULL;
3634	struct super_block *sb = elr->lr_super;
3635	ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3636	ext4_group_t group = elr->lr_next_group;
 
3637	unsigned int prefetch_ios = 0;
3638	int ret = 0;
3639	u64 start_time;
3640
3641	if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3642		elr->lr_next_group = ext4_mb_prefetch(sb, group,
3643				EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3644		if (prefetch_ios)
3645			ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3646					      prefetch_ios);
3647		trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3648					    prefetch_ios);
3649		if (group >= elr->lr_next_group) {
3650			ret = 1;
3651			if (elr->lr_first_not_zeroed != ngroups &&
3652			    !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3653				elr->lr_next_group = elr->lr_first_not_zeroed;
3654				elr->lr_mode = EXT4_LI_MODE_ITABLE;
3655				ret = 0;
3656			}
3657		}
3658		return ret;
3659	}
3660
3661	for (; group < ngroups; group++) {
3662		gdp = ext4_get_group_desc(sb, group, NULL);
3663		if (!gdp) {
3664			ret = 1;
3665			break;
3666		}
3667
3668		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3669			break;
3670	}
3671
3672	if (group >= ngroups)
3673		ret = 1;
3674
3675	if (!ret) {
3676		start_time = ktime_get_real_ns();
3677		ret = ext4_init_inode_table(sb, group,
3678					    elr->lr_timeout ? 0 : 1);
3679		trace_ext4_lazy_itable_init(sb, group);
3680		if (elr->lr_timeout == 0) {
3681			elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3682				EXT4_SB(elr->lr_super)->s_li_wait_mult);
 
3683		}
3684		elr->lr_next_sched = jiffies + elr->lr_timeout;
3685		elr->lr_next_group = group + 1;
3686	}
3687	return ret;
3688}
3689
3690/*
3691 * Remove lr_request from the list_request and free the
3692 * request structure. Should be called with li_list_mtx held
3693 */
3694static void ext4_remove_li_request(struct ext4_li_request *elr)
3695{
3696	if (!elr)
3697		return;
3698
3699	list_del(&elr->lr_request);
3700	EXT4_SB(elr->lr_super)->s_li_request = NULL;
3701	kfree(elr);
3702}
3703
3704static void ext4_unregister_li_request(struct super_block *sb)
3705{
3706	mutex_lock(&ext4_li_mtx);
3707	if (!ext4_li_info) {
3708		mutex_unlock(&ext4_li_mtx);
3709		return;
3710	}
3711
3712	mutex_lock(&ext4_li_info->li_list_mtx);
3713	ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3714	mutex_unlock(&ext4_li_info->li_list_mtx);
3715	mutex_unlock(&ext4_li_mtx);
3716}
3717
3718static struct task_struct *ext4_lazyinit_task;
3719
3720/*
3721 * This is the function where ext4lazyinit thread lives. It walks
3722 * through the request list searching for next scheduled filesystem.
3723 * When such a fs is found, run the lazy initialization request
3724 * (ext4_rn_li_request) and keep track of the time spend in this
3725 * function. Based on that time we compute next schedule time of
3726 * the request. When walking through the list is complete, compute
3727 * next waking time and put itself into sleep.
3728 */
3729static int ext4_lazyinit_thread(void *arg)
3730{
3731	struct ext4_lazy_init *eli = arg;
3732	struct list_head *pos, *n;
3733	struct ext4_li_request *elr;
3734	unsigned long next_wakeup, cur;
3735
3736	BUG_ON(NULL == eli);
3737	set_freezable();
3738
3739cont_thread:
3740	while (true) {
3741		next_wakeup = MAX_JIFFY_OFFSET;
3742
3743		mutex_lock(&eli->li_list_mtx);
3744		if (list_empty(&eli->li_request_list)) {
3745			mutex_unlock(&eli->li_list_mtx);
3746			goto exit_thread;
3747		}
3748		list_for_each_safe(pos, n, &eli->li_request_list) {
3749			int err = 0;
3750			int progress = 0;
3751			elr = list_entry(pos, struct ext4_li_request,
3752					 lr_request);
3753
3754			if (time_before(jiffies, elr->lr_next_sched)) {
3755				if (time_before(elr->lr_next_sched, next_wakeup))
3756					next_wakeup = elr->lr_next_sched;
3757				continue;
3758			}
3759			if (down_read_trylock(&elr->lr_super->s_umount)) {
3760				if (sb_start_write_trylock(elr->lr_super)) {
3761					progress = 1;
3762					/*
3763					 * We hold sb->s_umount, sb can not
3764					 * be removed from the list, it is
3765					 * now safe to drop li_list_mtx
3766					 */
3767					mutex_unlock(&eli->li_list_mtx);
3768					err = ext4_run_li_request(elr);
3769					sb_end_write(elr->lr_super);
3770					mutex_lock(&eli->li_list_mtx);
3771					n = pos->next;
3772				}
3773				up_read((&elr->lr_super->s_umount));
3774			}
3775			/* error, remove the lazy_init job */
3776			if (err) {
3777				ext4_remove_li_request(elr);
3778				continue;
3779			}
3780			if (!progress) {
3781				elr->lr_next_sched = jiffies +
3782					get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
 
3783			}
3784			if (time_before(elr->lr_next_sched, next_wakeup))
3785				next_wakeup = elr->lr_next_sched;
3786		}
3787		mutex_unlock(&eli->li_list_mtx);
3788
3789		try_to_freeze();
3790
3791		cur = jiffies;
3792		if ((time_after_eq(cur, next_wakeup)) ||
3793		    (MAX_JIFFY_OFFSET == next_wakeup)) {
3794			cond_resched();
3795			continue;
3796		}
3797
3798		schedule_timeout_interruptible(next_wakeup - cur);
3799
3800		if (kthread_should_stop()) {
3801			ext4_clear_request_list();
3802			goto exit_thread;
3803		}
3804	}
3805
3806exit_thread:
3807	/*
3808	 * It looks like the request list is empty, but we need
3809	 * to check it under the li_list_mtx lock, to prevent any
3810	 * additions into it, and of course we should lock ext4_li_mtx
3811	 * to atomically free the list and ext4_li_info, because at
3812	 * this point another ext4 filesystem could be registering
3813	 * new one.
3814	 */
3815	mutex_lock(&ext4_li_mtx);
3816	mutex_lock(&eli->li_list_mtx);
3817	if (!list_empty(&eli->li_request_list)) {
3818		mutex_unlock(&eli->li_list_mtx);
3819		mutex_unlock(&ext4_li_mtx);
3820		goto cont_thread;
3821	}
3822	mutex_unlock(&eli->li_list_mtx);
3823	kfree(ext4_li_info);
3824	ext4_li_info = NULL;
3825	mutex_unlock(&ext4_li_mtx);
3826
3827	return 0;
3828}
3829
3830static void ext4_clear_request_list(void)
3831{
3832	struct list_head *pos, *n;
3833	struct ext4_li_request *elr;
3834
3835	mutex_lock(&ext4_li_info->li_list_mtx);
3836	list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3837		elr = list_entry(pos, struct ext4_li_request,
3838				 lr_request);
3839		ext4_remove_li_request(elr);
3840	}
3841	mutex_unlock(&ext4_li_info->li_list_mtx);
3842}
3843
3844static int ext4_run_lazyinit_thread(void)
3845{
3846	ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3847					 ext4_li_info, "ext4lazyinit");
3848	if (IS_ERR(ext4_lazyinit_task)) {
3849		int err = PTR_ERR(ext4_lazyinit_task);
3850		ext4_clear_request_list();
3851		kfree(ext4_li_info);
3852		ext4_li_info = NULL;
3853		printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3854				 "initialization thread\n",
3855				 err);
3856		return err;
3857	}
3858	ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3859	return 0;
3860}
3861
3862/*
3863 * Check whether it make sense to run itable init. thread or not.
3864 * If there is at least one uninitialized inode table, return
3865 * corresponding group number, else the loop goes through all
3866 * groups and return total number of groups.
3867 */
3868static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3869{
3870	ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3871	struct ext4_group_desc *gdp = NULL;
3872
3873	if (!ext4_has_group_desc_csum(sb))
3874		return ngroups;
3875
3876	for (group = 0; group < ngroups; group++) {
3877		gdp = ext4_get_group_desc(sb, group, NULL);
3878		if (!gdp)
3879			continue;
3880
3881		if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3882			break;
3883	}
3884
3885	return group;
3886}
3887
3888static int ext4_li_info_new(void)
3889{
3890	struct ext4_lazy_init *eli = NULL;
3891
3892	eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3893	if (!eli)
3894		return -ENOMEM;
3895
3896	INIT_LIST_HEAD(&eli->li_request_list);
3897	mutex_init(&eli->li_list_mtx);
3898
3899	eli->li_state |= EXT4_LAZYINIT_QUIT;
3900
3901	ext4_li_info = eli;
3902
3903	return 0;
3904}
3905
3906static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3907					    ext4_group_t start)
3908{
3909	struct ext4_li_request *elr;
3910
3911	elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3912	if (!elr)
3913		return NULL;
3914
3915	elr->lr_super = sb;
3916	elr->lr_first_not_zeroed = start;
3917	if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3918		elr->lr_mode = EXT4_LI_MODE_ITABLE;
3919		elr->lr_next_group = start;
3920	} else {
3921		elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3922	}
3923
3924	/*
3925	 * Randomize first schedule time of the request to
3926	 * spread the inode table initialization requests
3927	 * better.
3928	 */
3929	elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
 
3930	return elr;
3931}
3932
3933int ext4_register_li_request(struct super_block *sb,
3934			     ext4_group_t first_not_zeroed)
3935{
3936	struct ext4_sb_info *sbi = EXT4_SB(sb);
3937	struct ext4_li_request *elr = NULL;
3938	ext4_group_t ngroups = sbi->s_groups_count;
3939	int ret = 0;
3940
3941	mutex_lock(&ext4_li_mtx);
3942	if (sbi->s_li_request != NULL) {
3943		/*
3944		 * Reset timeout so it can be computed again, because
3945		 * s_li_wait_mult might have changed.
3946		 */
3947		sbi->s_li_request->lr_timeout = 0;
3948		goto out;
3949	}
3950
3951	if (sb_rdonly(sb) ||
3952	    (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
3953	     (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE))))
3954		goto out;
3955
3956	elr = ext4_li_request_new(sb, first_not_zeroed);
3957	if (!elr) {
3958		ret = -ENOMEM;
3959		goto out;
3960	}
3961
3962	if (NULL == ext4_li_info) {
3963		ret = ext4_li_info_new();
3964		if (ret)
3965			goto out;
3966	}
3967
3968	mutex_lock(&ext4_li_info->li_list_mtx);
3969	list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3970	mutex_unlock(&ext4_li_info->li_list_mtx);
3971
3972	sbi->s_li_request = elr;
3973	/*
3974	 * set elr to NULL here since it has been inserted to
3975	 * the request_list and the removal and free of it is
3976	 * handled by ext4_clear_request_list from now on.
3977	 */
3978	elr = NULL;
3979
3980	if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3981		ret = ext4_run_lazyinit_thread();
3982		if (ret)
3983			goto out;
3984	}
3985out:
3986	mutex_unlock(&ext4_li_mtx);
3987	if (ret)
3988		kfree(elr);
3989	return ret;
3990}
3991
3992/*
3993 * We do not need to lock anything since this is called on
3994 * module unload.
3995 */
3996static void ext4_destroy_lazyinit_thread(void)
3997{
3998	/*
3999	 * If thread exited earlier
4000	 * there's nothing to be done.
4001	 */
4002	if (!ext4_li_info || !ext4_lazyinit_task)
4003		return;
4004
4005	kthread_stop(ext4_lazyinit_task);
4006}
4007
4008static int set_journal_csum_feature_set(struct super_block *sb)
4009{
4010	int ret = 1;
4011	int compat, incompat;
4012	struct ext4_sb_info *sbi = EXT4_SB(sb);
4013
4014	if (ext4_has_metadata_csum(sb)) {
4015		/* journal checksum v3 */
4016		compat = 0;
4017		incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
4018	} else {
4019		/* journal checksum v1 */
4020		compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4021		incompat = 0;
4022	}
4023
4024	jbd2_journal_clear_features(sbi->s_journal,
4025			JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4026			JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4027			JBD2_FEATURE_INCOMPAT_CSUM_V2);
4028	if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4029		ret = jbd2_journal_set_features(sbi->s_journal,
4030				compat, 0,
4031				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4032				incompat);
4033	} else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4034		ret = jbd2_journal_set_features(sbi->s_journal,
4035				compat, 0,
4036				incompat);
4037		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4038				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4039	} else {
4040		jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4041				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4042	}
4043
4044	return ret;
4045}
4046
4047/*
4048 * Note: calculating the overhead so we can be compatible with
4049 * historical BSD practice is quite difficult in the face of
4050 * clusters/bigalloc.  This is because multiple metadata blocks from
4051 * different block group can end up in the same allocation cluster.
4052 * Calculating the exact overhead in the face of clustered allocation
4053 * requires either O(all block bitmaps) in memory or O(number of block
4054 * groups**2) in time.  We will still calculate the superblock for
4055 * older file systems --- and if we come across with a bigalloc file
4056 * system with zero in s_overhead_clusters the estimate will be close to
4057 * correct especially for very large cluster sizes --- but for newer
4058 * file systems, it's better to calculate this figure once at mkfs
4059 * time, and store it in the superblock.  If the superblock value is
4060 * present (even for non-bigalloc file systems), we will use it.
4061 */
4062static int count_overhead(struct super_block *sb, ext4_group_t grp,
4063			  char *buf)
4064{
4065	struct ext4_sb_info	*sbi = EXT4_SB(sb);
4066	struct ext4_group_desc	*gdp;
4067	ext4_fsblk_t		first_block, last_block, b;
4068	ext4_group_t		i, ngroups = ext4_get_groups_count(sb);
4069	int			s, j, count = 0;
4070	int			has_super = ext4_bg_has_super(sb, grp);
4071
4072	if (!ext4_has_feature_bigalloc(sb))
4073		return (has_super + ext4_bg_num_gdb(sb, grp) +
4074			(has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
4075			sbi->s_itb_per_group + 2);
4076
4077	first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4078		(grp * EXT4_BLOCKS_PER_GROUP(sb));
4079	last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4080	for (i = 0; i < ngroups; i++) {
4081		gdp = ext4_get_group_desc(sb, i, NULL);
4082		b = ext4_block_bitmap(sb, gdp);
4083		if (b >= first_block && b <= last_block) {
4084			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4085			count++;
4086		}
4087		b = ext4_inode_bitmap(sb, gdp);
4088		if (b >= first_block && b <= last_block) {
4089			ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4090			count++;
4091		}
4092		b = ext4_inode_table(sb, gdp);
4093		if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4094			for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4095				int c = EXT4_B2C(sbi, b - first_block);
4096				ext4_set_bit(c, buf);
4097				count++;
4098			}
4099		if (i != grp)
4100			continue;
4101		s = 0;
4102		if (ext4_bg_has_super(sb, grp)) {
4103			ext4_set_bit(s++, buf);
4104			count++;
4105		}
4106		j = ext4_bg_num_gdb(sb, grp);
4107		if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4108			ext4_error(sb, "Invalid number of block group "
4109				   "descriptor blocks: %d", j);
4110			j = EXT4_BLOCKS_PER_GROUP(sb) - s;
4111		}
4112		count += j;
4113		for (; j > 0; j--)
4114			ext4_set_bit(EXT4_B2C(sbi, s++), buf);
4115	}
4116	if (!count)
4117		return 0;
4118	return EXT4_CLUSTERS_PER_GROUP(sb) -
4119		ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4120}
4121
4122/*
4123 * Compute the overhead and stash it in sbi->s_overhead
4124 */
4125int ext4_calculate_overhead(struct super_block *sb)
4126{
4127	struct ext4_sb_info *sbi = EXT4_SB(sb);
4128	struct ext4_super_block *es = sbi->s_es;
4129	struct inode *j_inode;
4130	unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
4131	ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4132	ext4_fsblk_t overhead = 0;
4133	char *buf = (char *) get_zeroed_page(GFP_NOFS);
4134
4135	if (!buf)
4136		return -ENOMEM;
4137
4138	/*
4139	 * Compute the overhead (FS structures).  This is constant
4140	 * for a given filesystem unless the number of block groups
4141	 * changes so we cache the previous value until it does.
4142	 */
4143
4144	/*
4145	 * All of the blocks before first_data_block are overhead
4146	 */
4147	overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4148
4149	/*
4150	 * Add the overhead found in each block group
4151	 */
4152	for (i = 0; i < ngroups; i++) {
4153		int blks;
4154
4155		blks = count_overhead(sb, i, buf);
4156		overhead += blks;
4157		if (blks)
4158			memset(buf, 0, PAGE_SIZE);
4159		cond_resched();
4160	}
4161
4162	/*
4163	 * Add the internal journal blocks whether the journal has been
4164	 * loaded or not
4165	 */
4166	if (sbi->s_journal && !sbi->s_journal_bdev)
4167		overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
4168	else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4169		/* j_inum for internal journal is non-zero */
4170		j_inode = ext4_get_journal_inode(sb, j_inum);
4171		if (j_inode) {
4172			j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4173			overhead += EXT4_NUM_B2C(sbi, j_blocks);
4174			iput(j_inode);
4175		} else {
4176			ext4_msg(sb, KERN_ERR, "can't get journal size");
4177		}
4178	}
4179	sbi->s_overhead = overhead;
4180	smp_wmb();
4181	free_page((unsigned long) buf);
4182	return 0;
4183}
4184
4185static void ext4_set_resv_clusters(struct super_block *sb)
4186{
4187	ext4_fsblk_t resv_clusters;
4188	struct ext4_sb_info *sbi = EXT4_SB(sb);
4189
4190	/*
4191	 * There's no need to reserve anything when we aren't using extents.
4192	 * The space estimates are exact, there are no unwritten extents,
4193	 * hole punching doesn't need new metadata... This is needed especially
4194	 * to keep ext2/3 backward compatibility.
4195	 */
4196	if (!ext4_has_feature_extents(sb))
4197		return;
4198	/*
4199	 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4200	 * This should cover the situations where we can not afford to run
4201	 * out of space like for example punch hole, or converting
4202	 * unwritten extents in delalloc path. In most cases such
4203	 * allocation would require 1, or 2 blocks, higher numbers are
4204	 * very rare.
4205	 */
4206	resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4207			 sbi->s_cluster_bits);
4208
4209	do_div(resv_clusters, 50);
4210	resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4211
4212	atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4213}
4214
4215static const char *ext4_quota_mode(struct super_block *sb)
4216{
4217#ifdef CONFIG_QUOTA
4218	if (!ext4_quota_capable(sb))
4219		return "none";
4220
4221	if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4222		return "journalled";
4223	else
4224		return "writeback";
4225#else
4226	return "disabled";
4227#endif
4228}
4229
4230static void ext4_setup_csum_trigger(struct super_block *sb,
4231				    enum ext4_journal_trigger_type type,
4232				    void (*trigger)(
4233					struct jbd2_buffer_trigger_type *type,
4234					struct buffer_head *bh,
4235					void *mapped_data,
4236					size_t size))
4237{
4238	struct ext4_sb_info *sbi = EXT4_SB(sb);
4239
4240	sbi->s_journal_triggers[type].sb = sb;
4241	sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4242}
4243
4244static void ext4_free_sbi(struct ext4_sb_info *sbi)
4245{
4246	if (!sbi)
4247		return;
4248
4249	kfree(sbi->s_blockgroup_lock);
4250	fs_put_dax(sbi->s_daxdev, NULL);
4251	kfree(sbi);
4252}
4253
4254static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4255{
4256	struct ext4_sb_info *sbi;
 
