1f5166768STheodore Ts'o // SPDX-License-Identifier: GPL-2.0 2ac27a0ecSDave Kleikamp /* 3617ba13bSMingming Cao * linux/fs/ext4/super.c 4ac27a0ecSDave Kleikamp * 5ac27a0ecSDave Kleikamp * Copyright (C) 1992, 1993, 1994, 1995 6ac27a0ecSDave Kleikamp * Remy Card (card@masi.ibp.fr) 7ac27a0ecSDave Kleikamp * Laboratoire MASI - Institut Blaise Pascal 8ac27a0ecSDave Kleikamp * Universite Pierre et Marie Curie (Paris VI) 9ac27a0ecSDave Kleikamp * 10ac27a0ecSDave Kleikamp * from 11ac27a0ecSDave Kleikamp * 12ac27a0ecSDave Kleikamp * linux/fs/minix/inode.c 13ac27a0ecSDave Kleikamp * 14ac27a0ecSDave Kleikamp * Copyright (C) 1991, 1992 Linus Torvalds 15ac27a0ecSDave Kleikamp * 16ac27a0ecSDave Kleikamp * Big-endian to little-endian byte-swapping/bitmaps by 17ac27a0ecSDave Kleikamp * David S. Miller (davem@caip.rutgers.edu), 1995 18ac27a0ecSDave Kleikamp */ 19ac27a0ecSDave Kleikamp 20ac27a0ecSDave Kleikamp #include <linux/module.h> 21ac27a0ecSDave Kleikamp #include <linux/string.h> 22ac27a0ecSDave Kleikamp #include <linux/fs.h> 23ac27a0ecSDave Kleikamp #include <linux/time.h> 24c5ca7c76STheodore Ts'o #include <linux/vmalloc.h> 25ac27a0ecSDave Kleikamp #include <linux/slab.h> 26ac27a0ecSDave Kleikamp #include <linux/init.h> 27ac27a0ecSDave Kleikamp #include <linux/blkdev.h> 2866114cadSTejun Heo #include <linux/backing-dev.h> 29ac27a0ecSDave Kleikamp #include <linux/parser.h> 30ac27a0ecSDave Kleikamp #include <linux/buffer_head.h> 31a5694255SChristoph Hellwig #include <linux/exportfs.h> 32ac27a0ecSDave Kleikamp #include <linux/vfs.h> 33ac27a0ecSDave Kleikamp #include <linux/random.h> 34ac27a0ecSDave Kleikamp #include <linux/mount.h> 35ac27a0ecSDave Kleikamp #include <linux/namei.h> 36ac27a0ecSDave Kleikamp #include <linux/quotaops.h> 37ac27a0ecSDave Kleikamp #include <linux/seq_file.h> 383197ebdbSTheodore Ts'o #include <linux/ctype.h> 391330593eSVignesh Babu #include <linux/log2.h> 40717d50e4SAndreas Dilger #include <linux/crc16.h> 41ef510424SDan Williams #include <linux/dax.h> 427abc52c2SDan Magenheimer #include <linux/cleancache.h> 437c0f6ba6SLinus Torvalds #include <linux/uaccess.h> 44ee73f9a5SJeff Layton #include <linux/iversion.h> 45c83ad55eSGabriel Krisman Bertazi #include <linux/unicode.h> 46c6a564ffSChristoph Hellwig #include <linux/part_stat.h> 47bfff6873SLukas Czerner #include <linux/kthread.h> 48bfff6873SLukas Czerner #include <linux/freezer.h> 49bfff6873SLukas Czerner 503dcf5451SChristoph Hellwig #include "ext4.h" 514a092d73STheodore Ts'o #include "ext4_extents.h" /* Needed for trace points definition */ 523dcf5451SChristoph Hellwig #include "ext4_jbd2.h" 53ac27a0ecSDave Kleikamp #include "xattr.h" 54ac27a0ecSDave Kleikamp #include "acl.h" 553661d286STheodore Ts'o #include "mballoc.h" 560c9ec4beSDarrick J. Wong #include "fsmap.h" 57ac27a0ecSDave Kleikamp 589bffad1eSTheodore Ts'o #define CREATE_TRACE_POINTS 599bffad1eSTheodore Ts'o #include <trace/events/ext4.h> 609bffad1eSTheodore Ts'o 610b75a840SLukas Czerner static struct ext4_lazy_init *ext4_li_info; 620b75a840SLukas Czerner static struct mutex ext4_li_mtx; 63e294a537STheodore Ts'o static struct ratelimit_state ext4_mount_msg_ratelimit; 649f6200bbSTheodore Ts'o 65617ba13bSMingming Cao static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 66ac27a0ecSDave Kleikamp unsigned long journal_devnum); 672adf6da8STheodore Ts'o static int ext4_show_options(struct seq_file *seq, struct dentry *root); 68e2d67052STheodore Ts'o static int ext4_commit_super(struct super_block *sb, int sync); 69617ba13bSMingming Cao static void ext4_mark_recovery_complete(struct super_block *sb, 70617ba13bSMingming Cao struct ext4_super_block *es); 71617ba13bSMingming Cao static void ext4_clear_journal_err(struct super_block *sb, 72617ba13bSMingming Cao struct ext4_super_block *es); 73617ba13bSMingming Cao static int ext4_sync_fs(struct super_block *sb, int wait); 74617ba13bSMingming Cao static int ext4_remount(struct super_block *sb, int *flags, char *data); 75617ba13bSMingming Cao static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 76c4be0c1dSTakashi Sato static int ext4_unfreeze(struct super_block *sb); 77c4be0c1dSTakashi Sato static int ext4_freeze(struct super_block *sb); 78152a0836SAl Viro static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 79152a0836SAl Viro const