 
 
4257
4258	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4259	if (!sbi)
4260		return NULL;
 
4261
4262	sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
4263					   NULL, NULL);
4264
 
4265	sbi->s_blockgroup_lock =
4266		kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4267
4268	if (!sbi->s_blockgroup_lock)
4269		goto err_out;
4270
4271	sb->s_fs_info = sbi;
4272	sbi->s_sb = sb;
4273	return sbi;
4274err_out:
4275	fs_put_dax(sbi->s_daxdev, NULL);
4276	kfree(sbi);
4277	return NULL;
4278}
4279
4280static void ext4_set_def_opts(struct super_block *sb,
4281			      struct ext4_super_block *es)
4282{
4283	unsigned long def_mount_opts;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4284
4285	/* Set defaults before we parse the mount options */
4286	def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4287	set_opt(sb, INIT_INODE_TABLE);
4288	if (def_mount_opts & EXT4_DEFM_DEBUG)
4289		set_opt(sb, DEBUG);
4290	if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4291		set_opt(sb, GRPID);
4292	if (def_mount_opts & EXT4_DEFM_UID16)
4293		set_opt(sb, NO_UID32);
4294	/* xattr user namespace & acls are now defaulted on */
4295	set_opt(sb, XATTR_USER);
4296#ifdef CONFIG_EXT4_FS_POSIX_ACL
4297	set_opt(sb, POSIX_ACL);
4298#endif
4299	if (ext4_has_feature_fast_commit(sb))
4300		set_opt2(sb, JOURNAL_FAST_COMMIT);
4301	/* don't forget to enable journal_csum when metadata_csum is enabled. */
4302	if (ext4_has_metadata_csum(sb))
4303		set_opt(sb, JOURNAL_CHECKSUM);
4304
4305	if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4306		set_opt(sb, JOURNAL_DATA);
4307	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4308		set_opt(sb, ORDERED_DATA);
4309	else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4310		set_opt(sb, WRITEBACK_DATA);
4311
4312	if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC)
4313		set_opt(sb, ERRORS_PANIC);
4314	else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE)
4315		set_opt(sb, ERRORS_CONT);
4316	else
4317		set_opt(sb, ERRORS_RO);
4318	/* block_validity enabled by default; disable with noblock_validity */
4319	set_opt(sb, BLOCK_VALIDITY);
4320	if (def_mount_opts & EXT4_DEFM_DISCARD)
4321		set_opt(sb, DISCARD);
4322
 
 
 
 
 
 
4323	if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4324		set_opt(sb, BARRIER);
4325
4326	/*
4327	 * enable delayed allocation by default
4328	 * Use -o nodelalloc to turn it off
4329	 */
4330	if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4331	    ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4332		set_opt(sb, DELALLOC);
4333
4334	if (sb->s_blocksize == PAGE_SIZE)
4335		set_opt(sb, DIOREAD_NOLOCK);
4336}
4337
4338static int ext4_handle_clustersize(struct super_block *sb)
4339{
4340	struct ext4_sb_info *sbi = EXT4_SB(sb);
4341	struct ext4_super_block *es = sbi->s_es;
4342	int clustersize;
4343
4344	/* Handle clustersize */
4345	clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4346	if (ext4_has_feature_bigalloc(sb)) {
4347		if (clustersize < sb->s_blocksize) {
4348			ext4_msg(sb, KERN_ERR,
4349				 "cluster size (%d) smaller than "
4350				 "block size (%lu)", clustersize, sb->s_blocksize);
4351			return -EINVAL;
4352		}
4353		sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4354			le32_to_cpu(es->s_log_block_size);
4355		sbi->s_clusters_per_group =
4356			le32_to_cpu(es->s_clusters_per_group);
4357		if (sbi->s_clusters_per_group > sb->s_blocksize * 8) {
4358			ext4_msg(sb, KERN_ERR,
4359				 "#clusters per group too big: %lu",
4360				 sbi->s_clusters_per_group);
4361			return -EINVAL;
4362		}
4363		if (sbi->s_blocks_per_group !=
4364		    (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) {
4365			ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4366				 "clusters per group (%lu) inconsistent",
4367				 sbi->s_blocks_per_group,
4368				 sbi->s_clusters_per_group);
4369			return -EINVAL;
4370		}
4371	} else {
4372		if (clustersize != sb->s_blocksize) {
4373			ext4_msg(sb, KERN_ERR,
4374				 "fragment/cluster size (%d) != "
4375				 "block size (%lu)", clustersize, sb->s_blocksize);
4376			return -EINVAL;
4377		}
4378		if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
4379			ext4_msg(sb, KERN_ERR,
4380				 "#blocks per group too big: %lu",
4381				 sbi->s_blocks_per_group);
4382			return -EINVAL;
4383		}
4384		sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4385		sbi->s_cluster_bits = 0;
4386	}
4387	sbi->s_cluster_ratio = clustersize / sb->s_blocksize;
4388
4389	/* Do we have standard group size of clustersize * 8 blocks ? */
4390	if (sbi->s_blocks_per_group == clustersize << 3)
4391		set_opt2(sb, STD_GROUP_SIZE);
4392
4393	return 0;
4394}
4395
4396static void ext4_fast_commit_init(struct super_block *sb)
4397{
4398	struct ext4_sb_info *sbi = EXT4_SB(sb);
4399
4400	/* Initialize fast commit stuff */
4401	atomic_set(&sbi->s_fc_subtid, 0);
4402	INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4403	INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4404	INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4405	INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4406	sbi->s_fc_bytes = 0;
4407	ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4408	sbi->s_fc_ineligible_tid = 0;
4409	spin_lock_init(&sbi->s_fc_lock);
4410	memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4411	sbi->s_fc_replay_state.fc_regions = NULL;
4412	sbi->s_fc_replay_state.fc_regions_size = 0;
4413	sbi->s_fc_replay_state.fc_regions_used = 0;
4414	sbi->s_fc_replay_state.fc_regions_valid = 0;
4415	sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4416	sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4417	sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4418}
4419
4420static int ext4_inode_info_init(struct super_block *sb,
4421				struct ext4_super_block *es)
4422{
4423	struct ext4_sb_info *sbi = EXT4_SB(sb);
4424
4425	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4426		sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4427		sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4428	} else {
4429		sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4430		sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4431		if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4432			ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4433				 sbi->s_first_ino);
4434			return -EINVAL;
4435		}
4436		if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4437		    (!is_power_of_2(sbi->s_inode_size)) ||
4438		    (sbi->s_inode_size > sb->s_blocksize)) {
4439			ext4_msg(sb, KERN_ERR,
4440			       "unsupported inode size: %d",
4441			       sbi->s_inode_size);
4442			ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize);
4443			return -EINVAL;
4444		}
4445		/*
4446		 * i_atime_extra is the last extra field available for
4447		 * [acm]times in struct ext4_inode. Checking for that
4448		 * field should suffice to ensure we have extra space
4449		 * for all three.
4450		 */
4451		if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4452			sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4453			sb->s_time_gran = 1;
4454			sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4455		} else {
4456			sb->s_time_gran = NSEC_PER_SEC;
4457			sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4458		}
4459		sb->s_time_min = EXT4_TIMESTAMP_MIN;
4460	}
4461
4462	if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4463		sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4464			EXT4_GOOD_OLD_INODE_SIZE;
4465		if (ext4_has_feature_extra_isize(sb)) {
4466			unsigned v, max = (sbi->s_inode_size -
4467					   EXT4_GOOD_OLD_INODE_SIZE);
4468
4469			v = le16_to_cpu(es->s_want_extra_isize);
4470			if (v > max) {
4471				ext4_msg(sb, KERN_ERR,
4472					 "bad s_want_extra_isize: %d", v);
4473				return -EINVAL;
4474			}
4475			if (sbi->s_want_extra_isize < v)
4476				sbi->s_want_extra_isize = v;
4477
4478			v = le16_to_cpu(es->s_min_extra_isize);
4479			if (v > max) {
4480				ext4_msg(sb, KERN_ERR,
4481					 "bad s_min_extra_isize: %d", v);
4482				return -EINVAL;
4483			}
4484			if (sbi->s_want_extra_isize < v)
4485				sbi->s_want_extra_isize = v;
4486		}
4487	}
4488
4489	return 0;
4490}
4491
4492#if IS_ENABLED(CONFIG_UNICODE)
4493static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4494{
4495	const struct ext4_sb_encodings *encoding_info;
4496	struct unicode_map *encoding;
4497	__u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
4498
4499	if (!ext4_has_feature_casefold(sb) || sb->s_encoding)
4500		return 0;
4501
4502	encoding_info = ext4_sb_read_encoding(es);
4503	if (!encoding_info) {
4504		ext4_msg(sb, KERN_ERR,
4505			"Encoding requested by superblock is unknown");
4506		return -EINVAL;
4507	}
4508
4509	encoding = utf8_load(encoding_info->version);
4510	if (IS_ERR(encoding)) {
4511		ext4_msg(sb, KERN_ERR,
4512			"can't mount with superblock charset: %s-%u.%u.%u "
4513			"not supported by the kernel. flags: 0x%x.",
4514			encoding_info->name,
4515			unicode_major(encoding_info->version),
4516			unicode_minor(encoding_info->version),
4517			unicode_rev(encoding_info->version),
4518			encoding_flags);
4519		return -EINVAL;
4520	}
4521	ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4522		"%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4523		unicode_major(encoding_info->version),
4524		unicode_minor(encoding_info->version),
4525		unicode_rev(encoding_info->version),
4526		encoding_flags);
4527
4528	sb->s_encoding = encoding;
4529	sb->s_encoding_flags = encoding_flags;
4530
4531	return 0;
4532}
4533#else
4534static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4535{
4536	return 0;
4537}
4538#endif
4539
4540static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es)
4541{
4542	struct ext4_sb_info *sbi = EXT4_SB(sb);
 
 
4543
4544	/* Warn if metadata_csum and gdt_csum are both set. */
4545	if (ext4_has_feature_metadata_csum(sb) &&
4546	    ext4_has_feature_gdt_csum(sb))
4547		ext4_warning(sb, "metadata_csum and uninit_bg are "
4548			     "redundant flags; please run fsck.");
 
4549
4550	/* Check for a known checksum algorithm */
4551	if (!ext4_verify_csum_type(sb, es)) {
4552		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4553			 "unknown checksum algorithm.");
4554		return -EINVAL;
4555	}
4556	ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4557				ext4_orphan_file_block_trigger);
 
 
 
 
4558
4559	/* Load the checksum driver */
4560	sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4561	if (IS_ERR(sbi->s_chksum_driver)) {
4562		int ret = PTR_ERR(sbi->s_chksum_driver);
4563		ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4564		sbi->s_chksum_driver = NULL;
4565		return ret;
4566	}
 
4567
4568	/* Check superblock checksum */
4569	if (!ext4_superblock_csum_verify(sb, es)) {
4570		ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4571			 "invalid superblock checksum.  Run e2fsck?");
4572		return -EFSBADCRC;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4573	}
4574
4575	/* Precompute checksum seed for all metadata */
4576	if (ext4_has_feature_csum_seed(sb))
4577		sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4578	else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4579		sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4580					       sizeof(es->s_uuid));
4581	return 0;
4582}
4583
4584static int ext4_check_feature_compatibility(struct super_block *sb,
4585					    struct ext4_super_block *es,
4586					    int silent)
4587{
4588	if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4589	    (ext4_has_compat_features(sb) ||
4590	     ext4_has_ro_compat_features(sb) ||
4591	     ext4_has_incompat_features(sb)))
4592		ext4_msg(sb, KERN_WARNING,
4593		       "feature flags set on rev 0 fs, "
4594		       "running e2fsck is recommended");
4595
4596	if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4597		set_opt2(sb, HURD_COMPAT);
4598		if (ext4_has_feature_64bit(sb)) {
4599			ext4_msg(sb, KERN_ERR,
4600				 "The Hurd can't support 64-bit file systems");
4601			return -EINVAL;
4602		}
4603
4604		/*
4605		 * ea_inode feature uses l_i_version field which is not
4606		 * available in HURD_COMPAT mode.
4607		 */
4608		if (ext4_has_feature_ea_inode(sb)) {
4609			ext4_msg(sb, KERN_ERR,
4610				 "ea_inode feature is not supported for Hurd");
4611			return -EINVAL;
4612		}
4613	}
4614
4615	if (IS_EXT2_SB(sb)) {
4616		if (ext2_feature_set_ok(sb))
4617			ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4618				 "using the ext4 subsystem");
4619		else {
4620			/*
4621			 * If we're probing be silent, if this looks like
4622			 * it's actually an ext[34] filesystem.
4623			 */
4624			if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4625				return -EINVAL;
4626			ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4627				 "to feature incompatibilities");
4628			return -EINVAL;
4629		}
4630	}
4631
4632	if (IS_EXT3_SB(sb)) {
4633		if (ext3_feature_set_ok(sb))
4634			ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4635				 "using the ext4 subsystem");
4636		else {
4637			/*
4638			 * If we're probing be silent, if this looks like
4639			 * it's actually an ext4 filesystem.
4640			 */
4641			if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4642				return -EINVAL;
4643			ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4644				 "to feature incompatibilities");
4645			return -EINVAL;
4646		}
4647	}
4648
4649	/*
4650	 * Check feature flags regardless of the revision level, since we
4651	 * previously didn't change the revision level when setting the flags,
4652	 * so there is a chance incompat flags are set on a rev 0 filesystem.
4653	 */
4654	if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4655		return -EINVAL;
4656
4657	return 0;
4658}
4659
4660static int ext4_geometry_check(struct super_block *sb,
4661			       struct ext4_super_block *es)
4662{
4663	struct ext4_sb_info *sbi = EXT4_SB(sb);
4664	__u64 blocks_count;
4665
4666	/* check blocks count against device size */
4667	blocks_count = sb_bdev_nr_blocks(sb);
4668	if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4669		ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4670		       "exceeds size of device (%llu blocks)",
4671		       ext4_blocks_count(es), blocks_count);
4672		return -EINVAL;
4673	}
4674
4675	/*
4676	 * It makes no sense for the first data block to be beyond the end
4677	 * of the filesystem.
4678	 */
4679	if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4680		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4681			 "block %u is beyond end of filesystem (%llu)",
4682			 le32_to_cpu(es->s_first_data_block),
4683			 ext4_blocks_count(es));
4684		return -EINVAL;
4685	}
4686	if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4687	    (sbi->s_cluster_ratio == 1)) {
4688		ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4689			 "block is 0 with a 1k block and cluster size");
4690		return -EINVAL;
4691	}
4692
4693	blocks_count = (ext4_blocks_count(es) -
4694			le32_to_cpu(es->s_first_data_block) +
4695			EXT4_BLOCKS_PER_GROUP(sb) - 1);
4696	do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4697	if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4698		ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4699		       "(block count %llu, first data block %u, "
4700		       "blocks per group %lu)", blocks_count,
4701		       ext4_blocks_count(es),
4702		       le32_to_cpu(es->s_first_data_block),
4703		       EXT4_BLOCKS_PER_GROUP(sb));
4704		return -EINVAL;
4705	}
4706	sbi->s_groups_count = blocks_count;
4707	sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4708			(EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4709	if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4710	    le32_to_cpu(es->s_inodes_count)) {
4711		ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4712			 le32_to_cpu(es->s_inodes_count),
4713			 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4714		return -EINVAL;
4715	}
4716
4717	return 0;
4718}
4719
4720static void ext4_group_desc_free(struct ext4_sb_info *sbi)
4721{
4722	struct buffer_head **group_desc;
4723	int i;
4724
4725	rcu_read_lock();
4726	group_desc = rcu_dereference(sbi->s_group_desc);
4727	for (i = 0; i < sbi->s_gdb_count; i++)
4728		brelse(group_desc[i]);
4729	kvfree(group_desc);
4730	rcu_read_unlock();
4731}
4732
4733static int ext4_group_desc_init(struct super_block *sb,
4734				struct ext4_super_block *es,
4735				ext4_fsblk_t logical_sb_block,
4736				ext4_group_t *first_not_zeroed)
4737{
4738	struct ext4_sb_info *sbi = EXT4_SB(sb);
4739	unsigned int db_count;
4740	ext4_fsblk_t block;
4741	int ret;
4742	int i;
4743
4744	db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4745		   EXT4_DESC_PER_BLOCK(sb);
4746	if (ext4_has_feature_meta_bg(sb)) {
4747		if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4748			ext4_msg(sb, KERN_WARNING,
4749				 "first meta block group too large: %u "
4750				 "(group descriptor block count %u)",
4751				 le32_to_cpu(es->s_first_meta_bg), db_count);
4752			return -EINVAL;
4753		}
4754	}
4755	rcu_assign_pointer(sbi->s_group_desc,
4756			   kvmalloc_array(db_count,
4757					  sizeof(struct buffer_head *),
4758					  GFP_KERNEL));
4759	if (sbi->s_group_desc == NULL) {
4760		ext4_msg(sb, KERN_ERR, "not enough memory");
4761		return -ENOMEM;
4762	}
4763
4764	bgl_lock_init(sbi->s_blockgroup_lock);
4765
4766	/* Pre-read the descriptors into the buffer cache */
4767	for (i = 0; i < db_count; i++) {
4768		block = descriptor_loc(sb, logical_sb_block, i);
4769		ext4_sb_breadahead_unmovable(sb, block);
4770	}
4771
4772	for (i = 0; i < db_count; i++) {
4773		struct buffer_head *bh;
4774
4775		block = descriptor_loc(sb, logical_sb_block, i);
4776		bh = ext4_sb_bread_unmovable(sb, block);
4777		if (IS_ERR(bh)) {
4778			ext4_msg(sb, KERN_ERR,
4779			       "can't read group descriptor %d", i);
4780			sbi->s_gdb_count = i;
4781			ret = PTR_ERR(bh);
4782			goto out;
4783		}
4784		rcu_read_lock();
4785		rcu_dereference(sbi->s_group_desc)[i] = bh;
4786		rcu_read_unlock();
4787	}
4788	sbi->s_gdb_count = db_count;
4789	if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) {
4790		ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4791		ret = -EFSCORRUPTED;
4792		goto out;
4793	}
4794	return 0;
4795out:
4796	ext4_group_desc_free(sbi);
4797	return ret;
4798}
4799
4800static int ext4_load_and_init_journal(struct super_block *sb,
4801				      struct ext4_super_block *es,
4802				      struct ext4_fs_context *ctx)
4803{
4804	struct ext4_sb_info *sbi = EXT4_SB(sb);
4805	int err;
4806
4807	err = ext4_load_journal(sb, es, ctx->journal_devnum);
4808	if (err)
4809		return err;
4810
4811	if (ext4_has_feature_64bit(sb) &&
4812	    !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4813				       JBD2_FEATURE_INCOMPAT_64BIT)) {
4814		ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4815		goto out;
4816	}
4817
4818	if (!set_journal_csum_feature_set(sb)) {
4819		ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4820			 "feature set");
4821		goto out;
4822	}
4823
4824	if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4825		!jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4826					  JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4827		ext4_msg(sb, KERN_ERR,
4828			"Failed to set fast commit journal feature");
4829		goto out;
4830	}
4831
4832	/* We have now updated the journal if required, so we can
4833	 * validate the data journaling mode. */
4834	switch (test_opt(sb, DATA_FLAGS)) {
4835	case 0:
4836		/* No mode set, assume a default based on the journal
4837		 * capabilities: ORDERED_DATA if the journal can
4838		 * cope, else JOURNAL_DATA
4839		 */
4840		if (jbd2_journal_check_available_features
4841		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4842			set_opt(sb, ORDERED_DATA);
4843			sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4844		} else {
4845			set_opt(sb, JOURNAL_DATA);
4846			sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4847		}
4848		break;
4849
4850	case EXT4_MOUNT_ORDERED_DATA:
4851	case EXT4_MOUNT_WRITEBACK_DATA:
4852		if (!jbd2_journal_check_available_features
4853		    (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4854			ext4_msg(sb, KERN_ERR, "Journal does not support "
4855			       "requested data journaling mode");
4856			goto out;
4857		}
4858		break;
4859	default:
4860		break;
4861	}
4862
4863	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4864	    test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4865		ext4_msg(sb, KERN_ERR, "can't mount with "
4866			"journal_async_commit in data=ordered mode");
4867		goto out;
4868	}
4869
4870	set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
4871
4872	sbi->s_journal->j_submit_inode_data_buffers =
4873		ext4_journal_submit_inode_data_buffers;
4874	sbi->s_journal->j_finish_inode_data_buffers =
4875		ext4_journal_finish_inode_data_buffers;
4876
4877	return 0;
4878
4879out:
4880	/* flush s_error_work before journal destroy. */
4881	flush_work(&sbi->s_error_work);
4882	jbd2_journal_destroy(sbi->s_journal);
4883	sbi->s_journal = NULL;
4884	return -EINVAL;
4885}
4886
4887static int ext4_journal_data_mode_check(struct super_block *sb)
4888{
4889	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4890		printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with "
4891			    "data=journal disables delayed allocation, "
4892			    "dioread_nolock, O_DIRECT and fast_commit support!\n");
4893		/* can't mount with both data=journal and dioread_nolock. */
4894		clear_opt(sb, DIOREAD_NOLOCK);
4895		clear_opt2(sb, JOURNAL_FAST_COMMIT);
4896		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4897			ext4_msg(sb, KERN_ERR, "can't mount with "
4898				 "both data=journal and delalloc");
4899			return -EINVAL;
4900		}
4901		if (test_opt(sb, DAX_ALWAYS)) {
4902			ext4_msg(sb, KERN_ERR, "can't mount with "
4903				 "both data=journal and dax");
4904			return -EINVAL;
4905		}
4906		if (ext4_has_feature_encrypt(sb)) {
4907			ext4_msg(sb, KERN_WARNING,
4908				 "encrypted files will use data=ordered "
4909				 "instead of data journaling mode");
4910		}
4911		if (test_opt(sb, DELALLOC))
4912			clear_opt(sb, DELALLOC);
4913	} else {
4914		sb->s_iflags |= SB_I_CGROUPWB;
4915	}
4916
4917	return 0;
4918}
4919
4920static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb,
4921			   int silent)
4922{
4923	struct ext4_sb_info *sbi = EXT4_SB(sb);
4924	struct ext4_super_block *es;
4925	ext4_fsblk_t logical_sb_block;
4926	unsigned long offset = 0;
4927	struct buffer_head *bh;
4928	int ret = -EINVAL;
4929	int blocksize;
4930
4931	blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
4932	if (!blocksize) {
4933		ext4_msg(sb, KERN_ERR, "unable to set blocksize");
4934		return -EINVAL;
4935	}
4936
4937	/*
4938	 * The ext4 superblock will not be buffer aligned for other than 1kB
4939	 * block sizes.  We need to calculate the offset from buffer start.
4940	 */
4941	if (blocksize != EXT4_MIN_BLOCK_SIZE) {
4942		logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
4943		offset = do_div(logical_sb_block, blocksize);
4944	} else {
4945		logical_sb_block = sbi->s_sb_block;
4946	}
4947
4948	bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4949	if (IS_ERR(bh)) {
4950		ext4_msg(sb, KERN_ERR, "unable to read superblock");
4951		return PTR_ERR(bh);
4952	}
4953	/*
4954	 * Note: s_es must be initialized as soon as possible because
4955	 *       some ext4 macro-instructions depend on its value
4956	 */
4957	es = (struct ext4_super_block *) (bh->b_data + offset);
4958	sbi->s_es = es;
4959	sb->s_magic = le16_to_cpu(es->s_magic);
4960	if (sb->s_magic != EXT4_SUPER_MAGIC) {
4961		if (!silent)
4962			ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4963		goto out;
4964	}
4965
4966	if (le32_to_cpu(es->s_log_block_size) >
4967	    (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4968		ext4_msg(sb, KERN_ERR,
4969			 "Invalid log block size: %u",
4970			 le32_to_cpu(es->s_log_block_size));
4971		goto out;
4972	}
4973	if (le32_to_cpu(es->s_log_cluster_size) >
4974	    (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4975		ext4_msg(sb, KERN_ERR,
4976			 "Invalid log cluster size: %u",
4977			 le32_to_cpu(es->s_log_cluster_size));
4978		goto out;
4979	}
4980
4981	blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4982
4983	/*
4984	 * If the default block size is not the same as the real block size,
4985	 * we need to reload it.
4986	 */
4987	if (sb->s_blocksize == blocksize) {
4988		*lsb = logical_sb_block;
4989		sbi->s_sbh = bh;
4990		return 0;
4991	}
4992
4993	/*
4994	 * bh must be released before kill_bdev(), otherwise
4995	 * it won't be freed and its page also. kill_bdev()
4996	 * is called by sb_set_blocksize().
4997	 */
4998	brelse(bh);
4999	/* Validate the filesystem blocksize */
5000	if (!sb_set_blocksize(sb, blocksize)) {
5001		ext4_msg(sb, KERN_ERR, "bad block size %d",
5002				blocksize);
5003		bh = NULL;
5004		goto out;
5005	}
5006
5007	logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5008	offset = do_div(logical_sb_block, blocksize);
5009	bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5010	if (IS_ERR(bh)) {
5011		ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try");
5012		ret = PTR_ERR(bh);
5013		bh = NULL;
5014		goto out;
5015	}
5016	es = (struct ext4_super_block *)(bh->b_data + offset);
5017	sbi->s_es = es;
5018	if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
5019		ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!");
5020		goto out;
5021	}
5022	*lsb = logical_sb_block;
5023	sbi->s_sbh = bh;
5024	return 0;
5025out:
5026	brelse(bh);
5027	return ret;
5028}
5029
5030static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
5031{
5032	struct ext4_super_block *es = NULL;
5033	struct ext4_sb_info *sbi = EXT4_SB(sb);
5034	struct flex_groups **flex_groups;
5035	ext4_fsblk_t block;
5036	ext4_fsblk_t logical_sb_block;
5037	struct inode *root;
5038	int ret = -ENOMEM;
5039	unsigned int i;
5040	int needs_recovery, has_huge_files;
5041	int err = 0;
5042	ext4_group_t first_not_zeroed;
5043	struct ext4_fs_context *ctx = fc->fs_private;
5044	int silent = fc->sb_flags & SB_SILENT;
5045
5046	/* Set defaults for the variables that will be set during parsing */
5047	if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
5048		ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5049
5050	sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
5051	sbi->s_sectors_written_start =
5052		part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
5053
5054	/* -EINVAL is default */
5055	ret = -EINVAL;
5056	err = ext4_load_super(sb, &logical_sb_block, silent);
5057	if (err)
5058		goto out_fail;
5059
5060	es = sbi->s_es;
5061	sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
5062
5063	err = ext4_init_metadata_csum(sb, es);
5064	if (err)
5065		goto failed_mount;
5066
5067	ext4_set_def_opts(sb, es);
5068
5069	sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
5070	sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
5071	sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
5072	sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
5073	sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
5074
5075	/*
5076	 * set default s_li_wait_mult for lazyinit, for the case there is
5077	 * no mount option specified.
5078	 */
5079	sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
5080
5081	if (ext4_inode_info_init(sb, es))
5082		goto failed_mount;
5083
5084	err = parse_apply_sb_mount_options(sb, ctx);
5085	if (err < 0)
5086		goto failed_mount;
5087
5088	sbi->s_def_mount_opt = sbi->s_mount_opt;
5089
5090	err = ext4_check_opt_consistency(fc, sb);
5091	if (err < 0)
5092		goto failed_mount;
5093
5094	ext4_apply_options(fc, sb);
5095
5096	if (ext4_encoding_init(sb, es))
5097		goto failed_mount;
5098
5099	if (ext4_journal_data_mode_check(sb))
5100		goto failed_mount;
5101
5102	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5103		(test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5104
5105	/* i_version is always enabled now */
5106	sb->s_flags |= SB_I_VERSION;
5107
5108	if (ext4_check_feature_compatibility(sb, es, silent))
5109		goto failed_mount;
5110
5111	if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) {
5112		ext4_msg(sb, KERN_ERR,
5113			 "Number of reserved GDT blocks insanely large: %d",
5114			 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
5115		goto failed_mount;
5116	}
5117
5118	if (sbi->s_daxdev) {
5119		if (sb->s_blocksize == PAGE_SIZE)
5120			set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
5121		else
5122			ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
5123	}
5124
5125	if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
5126		if (ext4_has_feature_inline_data(sb)) {
5127			ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
5128					" that may contain inline data");
5129			goto failed_mount;
5130		}
5131		if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
5132			ext4_msg(sb, KERN_ERR,
5133				"DAX unsupported by block device.");
5134			goto failed_mount;
5135		}
5136	}
5137
5138	if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
5139		ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
5140			 es->s_encryption_level);
5141		goto failed_mount;
5142	}
5143
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5144	has_huge_files = ext4_has_feature_huge_file(sb);
5145	sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
5146						      has_huge_files);
5147	sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
5148
5149	sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
5150	if (ext4_has_feature_64bit(sb)) {
5151		if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
5152		    sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
5153		    !is_power_of_2(sbi->s_desc_size)) {
5154			ext4_msg(sb, KERN_ERR,
5155			       "unsupported descriptor size %lu",
5156			       sbi->s_desc_size);
5157			goto failed_mount;
5158		}
5159	} else
5160		sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
5161
5162	sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
5163	sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
5164
5165	sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb);
5166	if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) {
5167		if (!silent)
5168			ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5169		goto failed_mount;
5170	}
5171	if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
5172	    sbi->s_inodes_per_group > sb->s_blocksize * 8) {
5173		ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
5174			 sbi->s_inodes_per_group);
5175		goto failed_mount;
5176	}
5177	sbi->s_itb_per_group = sbi->s_inodes_per_group /
5178					sbi->s_inodes_per_block;
5179	sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb);
5180	sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
 