char *dev_name, void *data); 802035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb); 812035e776STheodore Ts'o static inline int ext3_feature_set_ok(struct super_block *sb); 82d39195c3SAmir Goldstein static int ext4_feature_set_ok(struct super_block *sb, int readonly); 83bfff6873SLukas Czerner static void ext4_destroy_lazyinit_thread(void); 84bfff6873SLukas Czerner static void ext4_unregister_li_request(struct super_block *sb); 858f1f7453SEric Sandeen static void ext4_clear_request_list(void); 86c6cb7e77SEric Whitney static struct inode *ext4_get_journal_inode(struct super_block *sb, 87c6cb7e77SEric Whitney unsigned int journal_inum); 88ac27a0ecSDave Kleikamp 89e74031fdSJan Kara /* 90e74031fdSJan Kara * Lock ordering 91e74031fdSJan Kara * 92e74031fdSJan Kara * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and 93e74031fdSJan Kara * i_mmap_rwsem (inode->i_mmap_rwsem)! 94e74031fdSJan Kara * 95e74031fdSJan Kara * page fault path: 96e74031fdSJan Kara * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start -> 97e74031fdSJan Kara * page lock -> i_data_sem (rw) 98e74031fdSJan Kara * 99e74031fdSJan Kara * buffered write path: 100e74031fdSJan Kara * sb_start_write -> i_mutex -> mmap_sem 101e74031fdSJan Kara * sb_start_write -> i_mutex -> transaction start -> page lock -> 102e74031fdSJan Kara * i_data_sem (rw) 103e74031fdSJan Kara * 104e74031fdSJan Kara * truncate: 1051d39834fSNikolay Borisov * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock 1061d39834fSNikolay Borisov * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start -> 1071d39834fSNikolay Borisov * i_data_sem (rw) 108e74031fdSJan Kara * 109e74031fdSJan Kara * direct IO: 1101d39834fSNikolay Borisov * sb_start_write -> i_mutex -> mmap_sem 1111d39834fSNikolay Borisov * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw) 112e74031fdSJan Kara * 113e74031fdSJan Kara * writepages: 114e74031fdSJan Kara * transaction start -> page lock(s) -> i_data_sem (rw) 115e74031fdSJan Kara */ 116e74031fdSJan Kara 117c290ea01SJan Kara #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 1182035e776STheodore Ts'o static struct file_system_type ext2_fs_type = { 1192035e776STheodore Ts'o .owner = THIS_MODULE, 1202035e776STheodore Ts'o .name = "ext2", 1212035e776STheodore Ts'o .mount = ext4_mount, 1222035e776STheodore Ts'o .kill_sb = kill_block_super, 1232035e776STheodore Ts'o .fs_flags = FS_REQUIRES_DEV, 1242035e776STheodore Ts'o }; 1257f78e035SEric W. Biederman MODULE_ALIAS_FS("ext2"); 126fa7614ddSEric W. Biederman MODULE_ALIAS("ext2"); 1272035e776STheodore Ts'o #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 1282035e776STheodore Ts'o #else 1292035e776STheodore Ts'o #define IS_EXT2_SB(sb) (0) 1302035e776STheodore Ts'o #endif 1312035e776STheodore Ts'o 1322035e776STheodore Ts'o 133ba69f9abSJan Kara static struct file_system_type ext3_fs_type = { 134ba69f9abSJan Kara .owner = THIS_MODULE, 135ba69f9abSJan Kara .name = "ext3", 136152a0836SAl Viro .mount = ext4_mount, 137ba69f9abSJan Kara .kill_sb = kill_block_super, 138ba69f9abSJan Kara .fs_flags = FS_REQUIRES_DEV, 139ba69f9abSJan Kara }; 1407f78e035SEric W. Biederman MODULE_ALIAS_FS("ext3"); 141fa7614ddSEric W. Biederman MODULE_ALIAS("ext3"); 142ba69f9abSJan Kara #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 143bd81d8eeSLaurent Vivier 144fb265c9cSTheodore Ts'o /* 145fb265c9cSTheodore Ts'o * This works like sb_bread() except it uses ERR_PTR for error 146fb265c9cSTheodore Ts'o * returns. Currently with sb_bread it's impossible to distinguish 147fb265c9cSTheodore Ts'o * between ENOMEM and EIO situations (since both result in a NULL 148fb265c9cSTheodore Ts'o * return. 149fb265c9cSTheodore Ts'o */ 150fb265c9cSTheodore Ts'o struct buffer_head * 151fb265c9cSTheodore Ts'o ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags) 152fb265c9cSTheodore Ts'o { 153fb265c9cSTheodore Ts'o struct buffer_head *bh = sb_getblk(sb, block); 154fb265c9cSTheodore Ts'o 155fb265c9cSTheodore Ts'o if (bh == NULL) 156fb265c9cSTheodore Ts'o return ERR_PTR(-ENOMEM); 157cf2834a5STheodore Ts'o if (ext4_buffer_uptodate(bh)) 158fb265c9cSTheodore Ts'o return bh; 159fb265c9cSTheodore Ts'o ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh); 160fb265c9cSTheodore Ts'o wait_on_buffer(bh); 161fb265c9cSTheodore Ts'o if (buffer_uptodate(bh)) 162fb265c9cSTheodore Ts'o return bh; 163fb265c9cSTheodore Ts'o put_bh(bh); 164fb265c9cSTheodore Ts'o return ERR_PTR(-EIO); 165fb265c9cSTheodore Ts'o } 166fb265c9cSTheodore Ts'o 167d25425f8SDarrick J. Wong static int ext4_verify_csum_type(struct super_block *sb, 168d25425f8SDarrick J. Wong struct ext4_super_block *es) 169d25425f8SDarrick J. Wong { 170e2b911c5SDarrick J. Wong if (!ext4_has_feature_metadata_csum(sb)) 171d25425f8SDarrick J. Wong return 1; 172d25425f8SDarrick J. Wong 173d25425f8SDarrick J. Wong return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 174d25425f8SDarrick J. Wong } 175d25425f8SDarrick J. Wong 176a9c47317SDarrick J. Wong static __le32 ext4_superblock_csum(struct super_block *sb, 177a9c47317SDarrick J. Wong struct ext4_super_block *es) 178a9c47317SDarrick J. Wong { 179a9c47317SDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 180a9c47317SDarrick J. Wong int offset = offsetof(struct ext4_super_block, s_checksum); 181a9c47317SDarrick J. Wong __u32 csum; 182a9c47317SDarrick J. Wong 183a9c47317SDarrick J. Wong csum = ext4_chksum(sbi, ~0, (char *)es, offset); 184a9c47317SDarrick J. Wong 185a9c47317SDarrick J. Wong return cpu_to_le32(csum); 186a9c47317SDarrick J. Wong } 187a9c47317SDarrick J. Wong 188c197855eSStephen Hemminger static int ext4_superblock_csum_verify(struct super_block *sb, 189a9c47317SDarrick J. Wong struct ext4_super_block *es) 190a9c47317SDarrick J. Wong { 1919aa5d32bSDmitry Monakhov if (!ext4_has_metadata_csum(sb)) 192a9c47317SDarrick J. Wong return 1; 193a9c47317SDarrick J. Wong 194a9c47317SDarrick J. Wong return es->s_checksum == ext4_superblock_csum(sb, es); 195a9c47317SDarrick J. Wong } 196a9c47317SDarrick J. Wong 19706db49e6STheodore Ts'o void ext4_superblock_csum_set(struct super_block *sb) 198a9c47317SDarrick J. Wong { 19906db49e6STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 20006db49e6STheodore Ts'o 2019aa5d32bSDmitry Monakhov if (!ext4_has_metadata_csum(sb)) 202a9c47317SDarrick J. Wong return; 203a9c47317SDarrick J. Wong 204a9c47317SDarrick J. Wong es->s_checksum = ext4_superblock_csum(sb, es); 205a9c47317SDarrick J. Wong } 206a9c47317SDarrick J. Wong 2078fadc143SAlexandre Ratchov ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 2088fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 209bd81d8eeSLaurent Vivier { 2103a14589cSAneesh Kumar K.V return le32_to_cpu(bg->bg_block_bitmap_lo) | 2118fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 2128fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 213bd81d8eeSLaurent Vivier } 214bd81d8eeSLaurent Vivier 2158fadc143SAlexandre Ratchov ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 2168fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 217bd81d8eeSLaurent Vivier { 2185272f837SAneesh Kumar K.V return le32_to_cpu(bg->bg_inode_bitmap_lo) | 2198fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 2208fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 221bd81d8eeSLaurent Vivier } 222bd81d8eeSLaurent Vivier 2238fadc143SAlexandre Ratchov ext4_fsblk_t ext4_inode_table(struct super_block *sb, 2248fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 225bd81d8eeSLaurent Vivier { 2265272f837SAneesh Kumar K.V return le32_to_cpu(bg->bg_inode_table_lo) | 2278fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 2288fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 229bd81d8eeSLaurent Vivier } 230bd81d8eeSLaurent Vivier 231021b65bbSTheodore Ts'o __u32 ext4_free_group_clusters(struct super_block *sb, 232560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 233560671a0SAneesh Kumar K.V { 234560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_free_blocks_count_lo) | 235560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 236560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 237560671a0SAneesh Kumar K.V } 238560671a0SAneesh Kumar K.V 239560671a0SAneesh Kumar K.V __u32 ext4_free_inodes_count(struct super_block *sb, 240560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 241560671a0SAneesh Kumar K.V { 242560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_free_inodes_count_lo) | 243560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 244560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 245560671a0SAneesh Kumar K.V } 