5181	sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
5182	sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
5183
5184	for (i = 0; i < 4; i++)
5185		sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
5186	sbi->s_def_hash_version = es->s_def_hash_version;
5187	if (ext4_has_feature_dir_index(sb)) {
5188		i = le32_to_cpu(es->s_flags);
5189		if (i & EXT2_FLAGS_UNSIGNED_HASH)
5190			sbi->s_hash_unsigned = 3;
5191		else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
5192#ifdef __CHAR_UNSIGNED__
5193			if (!sb_rdonly(sb))
5194				es->s_flags |=
5195					cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
5196			sbi->s_hash_unsigned = 3;
5197#else
5198			if (!sb_rdonly(sb))
5199				es->s_flags |=
5200					cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
5201#endif
5202		}
5203	}
5204
5205	if (ext4_handle_clustersize(sb))
5206		goto failed_mount;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5207
5208	/*
5209	 * Test whether we have more sectors than will fit in sector_t,
5210	 * and whether the max offset is addressable by the page cache.
5211	 */
5212	err = generic_check_addressable(sb->s_blocksize_bits,
5213					ext4_blocks_count(es));
5214	if (err) {
5215		ext4_msg(sb, KERN_ERR, "filesystem"
5216			 " too large to mount safely on this system");
5217		goto failed_mount;
5218	}
5219
5220	if (ext4_geometry_check(sb, es))
 
 
 
 
 
 
 
 
5221		goto failed_mount;
 
5222
5223	err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed);
5224	if (err)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5225		goto failed_mount;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5226
5227	timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
5228	spin_lock_init(&sbi->s_error_lock);
5229	INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
5230
5231	/* Register extent status tree shrinker */
5232	if (ext4_es_register_shrinker(sbi))
5233		goto failed_mount3;
5234
5235	sbi->s_stripe = ext4_get_stripe_size(sbi);
5236	sbi->s_extent_max_zeroout_kb = 32;
5237
5238	/*
5239	 * set up enough so that it can read an inode
5240	 */
5241	sb->s_op = &ext4_sops;
5242	sb->s_export_op = &ext4_export_ops;
5243	sb->s_xattr = ext4_xattr_handlers;
5244#ifdef CONFIG_FS_ENCRYPTION
5245	sb->s_cop = &ext4_cryptops;
5246#endif
5247#ifdef CONFIG_FS_VERITY
5248	sb->s_vop = &ext4_verityops;
5249#endif
5250#ifdef CONFIG_QUOTA
5251	sb->dq_op = &ext4_quota_operations;
5252	if (ext4_has_feature_quota(sb))
5253		sb->s_qcop = &dquot_quotactl_sysfile_ops;
5254	else
5255		sb->s_qcop = &ext4_qctl_operations;
5256	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
5257#endif
5258	memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
5259
5260	INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
5261	mutex_init(&sbi->s_orphan_lock);
5262
5263	ext4_fast_commit_init(sb);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5264
5265	sb->s_root = NULL;
5266
5267	needs_recovery = (es->s_last_orphan != 0 ||
5268			  ext4_has_feature_orphan_present(sb) ||
5269			  ext4_has_feature_journal_needs_recovery(sb));
5270
5271	if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
5272		if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
5273			goto failed_mount3a;
5274
5275	/*
5276	 * The first inode we look at is the journal inode.  Don't try
5277	 * root first: it may be modified in the journal!
5278	 */
5279	if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
5280		err = ext4_load_and_init_journal(sb, es, ctx);
5281		if (err)
5282			goto failed_mount3a;
5283	} else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
5284		   ext4_has_feature_journal_needs_recovery(sb)) {
5285		ext4_msg(sb, KERN_ERR, "required journal recovery "
5286		       "suppressed and not mounted read-only");
5287		goto failed_mount3a;
5288	} else {
5289		/* Nojournal mode, all journal mount options are illegal */
5290		if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5291			ext4_msg(sb, KERN_ERR, "can't mount with "
5292				 "journal_async_commit, fs mounted w/o journal");
5293			goto failed_mount3a;
5294		}
5295
5296		if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
5297			ext4_msg(sb, KERN_ERR, "can't mount with "
5298				 "journal_checksum, fs mounted w/o journal");
5299			goto failed_mount3a;
 
 
 
 
 
5300		}
5301		if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5302			ext4_msg(sb, KERN_ERR, "can't mount with "
5303				 "commit=%lu, fs mounted w/o journal",
5304				 sbi->s_commit_interval / HZ);
5305			goto failed_mount3a;
5306		}
5307		if (EXT4_MOUNT_DATA_FLAGS &
5308		    (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5309			ext4_msg(sb, KERN_ERR, "can't mount with "
5310				 "data=, fs mounted w/o journal");
5311			goto failed_mount3a;
5312		}
5313		sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
5314		clear_opt(sb, JOURNAL_CHECKSUM);
5315		clear_opt(sb, DATA_FLAGS);
5316		clear_opt2(sb, JOURNAL_FAST_COMMIT);
5317		sbi->s_journal = NULL;
5318		needs_recovery = 0;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5319	}
5320
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5321	if (!test_opt(sb, NO_MBCACHE)) {
5322		sbi->s_ea_block_cache = ext4_xattr_create_cache();
5323		if (!sbi->s_ea_block_cache) {
5324			ext4_msg(sb, KERN_ERR,
5325				 "Failed to create ea_block_cache");
5326			goto failed_mount_wq;
5327		}
5328
5329		if (ext4_has_feature_ea_inode(sb)) {
5330			sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5331			if (!sbi->s_ea_inode_cache) {
5332				ext4_msg(sb, KERN_ERR,
5333					 "Failed to create ea_inode_cache");
5334				goto failed_mount_wq;
5335			}
5336		}
5337	}
5338
5339	if (ext4_has_feature_verity(sb) && sb->s_blocksize != PAGE_SIZE) {
5340		ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
5341		goto failed_mount_wq;
5342	}
5343
 
 
 
 
 
 
5344	/*
5345	 * Get the # of file system overhead blocks from the
5346	 * superblock if present.
5347	 */
5348	sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5349	/* ignore the precalculated value if it is ridiculous */
5350	if (sbi->s_overhead > ext4_blocks_count(es))
5351		sbi->s_overhead = 0;
5352	/*
5353	 * If the bigalloc feature is not enabled recalculating the
5354	 * overhead doesn't take long, so we might as well just redo
5355	 * it to make sure we are using the correct value.
5356	 */
5357	if (!ext4_has_feature_bigalloc(sb))
5358		sbi->s_overhead = 0;
5359	if (sbi->s_overhead == 0) {
5360		err = ext4_calculate_overhead(sb);
5361		if (err)
5362			goto failed_mount_wq;
5363	}
5364
5365	/*
5366	 * The maximum number of concurrent works can be high and
5367	 * concurrency isn't really necessary.  Limit it to 1.
5368	 */
5369	EXT4_SB(sb)->rsv_conversion_wq =
5370		alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5371	if (!EXT4_SB(sb)->rsv_conversion_wq) {
5372		printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
5373		ret = -ENOMEM;
5374		goto failed_mount4;
5375	}
5376
5377	/*
5378	 * The jbd2_journal_load will have done any necessary log recovery,
5379	 * so we can safely mount the rest of the filesystem now.
5380	 */
5381
5382	root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
5383	if (IS_ERR(root)) {
5384		ext4_msg(sb, KERN_ERR, "get root inode failed");
5385		ret = PTR_ERR(root);
5386		root = NULL;
5387		goto failed_mount4;
5388	}
5389	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
5390		ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
5391		iput(root);
5392		goto failed_mount4;
5393	}
5394
5395	sb->s_root = d_make_root(root);
5396	if (!sb->s_root) {
5397		ext4_msg(sb, KERN_ERR, "get root dentry failed");
5398		ret = -ENOMEM;
5399		goto failed_mount4;
5400	}
5401
5402	ret = ext4_setup_super(sb, es, sb_rdonly(sb));
5403	if (ret == -EROFS) {
5404		sb->s_flags |= SB_RDONLY;
5405		ret = 0;
5406	} else if (ret)
5407		goto failed_mount4a;
5408
5409	ext4_set_resv_clusters(sb);
5410
5411	if (test_opt(sb, BLOCK_VALIDITY)) {
5412		err = ext4_setup_system_zone(sb);
5413		if (err) {
5414			ext4_msg(sb, KERN_ERR, "failed to initialize system "
5415				 "zone (%d)", err);
5416			goto failed_mount4a;
5417		}
5418	}
5419	ext4_fc_replay_cleanup(sb);
5420
5421	ext4_ext_init(sb);
5422
5423	/*
5424	 * Enable optimize_scan if number of groups is > threshold. This can be
5425	 * turned off by passing "mb_optimize_scan=0". This can also be
5426	 * turned on forcefully by passing "mb_optimize_scan=1".
5427	 */
5428	if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5429		if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5430			set_opt2(sb, MB_OPTIMIZE_SCAN);
5431		else
5432			clear_opt2(sb, MB_OPTIMIZE_SCAN);
5433	}
5434
5435	err = ext4_mb_init(sb);
5436	if (err) {
5437		ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5438			 err);
5439		goto failed_mount5;
5440	}
5441
5442	/*
5443	 * We can only set up the journal commit callback once
5444	 * mballoc is initialized
5445	 */
5446	if (sbi->s_journal)
5447		sbi->s_journal->j_commit_callback =
5448			ext4_journal_commit_callback;
5449
5450	block = ext4_count_free_clusters(sb);
5451	ext4_free_blocks_count_set(sbi->s_es,
5452				   EXT4_C2B(sbi, block));
5453	err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
5454				  GFP_KERNEL);
5455	if (!err) {
5456		unsigned long freei = ext4_count_free_inodes(sb);
5457		sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
5458		err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
5459					  GFP_KERNEL);
5460	}
 
 
 
 
 
 
 