246560671a0SAneesh Kumar K.V 247560671a0SAneesh Kumar K.V __u32 ext4_used_dirs_count(struct super_block *sb, 248560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 249560671a0SAneesh Kumar K.V { 250560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_used_dirs_count_lo) | 251560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 252560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 253560671a0SAneesh Kumar K.V } 254560671a0SAneesh Kumar K.V 255560671a0SAneesh Kumar K.V __u32 ext4_itable_unused_count(struct super_block *sb, 256560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 257560671a0SAneesh Kumar K.V { 258560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_itable_unused_lo) | 259560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 260560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 261560671a0SAneesh Kumar K.V } 262560671a0SAneesh Kumar K.V 2638fadc143SAlexandre Ratchov void ext4_block_bitmap_set(struct super_block *sb, 2648fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 265bd81d8eeSLaurent Vivier { 2663a14589cSAneesh Kumar K.V bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 2678fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 2688fadc143SAlexandre Ratchov bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 269bd81d8eeSLaurent Vivier } 270bd81d8eeSLaurent Vivier 2718fadc143SAlexandre Ratchov void ext4_inode_bitmap_set(struct super_block *sb, 2728fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 273bd81d8eeSLaurent Vivier { 2745272f837SAneesh Kumar K.V bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 2758fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 2768fadc143SAlexandre Ratchov bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 277bd81d8eeSLaurent Vivier } 278bd81d8eeSLaurent Vivier 2798fadc143SAlexandre Ratchov void ext4_inode_table_set(struct super_block *sb, 2808fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 281bd81d8eeSLaurent Vivier { 2825272f837SAneesh Kumar K.V bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 2838fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 2848fadc143SAlexandre Ratchov bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 285bd81d8eeSLaurent Vivier } 286bd81d8eeSLaurent Vivier 287021b65bbSTheodore Ts'o void ext4_free_group_clusters_set(struct super_block *sb, 288560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 289560671a0SAneesh Kumar K.V { 290560671a0SAneesh Kumar K.V bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 291560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 292560671a0SAneesh Kumar K.V bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 293560671a0SAneesh Kumar K.V } 294560671a0SAneesh Kumar K.V 295560671a0SAneesh Kumar K.V void ext4_free_inodes_set(struct super_block *sb, 296560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 297560671a0SAneesh Kumar K.V { 298560671a0SAneesh Kumar K.V bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 299560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 300560671a0SAneesh Kumar K.V bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 301560671a0SAneesh Kumar K.V } 302560671a0SAneesh Kumar K.V 303560671a0SAneesh Kumar K.V void ext4_used_dirs_set(struct super_block *sb, 304560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 305560671a0SAneesh Kumar K.V { 306560671a0SAneesh Kumar K.V bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 307560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 308560671a0SAneesh Kumar K.V bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 309560671a0SAneesh Kumar K.V } 310560671a0SAneesh Kumar K.V 311560671a0SAneesh Kumar K.V void ext4_itable_unused_set(struct super_block *sb, 312560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 313560671a0SAneesh Kumar K.V { 314560671a0SAneesh Kumar K.V bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 315560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 316560671a0SAneesh Kumar K.V bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 317560671a0SAneesh Kumar K.V } 318560671a0SAneesh Kumar K.V 3196a0678a7SArnd Bergmann static void __ext4_update_tstamp(__le32 *lo, __u8 *hi) 3206a0678a7SArnd Bergmann { 3216a0678a7SArnd Bergmann time64_t now = ktime_get_real_seconds(); 3226a0678a7SArnd Bergmann 3236a0678a7SArnd Bergmann now = clamp_val(now, 0, (1ull << 40) - 1); 3246a0678a7SArnd Bergmann 3256a0678a7SArnd Bergmann *lo = cpu_to_le32(lower_32_bits(now)); 3266a0678a7SArnd Bergmann *hi = upper_32_bits(now); 3276a0678a7SArnd Bergmann } 3286a0678a7SArnd Bergmann 3296a0678a7SArnd Bergmann static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi) 3306a0678a7SArnd Bergmann { 3316a0678a7SArnd Bergmann return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo); 3326a0678a7SArnd Bergmann } 3336a0678a7SArnd Bergmann #define ext4_update_tstamp(es, tstamp) \ 3346a0678a7SArnd Bergmann __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 3356a0678a7SArnd Bergmann #define ext4_get_tstamp(es, tstamp) \ 3366a0678a7SArnd Bergmann __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 337d3d1faf6SCurt Wohlgemuth 33854d3adbcSTheodore Ts'o static void __save_error_info(struct super_block *sb, int error, 33954d3adbcSTheodore Ts'o __u32 ino, __u64 block, 34054d3adbcSTheodore Ts'o const char *func, unsigned int line) 3411c13d5c0STheodore Ts'o { 3421c13d5c0STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 34354d3adbcSTheodore Ts'o int err; 3441c13d5c0STheodore Ts'o 3451c13d5c0STheodore Ts'o EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 3461b46617bSTheodore Ts'o if (bdev_read_only(sb->s_bdev)) 3471b46617bSTheodore Ts'o return; 3481c13d5c0STheodore Ts'o es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 3496a0678a7SArnd Bergmann ext4_update_tstamp(es, s_last_error_time); 3501c13d5c0STheodore Ts'o strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); 3511c13d5c0STheodore Ts'o es->s_last_error_line = cpu_to_le32(line); 35254d3adbcSTheodore Ts'o es->s_last_error_ino = cpu_to_le32(ino); 35354d3adbcSTheodore Ts'o es->s_last_error_block = cpu_to_le64(block); 35454d3adbcSTheodore Ts'o switch (error) { 35554d3adbcSTheodore Ts'o case EIO: 35654d3adbcSTheodore Ts'o err = EXT4_ERR_EIO; 35754d3adbcSTheodore Ts'o break; 35854d3adbcSTheodore Ts'o case ENOMEM: 35954d3adbcSTheodore Ts'o err = EXT4_ERR_ENOMEM; 36054d3adbcSTheodore Ts'o break; 36154d3adbcSTheodore Ts'o case EFSBADCRC: 36254d3adbcSTheodore Ts'o err = EXT4_ERR_EFSBADCRC; 36354d3adbcSTheodore Ts'o break; 36454d3adbcSTheodore Ts'o case 0: 36554d3adbcSTheodore Ts'o case EFSCORRUPTED: 36654d3adbcSTheodore Ts'o err = EXT4_ERR_EFSCORRUPTED; 36754d3adbcSTheodore Ts'o break; 36854d3adbcSTheodore Ts'o case ENOSPC: 36954d3adbcSTheodore Ts'o err = EXT4_ERR_ENOSPC; 37054d3adbcSTheodore Ts'o break; 37154d3adbcSTheodore Ts'o case ENOKEY: 37254d3adbcSTheodore Ts'o err = EXT4_ERR_ENOKEY; 37354d3adbcSTheodore Ts'o break; 37454d3adbcSTheodore Ts'o case EROFS: 37554d3adbcSTheodore Ts'o err = EXT4_ERR_EROFS; 37654d3adbcSTheodore Ts'o break; 37754d3adbcSTheodore Ts'o case EFBIG: 37854d3adbcSTheodore Ts'o err = EXT4_ERR_EFBIG; 37954d3adbcSTheodore Ts'o break; 38054d3adbcSTheodore Ts'o case EEXIST: 38154d3adbcSTheodore Ts'o err = EXT4_ERR_EEXIST; 38254d3adbcSTheodore Ts'o break; 38354d3adbcSTheodore Ts'o case ERANGE: 38454d3adbcSTheodore Ts'o err = EXT4_ERR_ERANGE; 38554d3adbcSTheodore Ts'o break; 38654d3adbcSTheodore Ts'o case EOVERFLOW: 38754d3adbcSTheodore Ts'o err = EXT4_ERR_EOVERFLOW; 38854d3adbcSTheodore Ts'o break; 38954d3adbcSTheodore Ts'o case EBUSY: 39054d3adbcSTheodore Ts'o err = EXT4_ERR_EBUSY; 39154d3adbcSTheodore Ts'o break; 39254d3adbcSTheodore Ts'o case ENOTDIR: 39354d3adbcSTheodore Ts'o err = EXT4_ERR_ENOTDIR; 39454d3adbcSTheodore Ts'o break; 39554d3adbcSTheodore Ts'o case ENOTEMPTY: 39654d3adbcSTheodore Ts'o err = EXT4_ERR_ENOTEMPTY; 39754d3adbcSTheodore Ts'o break; 39854d3adbcSTheodore Ts'o case ESHUTDOWN: 39954d3adbcSTheodore Ts'o err = EXT4_ERR_ESHUTDOWN; 40054d3adbcSTheodore Ts'o break; 40154d3adbcSTheodore Ts'o case EFAULT: 40254d3adbcSTheodore Ts'o err = EXT4_ERR_EFAULT; 40354d3adbcSTheodore Ts'o break; 40454d3adbcSTheodore Ts'o default: 40554d3adbcSTheodore Ts'o err = EXT4_ERR_UNKNOWN; 40654d3adbcSTheodore Ts'o } 40754d3adbcSTheodore Ts'o es->s_last_error_errcode = err; 4081c13d5c0STheodore Ts'o if (!es->s_first_error_time) { 4091c13d5c0STheodore Ts'o es->s_first_error_time = es->s_last_error_time; 4106a0678a7SArnd Bergmann es->s_first_error_time_hi = es->s_last_error_time_hi; 4111c13d5c0STheodore Ts'o strncpy(es->s_first_error_func, func, 