 
5461	if (!err)
5462		err = percpu_counter_init(&sbi->s_dirs_counter,
5463					  ext4_count_dirs(sb), GFP_KERNEL);
5464	if (!err)
5465		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
5466					  GFP_KERNEL);
5467	if (!err)
5468		err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
5469					  GFP_KERNEL);
5470	if (!err)
5471		err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
5472
5473	if (err) {
5474		ext4_msg(sb, KERN_ERR, "insufficient memory");
5475		goto failed_mount6;
5476	}
5477
5478	if (ext4_has_feature_flex_bg(sb))
5479		if (!ext4_fill_flex_info(sb)) {
5480			ext4_msg(sb, KERN_ERR,
5481			       "unable to initialize "
5482			       "flex_bg meta info!");
5483			ret = -ENOMEM;
5484			goto failed_mount6;
5485		}
5486
5487	err = ext4_register_li_request(sb, first_not_zeroed);
5488	if (err)
5489		goto failed_mount6;
5490
5491	err = ext4_register_sysfs(sb);
5492	if (err)
5493		goto failed_mount7;
5494
5495	err = ext4_init_orphan_info(sb);
5496	if (err)
5497		goto failed_mount8;
5498#ifdef CONFIG_QUOTA
5499	/* Enable quota usage during mount. */
5500	if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5501		err = ext4_enable_quotas(sb);
5502		if (err)
5503			goto failed_mount9;
5504	}
5505#endif  /* CONFIG_QUOTA */
5506
5507	/*
5508	 * Save the original bdev mapping's wb_err value which could be
5509	 * used to detect the metadata async write error.
5510	 */
5511	spin_lock_init(&sbi->s_bdev_wb_lock);
5512	errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5513				 &sbi->s_bdev_wb_err);
5514	sb->s_bdev->bd_super = sb;
5515	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5516	ext4_orphan_cleanup(sb, es);
5517	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5518	/*
5519	 * Update the checksum after updating free space/inode counters and
5520	 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5521	 * checksum in the buffer cache until it is written out and
5522	 * e2fsprogs programs trying to open a file system immediately
5523	 * after it is mounted can fail.
5524	 */
5525	ext4_superblock_csum_set(sb);
5526	if (needs_recovery) {
5527		ext4_msg(sb, KERN_INFO, "recovery complete");
5528		err = ext4_mark_recovery_complete(sb, es);
5529		if (err)
5530			goto failed_mount9;
5531	}
 
 
 
 
 
 
 
 
 
5532
5533	if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev))
5534		ext4_msg(sb, KERN_WARNING,
5535			 "mounting with \"discard\" option, but the device does not support discard");
 
 
 
 
 
 
 
 
 
 
 
 
5536
5537	if (es->s_error_count)
5538		mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5539
5540	/* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5541	ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5542	ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5543	ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5544	atomic_set(&sbi->s_warning_count, 0);
5545	atomic_set(&sbi->s_msg_count, 0);
5546
 
5547	return 0;
5548
5549failed_mount9:
5550	ext4_release_orphan_info(sb);
 
 
 
5551failed_mount8:
5552	ext4_unregister_sysfs(sb);
5553	kobject_put(&sbi->s_kobj);
5554failed_mount7:
5555	ext4_unregister_li_request(sb);
5556failed_mount6:
5557	ext4_mb_release(sb);
5558	rcu_read_lock();
5559	flex_groups = rcu_dereference(sbi->s_flex_groups);
5560	if (flex_groups) {
5561		for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5562			kvfree(flex_groups[i]);
5563		kvfree(flex_groups);
5564	}
5565	rcu_read_unlock();
5566	percpu_counter_destroy(&sbi->s_freeclusters_counter);
5567	percpu_counter_destroy(&sbi->s_freeinodes_counter);
5568	percpu_counter_destroy(&sbi->s_dirs_counter);
5569	percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5570	percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
5571	percpu_free_rwsem(&sbi->s_writepages_rwsem);
5572failed_mount5:
5573	ext4_ext_release(sb);
5574	ext4_release_system_zone(sb);
5575failed_mount4a:
5576	dput(sb->s_root);
5577	sb->s_root = NULL;
5578failed_mount4:
5579	ext4_msg(sb, KERN_ERR, "mount failed");
5580	if (EXT4_SB(sb)->rsv_conversion_wq)
5581		destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5582failed_mount_wq:
5583	ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5584	sbi->s_ea_inode_cache = NULL;
5585
5586	ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5587	sbi->s_ea_block_cache = NULL;
5588
5589	if (sbi->s_journal) {
5590		/* flush s_error_work before journal destroy. */
5591		flush_work(&sbi->s_error_work);
5592		jbd2_journal_destroy(sbi->s_journal);
5593		sbi->s_journal = NULL;
5594	}
5595failed_mount3a:
5596	ext4_es_unregister_shrinker(sbi);
5597failed_mount3:
5598	/* flush s_error_work before sbi destroy */
5599	flush_work(&sbi->s_error_work);
5600	del_timer_sync(&sbi->s_err_report);
5601	ext4_stop_mmpd(sbi);
5602	ext4_group_desc_free(sbi);
 
 
 
 
 
 
5603failed_mount:
5604	if (sbi->s_chksum_driver)
5605		crypto_free_shash(sbi->s_chksum_driver);
5606
5607#if IS_ENABLED(CONFIG_UNICODE)
5608	utf8_unload(sb->s_encoding);
5609#endif
5610
5611#ifdef CONFIG_QUOTA
5612	for (i = 0; i < EXT4_MAXQUOTAS; i++)
5613		kfree(get_qf_name(sb, sbi, i));
5614#endif
5615	fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5616	/* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
5617	brelse(sbi->s_sbh);
5618	ext4_blkdev_remove(sbi);
5619out_fail:
5620	sb->s_fs_info = NULL;
 
 
 
 
 
5621	return err ? err : ret;
5622}
5623
5624static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
5625{
5626	struct ext4_fs_context *ctx = fc->fs_private;
5627	struct ext4_sb_info *sbi;
5628	const char *descr;
5629	int ret;
5630
5631	sbi = ext4_alloc_sbi(sb);
5632	if (!sbi)
5633		return -ENOMEM;
5634
5635	fc->s_fs_info = sbi;
5636
5637	/* Cleanup superblock name */
5638	strreplace(sb->s_id, '/', '!');
5639
5640	sbi->s_sb_block = 1;	/* Default super block location */
5641	if (ctx->spec & EXT4_SPEC_s_sb_block)
5642		sbi->s_sb_block = ctx->s_sb_block;
5643
5644	ret = __ext4_fill_super(fc, sb);
5645	if (ret < 0)
5646		goto free_sbi;
5647
5648	if (sbi->s_journal) {
5649		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5650			descr = " journalled data mode";
5651		else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5652			descr = " ordered data mode";
5653		else
5654			descr = " writeback data mode";
5655	} else
5656		descr = "out journal";
5657
5658	if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5659		ext4_msg(sb, KERN_INFO, "mounted filesystem %pU with%s. "
5660			 "Quota mode: %s.", &sb->s_uuid, descr,
5661			 ext4_quota_mode(sb));
5662
5663	/* Update the s_overhead_clusters if necessary */
5664	ext4_update_overhead(sb, false);
5665	return 0;
5666
5667free_sbi:
5668	ext4_free_sbi(sbi);
5669	fc->s_fs_info = NULL;
5670	return ret;
5671}
5672
5673static int ext4_get_tree(struct fs_context *fc)
5674{
5675	return get_tree_bdev(fc, ext4_fill_super);
5676}
5677
5678/*
5679 * Setup any per-fs journal parameters now.  We'll do this both on
5680 * initial mount, once the journal has been initialised but before we've
5681 * done any recovery; and again on any subsequent remount.
5682 */
5683static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5684{
5685	struct ext4_sb_info *sbi = EXT4_SB(sb);
5686
5687	journal->j_commit_interval = sbi->s_commit_interval;
5688	journal->j_min_batch_time = sbi->s_min_batch_time;
5689	journal->j_max_batch_time = sbi->s_max_batch_time;
5690	ext4_fc_init(sb, journal);
5691
5692	write_lock(&journal->j_state_lock);
5693	if (test_opt(sb, BARRIER))
5694		journal->j_flags |= JBD2_BARRIER;
5695	else
5696		journal->j_flags &= ~JBD2_BARRIER;
5697	if (test_opt(sb, DATA_ERR_ABORT))
5698		journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5699	else
5700		journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5701	write_unlock(&journal->j_state_lock);
5702}
5703
5704static struct inode *ext4_get_journal_inode(struct super_block *sb,
5705					     unsigned int journal_inum)
5706{
5707	struct inode *journal_inode;
5708
5709	/*
5710	 * Test for the existence of a valid inode on disk.  Bad things
5711	 * happen if we iget() an unused inode, as the subsequent iput()
5712	 * will try to delete it.
5713	 */
5714	journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5715	if (IS_ERR(journal_inode)) {
5716		ext4_msg(sb, KERN_ERR, "no journal found");
5717		return NULL;
5718	}
5719	if (!journal_inode->i_nlink) {
5720		make_bad_inode(journal_inode);
5721		iput(journal_inode);
5722		ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5723		return NULL;
5724	}
5725
5726	ext4_debug("Journal inode found at %p: %lld bytes\n",
5727		  journal_inode, journal_inode->i_size);
5728	if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
5729		ext4_msg(sb, KERN_ERR, "invalid journal inode");
5730		iput(journal_inode);
5731		return NULL;
5732	}
5733	return journal_inode;
5734}
5735
5736static journal_t *ext4_get_journal(struct super_block *sb,
5737				   unsigned int journal_inum)
5738{
5739	struct inode *journal_inode;
5740	journal_t *journal;
5741
5742	if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5743		return NULL;
5744
5745	journal_inode = ext4_get_journal_inode(sb, journal_inum);
5746	if (!journal_inode)
5747		return NULL;
5748
5749	journal = jbd2_journal_init_inode(journal_inode);
5750	if (!journal) {
5751		ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5752		iput(journal_inode);
5753		return NULL;
5754	}
5755	journal->j_private = sb;
5756	ext4_init_journal_params(sb, journal);
5757	return journal;
5758}
5759
5760static journal_t *ext4_get_dev_journal(struct super_block *sb,
5761				       dev_t j_dev)
5762{
5763	struct buffer_head *bh;
5764	journal_t *journal;
5765	ext4_fsblk_t start;
5766	ext4_fsblk_t len;
5767	int hblock, blocksize;
5768	ext4_fsblk_t sb_block;
5769	unsigned long offset;
5770	struct ext4_super_block *es;
5771	struct block_device *bdev;
5772
5773	if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5774		return NULL;
5775
5776	bdev = ext4_blkdev_get(j_dev, sb);
5777	if (bdev == NULL)
5778		return NULL;
5779
5780	blocksize = sb->s_blocksize;
5781	hblock = bdev_logical_block_size(bdev);
5782	if (blocksize < hblock) {
5783		ext4_msg(sb, KERN_ERR,
5784			"blocksize too small for journal device");
5785		goto out_bdev;
5786	}
5787
5788	sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5789	offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5790	set_blocksize(bdev, blocksize);
5791	if (!(bh = __bread(bdev, sb_block, blocksize))) {
5792		ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5793		       "external journal");
5794		goto out_bdev;
5795	}
5796
5797	es = (struct ext4_super_block *) (bh->b_data + offset);
5798	if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5799	    !(le32_to_cpu(es->s_feature_incompat) &
5800	      EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5801		ext4_msg(sb, KERN_ERR, "external journal has "
5802					"bad superblock");
5803		brelse(bh);
5804		goto out_bdev;
5805	}
5806
5807	if ((le32_to_cpu(es->s_feature_ro_compat) &
5808	     EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5809	    es->s_checksum != ext4_superblock_csum(sb, es)) {
5810		ext4_msg(sb, KERN_ERR, "external journal has "
5811				       "corrupt superblock");
5812		brelse(bh);
5813		goto out_bdev;
5814	}
5815
5816	if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5817		ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5818		brelse(bh);
5819		goto out_bdev;
5820	}
5821
5822	len = ext4_blocks_count(es);
5823	start = sb_block + 1;
5824	brelse(bh);	/* we're done with the superblock */
5825
5826	journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5827					start, len, blocksize);
5828	if (!journal) {
5829		ext4_msg(sb, KERN_ERR, "failed to create device journal");
5830		goto out_bdev;
5831	}
5832	journal->j_private = sb;
5833	if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5834		ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5835		goto out_journal;
5836	}
5837	if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5838		ext4_msg(sb, KERN_ERR, "External journal has more than one "
5839					"user (unsupported) - %d",
5840			be32_to_cpu(journal->j_superblock->s_nr_users));
5841		goto out_journal;
5842	}
5843	EXT4_SB(sb)->s_journal_bdev = bdev;
5844	ext4_init_journal_params(sb, journal);
5845	return journal;
5846
5847out_journal:
5848	jbd2_journal_destroy(journal);
5849out_bdev:
5850	ext4_blkdev_put(bdev);
5851	return NULL;
5852}
5853
5854static int ext4_load_journal(struct super_block *sb,
5855			     struct ext4_super_block *es,
5856			     unsigned long journal_devnum)
5857{
5858	journal_t *journal;
5859	unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5860	dev_t journal_dev;
5861	int err = 0;
5862	int really_read_only;
5863	int journal_dev_ro;
5864
5865	if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5866		return -EFSCORRUPTED;
5867
5868	if (journal_devnum &&
5869	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5870		ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5871			"numbers have changed");
5872		journal_dev = new_decode_dev(journal_devnum);
5873	} else
5874		journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5875
5876	if (journal_inum && journal_dev) {
5877		ext4_msg(sb, KERN_ERR,
5878			 "filesystem has both journal inode and journal device!");
5879		return -EINVAL;
5880	}
5881
5882	if (journal_inum) {
5883		journal = ext4_get_journal(sb, journal_inum);
5884		if (!journal)
5885			return -EINVAL;
5886	} else {
5887		journal = ext4_get_dev_journal(sb, journal_dev);
5888		if (!journal)
5889			return -EINVAL;
5890	}
5891
5892	journal_dev_ro = bdev_read_only(journal->j_dev);
5893	really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5894
5895	if (journal_dev_ro && !sb_rdonly(sb)) {
5896		ext4_msg(sb, KERN_ERR,
5897			 "journal device read-only, try mounting with '-o ro'");
5898		err = -EROFS;
5899		goto err_out;
5900	}
5901
5902	/*
5903	 * Are we loading a blank journal or performing recovery after a
5904	 * crash?  For recovery, we need to check in advance whether we
5905	 * can get read-write access to the device.
5906	 */
5907	if (ext4_has_feature_journal_needs_recovery(sb)) {
5908		if (sb_rdonly(sb)) {
5909			ext4_msg(sb, KERN_INFO, "INFO: recovery "
5910					"required on readonly filesystem");
5911			if (really_read_only) {
5912				ext4_msg(sb, KERN_ERR, "write access "
5913					"unavailable, cannot proceed "
5914					"(try mounting with noload)");
5915				err = -EROFS;
5916				goto err_out;
5917			}
5918			ext4_msg(sb, KERN_INFO, "write access will "
5919			       "be enabled during recovery");
5920		}
5921	}
5922
5923	if (!(journal->j_flags & JBD2_BARRIER))
5924		ext4_msg(sb, KERN_INFO, "barriers disabled");
5925
5926	if (!ext4_has_feature_journal_needs_recovery(sb))
5927		err = jbd2_journal_wipe(journal, !really_read_only);
5928	if (!err) {
5929		char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5930		if (save)
5931			memcpy(save, ((char *) es) +
5932			       EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5933		err = jbd2_journal_load(journal);
5934		if (save)
5935			memcpy(((char *) es) + EXT4_S_ERR_START,
5936			       save, EXT4_S_ERR_LEN);
5937		kfree(save);
5938	}
5939
5940	if (err) {
5941		ext4_msg(sb, KERN_ERR, "error loading journal");
5942		goto err_out;
5943	}
5944
5945	EXT4_SB(sb)->s_journal = journal;
5946	err = ext4_clear_journal_err(sb, es);
5947	if (err) {
5948		EXT4_SB(sb)->s_journal = NULL;
5949		jbd2_journal_destroy(journal);
5950		return err;
5951	}
5952
5953	if (!really_read_only && journal_devnum &&
5954	    journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5955		es->s_journal_dev = cpu_to_le32(journal_devnum);
5956
5957		/* Make sure we flush the recovery flag to disk. */
5958		ext4_commit_super(sb);
5959	}
5960
5961	return 0;
5962
5963err_out:
5964	jbd2_journal_destroy(journal);
5965	return err;
5966}
5967
5968/* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5969static void ext4_update_super(struct super_block *sb)
5970{
5971	struct ext4_sb_info *sbi = EXT4_SB(sb);
5972	struct ext4_super_block *es = sbi->s_es;
5973	struct buffer_head *sbh = sbi->s_sbh;
5974
5975	lock_buffer(sbh);
5976	/*
5977	 * If the file system is mounted read-only, don't update the
5978	 * superblock write time.  This avoids updating the superblock
5979	 * write time when we are mounting the root file system
5980	 * read/only but we need to replay the journal; at that point,
5981	 * for people who are east of GMT and who make their clock
5982	 * tick in localtime for Windows bug-for-bug compatibility,
5983	 * the clock is set in the future, and this will cause e2fsck
5984	 * to complain and force a full file system check.
5985	 */
5986	if (!(sb->s_flags & SB_RDONLY))
5987		ext4_update_tstamp(es, s_wtime);
5988	es->s_kbytes_written =
5989		cpu_to_le64(sbi->s_kbytes_written +
5990		    ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
5991		      sbi->s_sectors_written_start) >> 1));
5992	if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
5993		ext4_free_blocks_count_set(es,
5994			EXT4_C2B(sbi, percpu_counter_sum_positive(
5995				&sbi->s_freeclusters_counter)));
5996	if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
5997		es->s_free_inodes_count =
5998			cpu_to_le32(percpu_counter_sum_positive(
5999				&sbi->s_freeinodes_counter));
6000	/* Copy error information to the on-disk superblock */
6001	spin_lock(&sbi->s_error_lock);
6002	if (sbi->s_add_error_count > 0) {
6003		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6004		if (!es->s_first_error_time && !es->s_first_error_time_hi) {
6005			__ext4_update_tstamp(&es->s_first_error_time,
6006					     &es->s_first_error_time_hi,
6007					     sbi->s_first_error_time);
6008			strncpy(es->s_first_error_func, sbi->s_first_error_func,
6009				sizeof(es->s_first_error_func));
6010			es->s_first_error_line =
6011				cpu_to_le32(sbi->s_first_error_line);
6012			es->s_first_error_ino =
6013				cpu_to_le32(sbi->s_first_error_ino);
6014			es->s_first_error_block =
6015				cpu_to_le64(sbi->s_first_error_block);
6016			es->s_first_error_errcode =
6017				ext4_errno_to_code(sbi->s_first_error_code);
6018		}
6019		__ext4_update_tstamp(&es->s_last_error_time,
6020				     &es->s_last_error_time_hi,
6021				     sbi->s_last_error_time);
6022		strncpy(es->s_last_error_func, sbi->s_last_error_func,
6023			sizeof(es->s_last_error_func));
6024		es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
6025		es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
6026		es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
6027		es->s_last_error_errcode =
6028				ext4_errno_to_code(sbi->s_last_error_code);
6029		/*
6030		 * Start the daily error reporting function if it hasn't been
6031		 * started already
6032		 */
6033		if (!es->s_error_count)
6034			mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
6035		le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
6036		sbi->s_add_error_count = 0;
6037	}
6038	spin_unlock(&sbi->s_error_lock);
6039
6040	ext4_superblock_csum_set(sb);
6041	unlock_buffer(sbh);
6042}
6043
6044static int ext4_commit_super(struct super_block *sb)
6045{
6046	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
 