4121c13d5c0STheodore Ts'o sizeof(es->s_first_error_func)); 4131c13d5c0STheodore Ts'o es->s_first_error_line = cpu_to_le32(line); 4141c13d5c0STheodore Ts'o es->s_first_error_ino = es->s_last_error_ino; 4151c13d5c0STheodore Ts'o es->s_first_error_block = es->s_last_error_block; 416878520acSTheodore Ts'o es->s_first_error_errcode = es->s_last_error_errcode; 4171c13d5c0STheodore Ts'o } 41866e61a9eSTheodore Ts'o /* 41966e61a9eSTheodore Ts'o * Start the daily error reporting function if it hasn't been 42066e61a9eSTheodore Ts'o * started already 42166e61a9eSTheodore Ts'o */ 42266e61a9eSTheodore Ts'o if (!es->s_error_count) 42366e61a9eSTheodore Ts'o mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); 424ba39ebb6SWei Yongjun le32_add_cpu(&es->s_error_count, 1); 4251c13d5c0STheodore Ts'o } 4261c13d5c0STheodore Ts'o 42754d3adbcSTheodore Ts'o static void save_error_info(struct super_block *sb, int error, 42854d3adbcSTheodore Ts'o __u32 ino, __u64 block, 42954d3adbcSTheodore Ts'o const char *func, unsigned int line) 4301c13d5c0STheodore Ts'o { 43154d3adbcSTheodore Ts'o __save_error_info(sb, error, ino, block, func, line); 432c96e2b85SEric Sandeen if (!bdev_read_only(sb->s_bdev)) 4331c13d5c0STheodore Ts'o ext4_commit_super(sb, 1); 4341c13d5c0STheodore Ts'o } 4351c13d5c0STheodore Ts'o 436bdfe0cbdSTheodore Ts'o /* 437bdfe0cbdSTheodore Ts'o * The del_gendisk() function uninitializes the disk-specific data 438bdfe0cbdSTheodore Ts'o * structures, including the bdi structure, without telling anyone 439bdfe0cbdSTheodore Ts'o * else. Once this happens, any attempt to call mark_buffer_dirty() 440bdfe0cbdSTheodore Ts'o * (for example, by ext4_commit_super), will cause a kernel OOPS. 441bdfe0cbdSTheodore Ts'o * This is a kludge to prevent these oops until we can put in a proper 442bdfe0cbdSTheodore Ts'o * hook in del_gendisk() to inform the VFS and file system layers. 443bdfe0cbdSTheodore Ts'o */ 444bdfe0cbdSTheodore Ts'o static int block_device_ejected(struct super_block *sb) 445bdfe0cbdSTheodore Ts'o { 446bdfe0cbdSTheodore Ts'o struct inode *bd_inode = sb->s_bdev->bd_inode; 447bdfe0cbdSTheodore Ts'o struct backing_dev_info *bdi = inode_to_bdi(bd_inode); 448bdfe0cbdSTheodore Ts'o 449bdfe0cbdSTheodore Ts'o return bdi->dev == NULL; 450bdfe0cbdSTheodore Ts'o } 451bdfe0cbdSTheodore Ts'o 45218aadd47SBobi Jam static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 45318aadd47SBobi Jam { 45418aadd47SBobi Jam struct super_block *sb = journal->j_private; 45518aadd47SBobi Jam struct ext4_sb_info *sbi = EXT4_SB(sb); 45618aadd47SBobi Jam int error = is_journal_aborted(journal); 4575d3ee208SDmitry Monakhov struct ext4_journal_cb_entry *jce; 45818aadd47SBobi Jam 4595d3ee208SDmitry Monakhov BUG_ON(txn->t_state == T_FINISHED); 460a0154344SDaeho Jeong 461a0154344SDaeho Jeong ext4_process_freed_data(sb, txn->t_tid); 462a0154344SDaeho Jeong 46318aadd47SBobi Jam spin_lock(&sbi->s_md_lock); 4645d3ee208SDmitry Monakhov while (!list_empty(&txn->t_private_list)) { 4655d3ee208SDmitry Monakhov jce = list_entry(txn->t_private_list.next, 4665d3ee208SDmitry Monakhov struct ext4_journal_cb_entry, jce_list); 46718aadd47SBobi Jam list_del_init(&jce->jce_list); 46818aadd47SBobi Jam spin_unlock(&sbi->s_md_lock); 46918aadd47SBobi Jam jce->jce_func(sb, jce, error); 47018aadd47SBobi Jam spin_lock(&sbi->s_md_lock); 47118aadd47SBobi Jam } 47218aadd47SBobi Jam spin_unlock(&sbi->s_md_lock); 47318aadd47SBobi Jam } 4741c13d5c0STheodore Ts'o 4751dc1097fSJan Kara static bool system_going_down(void) 4761dc1097fSJan Kara { 4771dc1097fSJan Kara return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 4781dc1097fSJan Kara || system_state == SYSTEM_RESTART; 4791dc1097fSJan Kara } 4801dc1097fSJan Kara 481ac27a0ecSDave Kleikamp /* Deal with the reporting of failure conditions on a filesystem such as 482ac27a0ecSDave Kleikamp * inconsistencies detected or read IO failures. 483ac27a0ecSDave Kleikamp * 484ac27a0ecSDave Kleikamp * On ext2, we can store the error state of the filesystem in the 485617ba13bSMingming Cao * superblock. That is not possible on ext4, because we may have other 486ac27a0ecSDave Kleikamp * write ordering constraints on the superblock which prevent us from 487ac27a0ecSDave Kleikamp * writing it out straight away; and given that the journal is about to 488ac27a0ecSDave Kleikamp * be aborted, we can't rely on the current, or future, transactions to 489ac27a0ecSDave Kleikamp * write out the superblock safely. 490ac27a0ecSDave Kleikamp * 491dab291afSMingming Cao * We'll just use the jbd2_journal_abort() error code to record an error in 492d6b198bcSThadeu Lima de Souza Cascardo * the journal instead. On recovery, the journal will complain about 493ac27a0ecSDave Kleikamp * that error until we've noted it down and cleared it. 494ac27a0ecSDave Kleikamp */ 495ac27a0ecSDave Kleikamp 496617ba13bSMingming Cao static void ext4_handle_error(struct super_block *sb) 497ac27a0ecSDave Kleikamp { 498327eaf73STheodore Ts'o if (test_opt(sb, WARN_ON_ERROR)) 499327eaf73STheodore Ts'o WARN_ON_ONCE(1); 500327eaf73STheodore Ts'o 501bc98a42cSDavid Howells if (sb_rdonly(sb)) 502ac27a0ecSDave Kleikamp return; 503ac27a0ecSDave Kleikamp 504ac27a0ecSDave Kleikamp if (!test_opt(sb, ERRORS_CONT)) { 505617ba13bSMingming Cao journal_t *journal = EXT4_SB(sb)->s_journal; 506ac27a0ecSDave Kleikamp 5074ab2f15bSTheodore Ts'o EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 508ac27a0ecSDave Kleikamp if (journal) 509dab291afSMingming Cao jbd2_journal_abort(journal, -EIO); 510ac27a0ecSDave Kleikamp } 5111dc1097fSJan Kara /* 5121dc1097fSJan Kara * We force ERRORS_RO behavior when system is rebooting. Otherwise we 5131dc1097fSJan Kara * could panic during 'reboot -f' as the underlying device got already 5141dc1097fSJan Kara * disabled. 5151dc1097fSJan Kara */ 5161dc1097fSJan Kara if (test_opt(sb, ERRORS_RO) || system_going_down()) { 517b31e1552SEric Sandeen ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 5184418e141SDmitry Monakhov /* 5194418e141SDmitry Monakhov * Make sure updated value of ->s_mount_flags will be visible 5204418e141SDmitry Monakhov * before ->s_flags update 5214418e141SDmitry Monakhov */ 5224418e141SDmitry Monakhov smp_wmb(); 5231751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 5241dc1097fSJan Kara } else if (test_opt(sb, ERRORS_PANIC)) { 5254327ba52SDaeho Jeong if (EXT4_SB(sb)->s_journal && 5264327ba52SDaeho Jeong !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) 5274327ba52SDaeho Jeong return; 528617ba13bSMingming Cao panic("EXT4-fs (device %s): panic forced after error\n", 529ac27a0ecSDave Kleikamp sb->s_id); 530ac27a0ecSDave Kleikamp } 5314327ba52SDaeho Jeong } 532ac27a0ecSDave Kleikamp 533efbed4dcSTheodore Ts'o #define ext4_error_ratelimit(sb) \ 534efbed4dcSTheodore Ts'o ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 535efbed4dcSTheodore Ts'o "EXT4-fs error") 536efbed4dcSTheodore Ts'o 53712062dddSEric Sandeen void __ext4_error(struct super_block *sb, const char *function, 53854d3adbcSTheodore Ts'o unsigned int line, int error, __u64 block, 53954d3adbcSTheodore Ts'o const char *fmt, ...) 540ac27a0ecSDave Kleikamp { 5410ff2ea7dSJoe Perches struct va_format vaf; 542ac27a0ecSDave Kleikamp va_list args; 543ac27a0ecSDave Kleikamp 5440db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 5450db1ff22STheodore Ts'o return; 5460db1ff22STheodore Ts'o 547ccf0f32aSTheodore Ts'o trace_ext4_error(sb, function, line); 548efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 549ac27a0ecSDave Kleikamp va_start(args, fmt); 5500ff2ea7dSJoe Perches vaf.fmt = fmt; 5510ff2ea7dSJoe Perches vaf.va = &args; 552efbed4dcSTheodore Ts'o printk(KERN_CRIT 553efbed4dcSTheodore Ts'o "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 5540ff2ea7dSJoe Perches sb->s_id, function, line, current->comm, &vaf); 555ac27a0ecSDave Kleikamp va_end(args); 556efbed4dcSTheodore Ts'o } 55754d3adbcSTheodore Ts'o save_error_info(sb, error, 0, block, function, line); 558617ba13bSMingming Cao ext4_handle_error(sb); 559ac27a0ecSDave Kleikamp } 560ac27a0ecSDave Kleikamp 561e7c96e8eSJoe Perches void __ext4_error_inode(struct inode *inode, const char *function, 56254d3adbcSTheodore Ts'o unsigned int line, ext4_fsblk_t block, int error, 563273df556SFrank Mayhar const char *fmt, ...) 