6047
6048	if (!sbh)
6049		return -EINVAL;
6050	if (block_device_ejected(sb))
6051		return -ENODEV;
6052
6053	ext4_update_super(sb);
6054
6055	lock_buffer(sbh);
6056	/* Buffer got discarded which means block device got invalidated */
6057	if (!buffer_mapped(sbh)) {
6058		unlock_buffer(sbh);
6059		return -EIO;
6060	}
6061
6062	if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
6063		/*
6064		 * Oh, dear.  A previous attempt to write the
6065		 * superblock failed.  This could happen because the
6066		 * USB device was yanked out.  Or it could happen to
6067		 * be a transient write error and maybe the block will
6068		 * be remapped.  Nothing we can do but to retry the
6069		 * write and hope for the best.
6070		 */
6071		ext4_msg(sb, KERN_ERR, "previous I/O error to "
6072		       "superblock detected");
6073		clear_buffer_write_io_error(sbh);
6074		set_buffer_uptodate(sbh);
6075	}
6076	get_bh(sbh);
6077	/* Clear potential dirty bit if it was journalled update */
6078	clear_buffer_dirty(sbh);
6079	sbh->b_end_io = end_buffer_write_sync;
6080	submit_bh(REQ_OP_WRITE | REQ_SYNC |
6081		  (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
6082	wait_on_buffer(sbh);
6083	if (buffer_write_io_error(sbh)) {
6084		ext4_msg(sb, KERN_ERR, "I/O error while writing "
6085		       "superblock");
6086		clear_buffer_write_io_error(sbh);
6087		set_buffer_uptodate(sbh);
6088		return -EIO;
6089	}
6090	return 0;
6091}
6092
6093/*
6094 * Have we just finished recovery?  If so, and if we are mounting (or
6095 * remounting) the filesystem readonly, then we will end up with a
6096 * consistent fs on disk.  Record that fact.
6097 */
6098static int ext4_mark_recovery_complete(struct super_block *sb,
6099				       struct ext4_super_block *es)
6100{
6101	int err;
6102	journal_t *journal = EXT4_SB(sb)->s_journal;
6103
6104	if (!ext4_has_feature_journal(sb)) {
6105		if (journal != NULL) {
6106			ext4_error(sb, "Journal got removed while the fs was "
6107				   "mounted!");
6108			return -EFSCORRUPTED;
6109		}
6110		return 0;
6111	}
6112	jbd2_journal_lock_updates(journal);
6113	err = jbd2_journal_flush(journal, 0);
6114	if (err < 0)
6115		goto out;
6116
6117	if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6118	    ext4_has_feature_orphan_present(sb))) {
6119		if (!ext4_orphan_file_empty(sb)) {
6120			ext4_error(sb, "Orphan file not empty on read-only fs.");
6121			err = -EFSCORRUPTED;
6122			goto out;
6123		}
6124		ext4_clear_feature_journal_needs_recovery(sb);
6125		ext4_clear_feature_orphan_present(sb);
6126		ext4_commit_super(sb);
6127	}
6128out:
6129	jbd2_journal_unlock_updates(journal);
6130	return err;
6131}
6132
6133/*
6134 * If we are mounting (or read-write remounting) a filesystem whose journal
6135 * has recorded an error from a previous lifetime, move that error to the
6136 * main filesystem now.
6137 */
6138static int ext4_clear_journal_err(struct super_block *sb,
6139				   struct ext4_super_block *es)
6140{
6141	journal_t *journal;
6142	int j_errno;
6143	const char *errstr;
6144
6145	if (!ext4_has_feature_journal(sb)) {
6146		ext4_error(sb, "Journal got removed while the fs was mounted!");
6147		return -EFSCORRUPTED;
6148	}
6149
6150	journal = EXT4_SB(sb)->s_journal;
6151
6152	/*
6153	 * Now check for any error status which may have been recorded in the
6154	 * journal by a prior ext4_error() or ext4_abort()
6155	 */
6156
6157	j_errno = jbd2_journal_errno(journal);
6158	if (j_errno) {
6159		char nbuf[16];
6160
6161		errstr = ext4_decode_error(sb, j_errno, nbuf);
6162		ext4_warning(sb, "Filesystem error recorded "
6163			     "from previous mount: %s", errstr);
6164		ext4_warning(sb, "Marking fs in need of filesystem check.");
6165
6166		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6167		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6168		ext4_commit_super(sb);
6169
6170		jbd2_journal_clear_err(journal);
6171		jbd2_journal_update_sb_errno(journal);
6172	}
6173	return 0;
6174}
6175
6176/*
6177 * Force the running and committing transactions to commit,
6178 * and wait on the commit.
6179 */
6180int ext4_force_commit(struct super_block *sb)
6181{
6182	journal_t *journal;
6183
6184	if (sb_rdonly(sb))
6185		return 0;
6186
6187	journal = EXT4_SB(sb)->s_journal;
6188	return ext4_journal_force_commit(journal);
6189}
6190
6191static int ext4_sync_fs(struct super_block *sb, int wait)
6192{
6193	int ret = 0;
6194	tid_t target;
6195	bool needs_barrier = false;
6196	struct ext4_sb_info *sbi = EXT4_SB(sb);
6197
6198	if (unlikely(ext4_forced_shutdown(sbi)))
6199		return 0;
6200
6201	trace_ext4_sync_fs(sb, wait);
6202	flush_workqueue(sbi->rsv_conversion_wq);
6203	/*
6204	 * Writeback quota in non-journalled quota case - journalled quota has
6205	 * no dirty dquots
6206	 */
6207	dquot_writeback_dquots(sb, -1);
6208	/*
6209	 * Data writeback is possible w/o journal transaction, so barrier must
6210	 * being sent at the end of the function. But we can skip it if
6211	 * transaction_commit will do it for us.
6212	 */
6213	if (sbi->s_journal) {
6214		target = jbd2_get_latest_transaction(sbi->s_journal);
6215		if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6216		    !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6217			needs_barrier = true;
6218
6219		if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6220			if (wait)
6221				ret = jbd2_log_wait_commit(sbi->s_journal,
6222							   target);
6223		}
6224	} else if (wait && test_opt(sb, BARRIER))
6225		needs_barrier = true;
6226	if (needs_barrier) {
6227		int err;
6228		err = blkdev_issue_flush(sb->s_bdev);
6229		if (!ret)
6230			ret = err;
6231	}
6232
6233	return ret;
6234}
6235
6236/*
6237 * LVM calls this function before a (read-only) snapshot is created.  This
6238 * gives us a chance to flush the journal completely and mark the fs clean.
6239 *
6240 * Note that only this function cannot bring a filesystem to be in a clean
6241 * state independently. It relies on upper layer to stop all data & metadata
6242 * modifications.
6243 */
6244static int ext4_freeze(struct super_block *sb)
6245{
6246	int error = 0;
6247	journal_t *journal;
6248
6249	if (sb_rdonly(sb))
6250		return 0;
6251
6252	journal = EXT4_SB(sb)->s_journal;
6253
6254	if (journal) {
6255		/* Now we set up the journal barrier. */
6256		jbd2_journal_lock_updates(journal);
6257
6258		/*
6259		 * Don't clear the needs_recovery flag if we failed to
6260		 * flush the journal.
6261		 */
6262		error = jbd2_journal_flush(journal, 0);
6263		if (error < 0)
6264			goto out;
6265
6266		/* Journal blocked and flushed, clear needs_recovery flag. */
6267		ext4_clear_feature_journal_needs_recovery(sb);
6268		if (ext4_orphan_file_empty(sb))
6269			ext4_clear_feature_orphan_present(sb);
6270	}
6271
6272	error = ext4_commit_super(sb);
6273out:
6274	if (journal)
6275		/* we rely on upper layer to stop further updates */
6276		jbd2_journal_unlock_updates(journal);
6277	return error;
6278}
6279
6280/*
6281 * Called by LVM after the snapshot is done.  We need to reset the RECOVER
6282 * flag here, even though the filesystem is not technically dirty yet.
6283 */
6284static int ext4_unfreeze(struct super_block *sb)
6285{
6286	if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
6287		return 0;
6288
6289	if (EXT4_SB(sb)->s_journal) {
6290		/* Reset the needs_recovery flag before the fs is unlocked. */
6291		ext4_set_feature_journal_needs_recovery(sb);
6292		if (ext4_has_feature_orphan_file(sb))
6293			ext4_set_feature_orphan_present(sb);
6294	}
6295
6296	ext4_commit_super(sb);
6297	return 0;
6298}
6299
6300/*
6301 * Structure to save mount options for ext4_remount's benefit
6302 */
6303struct ext4_mount_options {
6304	unsigned long s_mount_opt;
6305	unsigned long s_mount_opt2;
6306	kuid_t s_resuid;
6307	kgid_t s_resgid;
6308	unsigned long s_commit_interval;
6309	u32 s_min_batch_time, s_max_batch_time;
6310#ifdef CONFIG_QUOTA
6311	int s_jquota_fmt;
6312	char *s_qf_names[EXT4_MAXQUOTAS];
6313#endif
6314};
6315
6316static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
6317{
6318	struct ext4_fs_context *ctx = fc->fs_private;
6319	struct ext4_super_block *es;
6320	struct ext4_sb_info *sbi = EXT4_SB(sb);
6321	unsigned long old_sb_flags;
6322	struct ext4_mount_options old_opts;
 
6323	ext4_group_t g;
6324	int err = 0;
6325#ifdef CONFIG_QUOTA
6326	int enable_quota = 0;
6327	int i, j;
6328	char *to_free[EXT4_MAXQUOTAS];
6329#endif
 
 
6330
 
 
 
 
 
6331
6332	/* Store the original options */
6333	old_sb_flags = sb->s_flags;
6334	old_opts.s_mount_opt = sbi->s_mount_opt;
6335	old_opts.s_mount_opt2 = sbi->s_mount_opt2;
6336	old_opts.s_resuid = sbi->s_resuid;
6337	old_opts.s_resgid = sbi->s_resgid;
6338	old_opts.s_commit_interval = sbi->s_commit_interval;
6339	old_opts.s_min_batch_time = sbi->s_min_batch_time;
6340	old_opts.s_max_batch_time = sbi->s_max_batch_time;
6341#ifdef CONFIG_QUOTA
6342	old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
6343	for (i = 0; i < EXT4_MAXQUOTAS; i++)
6344		if (sbi->s_qf_names[i]) {
6345			char *qf_name = get_qf_name(sb, sbi, i);
6346
6347			old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
6348			if (!old_opts.s_qf_names[i]) {
6349				for (j = 0; j < i; j++)
6350					kfree(old_opts.s_qf_names[j]);
 
6351				return -ENOMEM;
6352			}
6353		} else
6354			old_opts.s_qf_names[i] = NULL;
6355#endif
6356	if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6357		if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6358			ctx->journal_ioprio =
6359				sbi->s_journal->j_task->io_context->ioprio;
6360		else
6361			ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6362
6363	}
 
 
 
 
 
 
6364
6365	ext4_apply_options(fc, sb);
 
 
 
6366
6367	if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
6368	    test_opt(sb, JOURNAL_CHECKSUM)) {
6369		ext4_msg(sb, KERN_ERR, "changing journal_checksum "
6370			 "during remount not supported; ignoring");
6371		sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6372	}
6373
6374	if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6375		if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6376			ext4_msg(sb, KERN_ERR, "can't mount with "
6377				 "both data=journal and delalloc");
6378			err = -EINVAL;
6379			goto restore_opts;
6380		}
6381		if (test_opt(sb, DIOREAD_NOLOCK)) {
6382			ext4_msg(sb, KERN_ERR, "can't mount with "
6383				 "both data=journal and dioread_nolock");
6384			err = -EINVAL;
6385			goto restore_opts;
6386		}
6387	} else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6388		if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6389			ext4_msg(sb, KERN_ERR, "can't mount with "
6390				"journal_async_commit in data=ordered mode");
6391			err = -EINVAL;
6392			goto restore_opts;
6393		}
6394	}
6395
6396	if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6397		ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6398		err = -EINVAL;
6399		goto restore_opts;
6400	}
6401
6402	if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
6403		ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
6404
6405	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6406		(test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
6407
6408	es = sbi->s_es;
6409
6410	if (sbi->s_journal) {
6411		ext4_init_journal_params(sb, sbi->s_journal);
6412		set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
6413	}
6414
6415	/* Flush outstanding errors before changing fs state */
6416	flush_work(&sbi->s_error_work);
6417
6418	if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
6419		if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
6420			err = -EROFS;
6421			goto restore_opts;
6422		}
6423
6424		if (fc->sb_flags & SB_RDONLY) {
6425			err = sync_filesystem(sb);
6426			if (err < 0)
6427				goto restore_opts;
6428			err = dquot_suspend(sb, -1);
6429			if (err < 0)
6430				goto restore_opts;
6431
6432			/*
6433			 * First of all, the unconditional stuff we have to do
6434			 * to disable replay of the journal when we next remount
6435			 */
6436			sb->s_flags |= SB_RDONLY;
6437
6438			/*
6439			 * OK, test if we are remounting a valid rw partition
6440			 * readonly, and if so set the rdonly flag and then
6441			 * mark the partition as valid again.
6442			 */
6443			if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6444			    (sbi->s_mount_state & EXT4_VALID_FS))
6445				es->s_state = cpu_to_le16(sbi->s_mount_state);
6446
6447			if (sbi->s_journal) {
6448				/*
6449				 * We let remount-ro finish even if marking fs
6450				 * as clean failed...
6451				 */
6452				ext4_mark_recovery_complete(sb, es);
6453			}
6454		} else {
6455			/* Make sure we can mount this feature set readwrite */
6456			if (ext4_has_feature_readonly(sb) ||
6457			    !ext4_feature_set_ok(sb, 0)) {
6458				err = -EROFS;
6459				goto restore_opts;
6460			}
6461			/*
6462			 * Make sure the group descriptor checksums
6463			 * are sane.  If they aren't, refuse to remount r/w.
6464			 */
6465			for (g = 0; g < sbi->s_groups_count; g++) {
6466				struct ext4_group_desc *gdp =
6467					ext4_get_group_desc(sb, g, NULL);
6468
6469				if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
6470					ext4_msg(sb, KERN_ERR,
6471	       "ext4_remount: Checksum for group %u failed (%u!=%u)",
6472		g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6473					       le16_to_cpu(gdp->bg_checksum));
6474					err = -EFSBADCRC;
6475					goto restore_opts;
6476				}
6477			}
6478
6479			/*
6480			 * If we have an unprocessed orphan list hanging
6481			 * around from a previously readonly bdev mount,
6482			 * require a full umount/remount for now.
6483			 */
6484			if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
6485				ext4_msg(sb, KERN_WARNING, "Couldn't "
6486				       "remount RDWR because of unprocessed "
6487				       "orphan inode list.  Please "
6488				       "umount/remount instead");
6489				err = -EINVAL;
6490				goto restore_opts;
6491			}
6492
6493			/*
6494			 * Mounting a RDONLY partition read-write, so reread
6495			 * and store the current valid flag.  (It may have
6496			 * been changed by e2fsck since we originally mounted
6497			 * the partition.)
6498			 */
6499			if (sbi->s_journal) {
6500				err = ext4_clear_journal_err(sb, es);
6501				if (err)
6502					goto restore_opts;
6503			}
6504			sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6505					      ~EXT4_FC_REPLAY);
6506
6507			err = ext4_setup_super(sb, es, 0);
6508			if (err)
6509				goto restore_opts;
6510
6511			sb->s_flags &= ~SB_RDONLY;
6512			if (ext4_has_feature_mmp(sb))
6513				if (ext4_multi_mount_protect(sb,
6514						le64_to_cpu(es->s_mmp_block))) {
6515					err = -EROFS;
6516					goto restore_opts;
6517				}
6518#ifdef CONFIG_QUOTA
6519			enable_quota = 1;
6520#endif
6521		}
6522	}
6523
6524	/*
6525	 * Reinitialize lazy itable initialization thread based on
6526	 * current settings
6527	 */
6528	if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6529		ext4_unregister_li_request(sb);
6530	else {
6531		ext4_group_t first_not_zeroed;
6532		first_not_zeroed = ext4_has_uninit_itable(sb);
6533		ext4_register_li_request(sb, first_not_zeroed);
6534	}
6535
6536	/*
6537	 * Handle creation of system zone data early because it can fail.
6538	 * Releasing of existing data is done when we are sure remount will
6539	 * succeed.
6540	 */
6541	if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6542		err = ext4_setup_system_zone(sb);
6543		if (err)
6544			goto restore_opts;
6545	}
6546
6547	if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6548		err = ext4_commit_super(sb);
6549		if (err)
6550			goto restore_opts;
6551	}
6552
6553#ifdef CONFIG_QUOTA
6554	/* Release old quota file names */
6555	for (i = 0; i < EXT4_MAXQUOTAS; i++)
6556		kfree(old_opts.s_qf_names[i]);
6557	if (enable_quota) {
6558		if (sb_any_quota_suspended(sb))
6559			dquot_resume(sb, -1);
6560		else if (ext4_has_feature_quota(sb)) {
6561			err = ext4_enable_quotas(sb);
6562			if (err)
6563				goto restore_opts;
6564		}
6565	}
6566#endif
6567	if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6568		ext4_release_system_zone(sb);
6569
6570	if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6571		ext4_stop_mmpd(sbi);
6572
 
 
 
 
 
 
 
 
 
 
6573	return 0;
6574
6575restore_opts:
6576	sb->s_flags = old_sb_flags;
6577	sbi->s_mount_opt = old_opts.s_mount_opt;
6578	sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6579	sbi->s_resuid = old_opts.s_resuid;
6580	sbi->s_resgid = old_opts.s_resgid;
6581	sbi->s_commit_interval = old_opts.s_commit_interval;
6582	sbi->s_min_batch_time = old_opts.s_min_batch_time;
6583	sbi->s_max_batch_time = old_opts.s_max_batch_time;
6584	if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6585		ext4_release_system_zone(sb);
6586#ifdef CONFIG_QUOTA
6587	sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6588	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6589		to_free[i] = get_qf_name(sb, sbi, i);
6590		rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6591	}
6592	synchronize_rcu();
6593	for (i = 0; i < EXT4_MAXQUOTAS; i++)
6594		kfree(to_free[i]);
6595#endif
6596	if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6597		ext4_stop_mmpd(sbi);
 
6598	return err;
6599}
6600
6601static int ext4_reconfigure(struct fs_context *fc)
6602{
6603	struct super_block *sb = fc->root->d_sb;
6604	int ret;
6605
6606	fc->s_fs_info = EXT4_SB(sb);
6607
6608	ret = ext4_check_opt_consistency(fc, sb);
6609	if (ret < 0)
6610		return ret;
6611
6612	ret = __ext4_remount(fc, sb);
6613	if (ret < 0)
6614		return ret;
6615
6616	ext4_msg(sb, KERN_INFO, "re-mounted %pU. Quota mode: %s.",
6617		 &sb->s_uuid, ext4_quota_mode(sb));
6618
6619	return 0;
6620}
6621
6622#ifdef CONFIG_QUOTA
6623static int ext4_statfs_project(struct super_block *sb,
6624			       kprojid_t projid, struct kstatfs *buf)
6625{
6626	struct kqid qid;
6627	struct dquot *dquot;
6628	u64 limit;
6629	u64 curblock;
6630
6631	qid = make_kqid_projid(projid);
6632	dquot = dqget(sb, qid);
6633	if (IS_ERR(dquot))
6634		return PTR_ERR(dquot);
6635	spin_lock(&dquot->dq_dqb_lock);
6636
6637	limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6638			     dquot->dq_dqb.dqb_bhardlimit);
6639	limit >>= sb->s_blocksize_bits;
6640
6641	if (limit && buf->f_blocks > limit) {
6642		curblock = (dquot->dq_dqb.dqb_curspace +
6643			    dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6644		buf->f_blocks = limit;
6645		buf->f_bfree = buf->f_bavail =
6646			(buf->f_blocks > curblock) ?
6647			 (buf->f_blocks - curblock) : 0;
6648	}
6649
6650	limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6651			     dquot->dq_dqb.dqb_ihardlimit);
6652	if (limit && buf->f_files > limit) {
6653		buf->f_files = limit;
6654		buf->f_ffree =
6655			(buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6656			 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6657	}
6658
6659	spin_unlock(&dquot->dq_dqb_lock);
6660	dqput(dquot);
6661	return 0;
6662}
6663#endif
6664
6665static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6666{
6667	struct super_block *sb = dentry->d_sb;
6668	struct ext4_sb_info *sbi = EXT4_SB(sb);
6669	struct ext4_super_block *es = sbi->s_es;
6670	ext4_fsblk_t overhead = 0, resv_blocks;
6671	s64 bfree;
6672	resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6673
6674	if (!test_opt(sb, MINIX_DF))
6675		overhead = sbi->s_overhead;
6676
6677	buf->f_type = EXT4_SUPER_MAGIC;
6678	buf->f_bsize = sb->s_blocksize;
6679	buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6680	bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6681		percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6682	/* prevent underflow in case that few free space is available */
6683	buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6684	buf->f_bavail = buf->f_bfree -
6685			(ext4_r_blocks_count(es) + resv_blocks);
6686	if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6687		buf->f_bavail = 0;
6688	buf->f_files = le32_to_cpu(es->s_inodes_count);
6689	buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6690	buf->f_namelen = EXT4_NAME_LEN;
6691	buf->f_fsid = uuid_to_fsid(es->s_uuid);
6692
6693#ifdef CONFIG_QUOTA
6694	if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6695	    sb_has_quota_limits_enabled(sb, PRJQUOTA))
6696		ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6697#endif
6698	return 0;
6699}
6700
6701
6702#ifdef CONFIG_QUOTA
6703
6704/*
6705 * Helper functions so that transaction is started before we acquire dqio_sem
6706 * to keep correct lock ordering of transaction > dqio_sem
6707 */
6708static inline struct inode *dquot_to_inode(struct dquot *dquot)
6709{
6710	return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6711}
6712
6713static int ext4_write_dquot(struct dquot *dquot)
6714{
6715	int ret, err;
6716	handle_t *handle;
6717	struct inode *inode;
6718
6719	inode = dquot_to_inode(dquot);
6720	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6721				    EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6722	if (IS_ERR(handle))
6723		return PTR_ERR(handle);
6724	ret = dquot_commit(dquot);
6725	err = ext4_journal_stop(handle);
6726	if (!ret)
6727		ret = err;
6728	return ret;
6729}
6730
6731static int ext4_acquire_dquot(struct dquot *dquot)
6732{
6733	int ret, err;
6734	handle_t *handle;
6735
6736	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6737				    EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6738	if (IS_ERR(handle))
6739		return PTR_ERR(handle);
6740	ret = dquot_acquire(dquot);
6741	err = ext4_journal_stop(handle);
6742	if (!ret)
6743		ret = err;
6744	return ret;
6745}
6746
6747static int ext4_release_dquot(struct dquot *dquot)
6748{
6749	int ret, err;
6750	handle_t *handle;
6751
6752	handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6753				    EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6754	if (IS_ERR(handle)) {
6755		/* Release dquot anyway to avoid endless cycle in dqput() */
6756		dquot_release(dquot);
6757		return PTR_ERR(handle);
6758	}
6759	ret = dquot_release(dquot);
6760	err = ext4_journal_stop(handle);
6761	if (!ret)
6762		ret = err;
6763	return ret;
6764}
6765
6766static int ext4_mark_dquot_dirty(struct dquot *dquot)
6767{
6768	struct super_block *sb = dquot->dq_sb;
6769
6770	if (ext4_is_quota_journalled(sb)) {
6771		dquot_mark_dquot_dirty(dquot);
6772		return ext4_write_dquot(dquot);
6773	} else {
6774		return dquot_mark_dquot_dirty(dquot);
6775	}
6776}
6777
6778static int ext4_write_info(struct super_block *sb, int type)
6779{
6780	int ret, err;
6781	handle_t *handle;
6782
6783	/* Data block + inode block */
6784	handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
6785	if (IS_ERR(handle))
6786		return PTR_ERR(handle);
6787	ret = dquot_commit_info(sb, type);
6788	err = ext4_journal_stop(handle);
6789	if (!ret)
6790		ret = err;
6791	return ret;
6792}
6793
 
 
 
 
 
 
 
 
 
 
6794static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6795{
6796	struct ext4_inode_info *ei = EXT4_I(inode);
6797
6798	/* The first argument of lockdep_set_subclass has to be
6799	 * *exactly* the same as the argument to init_rwsem() --- in
6800	 * this case, in init_once() --- or lockdep gets unhappy
6801	 * because the name of the lock is set using the
6802	 * stringification of the argument to init_rwsem().
6803	 */
6804	(void) ei;	/* shut up clang warning if !CONFIG_LOCKDEP */
6805	lockdep_set_subclass(&ei->i_data_sem, subclass);
6806}
6807
6808/*
6809 * Standard function to be called on quota_on
6810 */
6811static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6812			 const struct path *path)
6813{
6814	int err;
6815
6816	if (!test_opt(sb, QUOTA))
6817		return -EINVAL;
6818
6819	/* Quotafile not on the same filesystem? */
6820	if (path->dentry->d_sb != sb)
6821		return -EXDEV;
6822
6823	/* Quota already enabled for this file? */
6824	if (IS_NOQUOTA(d_inode(path->dentry)))
6825		return -EBUSY;
6826
6827	/* Journaling quota? */
6828	if (EXT4_SB(sb)->s_qf_names[type]) {
6829		/* Quotafile not in fs root? */
6830		if (path->dentry->d_parent != sb->s_root)
6831			ext4_msg(sb, KERN_WARNING,
6832				"Quota file not on filesystem root. "
6833				"Journaled quota will not work");
6834		sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6835	} else {
6836		/*
6837		 * Clear the flag just in case mount options changed since
6838		 * last time.
6839		 */
6840		sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6841	}
6842
6843	/*
6844	 * When we journal data on quota file, we have to flush journal to see
6845	 * all updates to the file when we bypass pagecache...
6846	 */
6847	if (EXT4_SB(sb)->s_journal &&
6848	    ext4_should_journal_data(d_inode(path->dentry))) {
6849		/*
6850		 * We don't need to lock updates but journal_flush() could
6851		 * otherwise be livelocked...
6852		 */
6853		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6854		err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
6855		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6856		if (err)
6857			return err;
6858	}
6859
6860	lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6861	err = dquot_quota_on(sb, type, format_id, path);
6862	if (!err) {
 
 
 
6863		struct inode *inode = d_inode(path->dentry);
6864		handle_t *handle;
6865
6866		/*
6867		 * Set inode flags to prevent userspace from messing with quota
6868		 * files. If this fails, we return success anyway since quotas
6869		 * are already enabled and this is not a hard failure.
6870		 */
6871		inode_lock(inode);
6872		handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6873		if (IS_ERR(handle))
6874			goto unlock_inode;
6875		EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6876		inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6877				S_NOATIME | S_IMMUTABLE);
6878		err = ext4_mark_inode_dirty(handle, inode);
6879		ext4_journal_stop(handle);
6880	unlock_inode:
6881		inode_unlock(inode);
6882		if (err)
6883			dquot_quota_off(sb, type);
6884	}
6885	if (err)
6886		lockdep_set_quota_inode(path->dentry->d_inode,
6887					     I_DATA_SEM_NORMAL);
6888	return err;
6889}
6890
6891static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
6892{
6893	switch (type) {
6894	case USRQUOTA:
6895		return qf_inum == EXT4_USR_QUOTA_INO;
6896	case GRPQUOTA:
6897		return qf_inum == EXT4_GRP_QUOTA_INO;
6898	case PRJQUOTA:
6899		return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
6900	default:
6901		BUG();
6902	}
6903}
6904
6905static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6906			     unsigned int flags)
6907{
6908	int err;
6909	struct inode *qf_inode;
6910	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6911		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6912		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6913		le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6914	};
6915
6916	BUG_ON(!ext4_has_feature_quota(sb));
6917
6918	if (!qf_inums[type])
6919		return -EPERM;
6920
6921	if (!ext4_check_quota_inum(type, qf_inums[type])) {
6922		ext4_error(sb, "Bad quota inum: %lu, type: %d",
6923				qf_inums[type], type);
6924		return -EUCLEAN;
6925	}
6926
6927	qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6928	if (IS_ERR(qf_inode)) {
6929		ext4_error(sb, "Bad quota inode: %lu, type: %d",
6930				qf_inums[type], type);
6931		return PTR_ERR(qf_inode);
6932	}
6933
6934	/* Don't account quota for quota files to avoid recursion */
6935	qf_inode->i_flags |= S_NOQUOTA;
6936	lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6937	err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6938	if (err)
6939		lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6940	iput(qf_inode);
6941
6942	return err;
6943}
6944
6945/* Enable usage tracking for all quota types. */
6946int ext4_enable_quotas(struct super_block *sb)
6947{
6948	int type, err = 0;
6949	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6950		le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6951		le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6952		le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6953	};
6954	bool quota_mopt[EXT4_MAXQUOTAS] = {
6955		test_opt(sb, USRQUOTA),
6956		test_opt(sb, GRPQUOTA),
6957		test_opt(sb, PRJQUOTA),
6958	};
6959
6960	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6961	for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6962		if (qf_inums[type]) {
6963			err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6964				DQUOT_USAGE_ENABLED |
6965				(quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6966			if (err) {
6967				ext4_warning(sb,
6968					"Failed to enable quota tracking "
6969					"(type=%d, err=%d, ino=%lu). "
6970					"Please run e2fsck to fix.", type,
6971					err, qf_inums[type]);
6972				for (type--; type >= 0; type--) {
6973					struct inode *inode;
6974
6975					inode = sb_dqopt(sb)->files[type];
6976					if (inode)
6977						inode = igrab(inode);
6978					dquot_quota_off(sb, type);
6979					if (inode) {
6980						lockdep_set_quota_inode(inode,
6981							I_DATA_SEM_NORMAL);
6982						iput(inode);
6983					}
6984				}
6985
6986				return err;
6987			}
6988		}
6989	}
6990	return 0;
6991}
6992
6993static int ext4_quota_off(struct super_block *sb, int type)
6994{
6995	struct inode *inode = sb_dqopt(sb)->files[type];
6996	handle_t *handle;
6997	int err;
6998
6999	/* Force all delayed allocation blocks to be allocated.
7000	 * Caller already holds s_umount sem */
7001	if (test_opt(sb, DELALLOC))
7002		sync_filesystem(sb);
7003
7004	if (!inode || !igrab(inode))
7005		goto out;
7006
7007	err = dquot_quota_off(sb, type);
7008	if (err || ext4_has_feature_quota(sb))
7009		goto out_put;
7010
7011	inode_lock(inode);
7012	/*
7013	 * Update modification times of quota files when userspace can
7014	 * start looking at them. If we fail, we return success anyway since
7015	 * this is not a hard failure and quotas are already disabled.
7016	 */
7017	handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
7018	if (IS_ERR(handle)) {
7019		err = PTR_ERR(handle);
7020		goto out_unlock;
7021	}
7022	EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
7023	inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
7024	inode->i_mtime = inode->i_ctime = current_time(inode);
7025	err = ext4_mark_inode_dirty(handle, inode);
7026	ext4_journal_stop(handle);
7027out_unlock:
7028	inode_unlock(inode);
7029out_put:
7030	lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
7031	iput(inode);
7032	return err;
7033out:
7034	return dquot_quota_off(sb, type);
7035}
7036
7037/* Read data from quotafile - avoid pagecache and such because we cannot afford
7038 * acquiring the locks... As quota files are never truncated and quota code
7039 * itself serializes the operations (and no one else should touch the files)
7040 * we don't have to be afraid of races */
7041static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
7042			       size_t len, loff_t off)
7043{
7044	struct inode *inode = sb_dqopt(sb)->files[type];
7045	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7046	int offset = off & (sb->s_blocksize - 1);
7047	int tocopy;
7048	size_t toread;
7049	struct buffer_head *bh;
7050	loff_t i_size = i_size_read(inode);
7051
7052	if (off > i_size)
7053		return 0;
7054	if (off+len > i_size)
7055		len = i_size-off;
7056	toread = len;
7057	while (toread > 0) {
7058		tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
 
7059		bh = ext4_bread(NULL, inode, blk, 0);
7060		if (IS_ERR(bh))
7061			return PTR_ERR(bh);
7062		if (!bh)	/* A hole? */
7063			memset(data, 0, tocopy);
7064		else
7065			memcpy(data, bh->b_data+offset, tocopy);
7066		brelse(bh);
7067		offset = 0;
7068		toread -= tocopy;
7069		data += tocopy;
7070		blk++;
7071	}
7072	return len;
7073}
7074
7075/* Write to quotafile (we know the transaction is already started and has
7076 * enough credits) */
7077static ssize_t ext4_quota_write(struct super_block *sb, int type,
7078				const char *data, size_t len, loff_t off)
7079{
7080	struct inode *inode = sb_dqopt(sb)->files[type];
7081	ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7082	int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
7083	int retries = 0;
7084	struct buffer_head *bh;
7085	handle_t *handle = journal_current_handle();
7086
7087	if (!handle) {
7088		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7089			" cancelled because transaction is not started",
7090			(unsigned long long)off, (unsigned long long)len);
7091		return -EIO;
7092	}
7093	/*
7094	 * Since we account only one data block in transaction credits,
7095	 * then it is impossible to cross a block boundary.
7096	 */
7097	if (sb->s_blocksize - offset < len) {
7098		ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7099			" cancelled because not block aligned",
7100			(unsigned long long)off, (unsigned long long)len);
7101		return -EIO;
7102	}
7103
7104	do {
7105		bh = ext4_bread(handle, inode, blk,
7106				EXT4_GET_BLOCKS_CREATE |
7107				EXT4_GET_BLOCKS_METADATA_NOFAIL);
7108	} while (PTR_ERR(bh) == -ENOSPC &&
7109		 ext4_should_retry_alloc(inode->i_sb, &retries));
7110	if (IS_ERR(bh))
7111		return PTR_ERR(bh);
7112	if (!bh)
7113		goto out;
7114	BUFFER_TRACE(bh, "get write access");
7115	err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
7116	if (err) {
7117		brelse(bh);
7118		return err;
7119	}
7120	lock_buffer(bh);
7121	memcpy(bh->b_data+offset, data, len);
7122	flush_dcache_page(bh->b_page);
7123	unlock_buffer(bh);
7124	err = ext4_handle_dirty_metadata(handle, NULL, bh);
7125	brelse(bh);
7126out:
7127	if (inode->i_size < off + len) {
7128		i_size_write(inode, off + len);
7129		EXT4_I(inode)->i_disksize = inode->i_size;
7130		err2 = ext4_mark_inode_dirty(handle, inode);
7131		if (unlikely(err2 && !err))
7132			err = err2;
7133	}
7134	return err ? err : len;
7135}
7136#endif
7137
 
 
 
 
 
 
7138#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
7139static inline void register_as_ext2(void)
7140{
7141	int err = register_filesystem(&ext2_fs_type);
7142	if (err)
7143		printk(KERN_WARNING
7144		       "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7145}
7146
7147static inline void unregister_as_ext2(void)
7148{
7149	unregister_filesystem(&ext2_fs_type);
7150}
7151
7152static inline int ext2_feature_set_ok(struct super_block *sb)
7153{
7154	if (ext4_has_unknown_ext2_incompat_features(sb))
7155		return 0;
7156	if (sb_rdonly(sb))
7157		return 1;
7158	if (ext4_has_unknown_ext2_ro_compat_features(sb))
7159		return 0;
7160	return 1;
7161}
7162#else
7163static inline void register_as_ext2(void) { }
7164static inline void unregister_as_ext2(void) { }
7165static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
7166#endif
7167
7168static inline void register_as_ext3(void)
7169{
7170	int err = register_filesystem(&ext3_fs_type);
7171	if (err)
7172		printk(KERN_WARNING
7173		       "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7174}
7175
7176static inline void unregister_as_ext3(void)
7177{
7178	unregister_filesystem(&ext3_fs_type);
7179}
7180
7181static inline int ext3_feature_set_ok(struct super_block *sb)
7182{
7183	if (ext4_has_unknown_ext3_incompat_features(sb))
7184		return 0;
7185	if (!ext4_has_feature_journal(sb))
7186		return 0;
7187	if (sb_rdonly(sb))
7188		return 1;
7189	if (ext4_has_unknown_ext3_ro_compat_features(sb))
7190		return 0;
7191	return 1;
7192}
7193
7194static struct file_system_type ext4_fs_type = {
7195	.owner			= THIS_MODULE,
7196	.name			= "ext4",
7197	.init_fs_context	= ext4_init_fs_context,
7198	.parameters		= ext4_param_specs,
7199	.kill_sb		= kill_block_super,
7200	.fs_flags		= FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
7201};
7202MODULE_ALIAS_FS("ext4");
7203
7204/* Shared across all ext4 file systems */
7205wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
7206
7207static int __init ext4_init_fs(void)
7208{
7209	int i, err;
7210
7211	ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
7212	ext4_li_info = NULL;
7213
7214	/* Build-time check for flags consistency */
7215	ext4_check_flag_values();
7216
7217	for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
7218		init_waitqueue_head(&ext4__ioend_wq[i]);
7219
7220	err = ext4_init_es();
7221	if (err)
7222		return err;
7223
7224	err = ext4_init_pending();
7225	if (err)
7226		goto out7;
7227
7228	err = ext4_init_post_read_processing();
7229	if (err)
7230		goto out6;
7231
7232	err = ext4_init_pageio();
7233	if (err)
7234		goto out5;
7235
7236	err = ext4_init_system_zone();
7237	if (err)
7238		goto out4;
7239
7240	err = ext4_init_sysfs();
7241	if (err)
7242		goto out3;
7243
7244	err = ext4_init_mballoc();
7245	if (err)
7246		goto out2;
7247	err = init_inodecache();
7248	if (err)
7249		goto out1;
7250
7251	err = ext4_fc_init_dentry_cache();
7252	if (err)
7253		goto out05;
7254
7255	register_as_ext3();
7256	register_as_ext2();
7257	err = register_filesystem(&ext4_fs_type);
7258	if (err)
7259		goto out;
7260
7261	return 0;
7262out:
7263	unregister_as_ext2();
7264	unregister_as_ext3();
7265	ext4_fc_destroy_dentry_cache();
7266out05:
7267	destroy_inodecache();
7268out1:
7269	ext4_exit_mballoc();
7270out2:
7271	ext4_exit_sysfs();
7272out3:
7273	ext4_exit_system_zone();
7274out4:
7275	ext4_exit_pageio();
7276out5:
7277	ext4_exit_post_read_processing();
7278out6:
7279	ext4_exit_pending();
7280out7:
7281	ext4_exit_es();
7282
7283	return err;
7284}
7285
7286static void __exit ext4_exit_fs(void)
7287{
7288	ext4_destroy_lazyinit_thread();
7289	unregister_as_ext2();
7290	unregister_as_ext3();
7291	unregister_filesystem(&ext4_fs_type);
7292	ext4_fc_destroy_dentry_cache();
7293	destroy_inodecache();
7294	ext4_exit_mballoc();
7295	ext4_exit_sysfs();
7296	ext4_exit_system_zone();
7297	ext4_exit_pageio();
7298	ext4_exit_post_read_processing();
7299	ext4_exit_es();
7300	ext4_exit_pending();
7301}
7302
7303MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
7304MODULE_DESCRIPTION("Fourth Extended Filesystem");
7305MODULE_LICENSE("GPL");
7306MODULE_SOFTDEP("pre: crc32c");
7307module_init(ext4_init_fs)
7308module_exit(ext4_exit_fs)