564273df556SFrank Mayhar { 565273df556SFrank Mayhar va_list args; 566f7c21177STheodore Ts'o struct va_format vaf; 567273df556SFrank Mayhar 5680db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 5690db1ff22STheodore Ts'o return; 5700db1ff22STheodore Ts'o 571ccf0f32aSTheodore Ts'o trace_ext4_error(inode->i_sb, function, line); 572efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(inode->i_sb)) { 573273df556SFrank Mayhar va_start(args, fmt); 574f7c21177STheodore Ts'o vaf.fmt = fmt; 575f7c21177STheodore Ts'o vaf.va = &args; 576c398eda0STheodore Ts'o if (block) 577d9ee81daSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 578d9ee81daSJoe Perches "inode #%lu: block %llu: comm %s: %pV\n", 579d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 580d9ee81daSJoe Perches block, current->comm, &vaf); 581d9ee81daSJoe Perches else 582d9ee81daSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 583d9ee81daSJoe Perches "inode #%lu: comm %s: %pV\n", 584d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 585d9ee81daSJoe Perches current->comm, &vaf); 586273df556SFrank Mayhar va_end(args); 587efbed4dcSTheodore Ts'o } 58854d3adbcSTheodore Ts'o save_error_info(inode->i_sb, error, inode->i_ino, block, 58954d3adbcSTheodore Ts'o function, line); 590273df556SFrank Mayhar ext4_handle_error(inode->i_sb); 591273df556SFrank Mayhar } 592273df556SFrank Mayhar 593e7c96e8eSJoe Perches void __ext4_error_file(struct file *file, const char *function, 594f7c21177STheodore Ts'o unsigned int line, ext4_fsblk_t block, 595f7c21177STheodore Ts'o const char *fmt, ...) 596273df556SFrank Mayhar { 597273df556SFrank Mayhar va_list args; 598f7c21177STheodore Ts'o struct va_format vaf; 599496ad9aaSAl Viro struct inode *inode = file_inode(file); 600273df556SFrank Mayhar char pathname[80], *path; 601273df556SFrank Mayhar 6020db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 6030db1ff22STheodore Ts'o return; 6040db1ff22STheodore Ts'o 605ccf0f32aSTheodore Ts'o trace_ext4_error(inode->i_sb, function, line); 606efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(inode->i_sb)) { 6079bf39ab2SMiklos Szeredi path = file_path(file, pathname, sizeof(pathname)); 608f9a62d09SDan Carpenter if (IS_ERR(path)) 609273df556SFrank Mayhar path = "(unknown)"; 610f7c21177STheodore Ts'o va_start(args, fmt); 611f7c21177STheodore Ts'o vaf.fmt = fmt; 612f7c21177STheodore Ts'o vaf.va = &args; 613d9ee81daSJoe Perches if (block) 614d9ee81daSJoe Perches printk(KERN_CRIT 615d9ee81daSJoe Perches "EXT4-fs error (device %s): %s:%d: inode #%lu: " 616d9ee81daSJoe Perches "block %llu: comm %s: path %s: %pV\n", 617d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 618d9ee81daSJoe Perches block, current->comm, path, &vaf); 619d9ee81daSJoe Perches else 620d9ee81daSJoe Perches printk(KERN_CRIT 621d9ee81daSJoe Perches "EXT4-fs error (device %s): %s:%d: inode #%lu: " 622d9ee81daSJoe Perches "comm %s: path %s: %pV\n", 623d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 624d9ee81daSJoe Perches current->comm, path, &vaf); 625273df556SFrank Mayhar va_end(args); 626efbed4dcSTheodore Ts'o } 62754d3adbcSTheodore Ts'o save_error_info(inode->i_sb, EFSCORRUPTED, inode->i_ino, block, 62854d3adbcSTheodore Ts'o function, line); 629273df556SFrank Mayhar ext4_handle_error(inode->i_sb); 630273df556SFrank Mayhar } 631273df556SFrank Mayhar 632722887ddSTheodore Ts'o const char *ext4_decode_error(struct super_block *sb, int errno, 633ac27a0ecSDave Kleikamp char nbuf[16]) 634ac27a0ecSDave Kleikamp { 635ac27a0ecSDave Kleikamp char *errstr = NULL; 636ac27a0ecSDave Kleikamp 637ac27a0ecSDave Kleikamp switch (errno) { 6386a797d27SDarrick J. Wong case -EFSCORRUPTED: 6396a797d27SDarrick J. Wong errstr = "Corrupt filesystem"; 6406a797d27SDarrick J. Wong break; 6416a797d27SDarrick J. Wong case -EFSBADCRC: 6426a797d27SDarrick J. Wong errstr = "Filesystem failed CRC"; 6436a797d27SDarrick J. Wong break; 644ac27a0ecSDave Kleikamp case -EIO: 645ac27a0ecSDave Kleikamp errstr = "IO failure"; 646ac27a0ecSDave Kleikamp break; 647ac27a0ecSDave Kleikamp case -ENOMEM: 648ac27a0ecSDave Kleikamp errstr = "Out of memory"; 649ac27a0ecSDave Kleikamp break; 650ac27a0ecSDave Kleikamp case -EROFS: 65178f1ddbbSTheodore Ts'o if (!sb || (EXT4_SB(sb)->s_journal && 65278f1ddbbSTheodore Ts'o EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 653ac27a0ecSDave Kleikamp errstr = "Journal has aborted"; 654ac27a0ecSDave Kleikamp else 655ac27a0ecSDave Kleikamp errstr = "Readonly filesystem"; 656ac27a0ecSDave Kleikamp break; 657ac27a0ecSDave Kleikamp default: 658ac27a0ecSDave Kleikamp /* If the caller passed in an extra buffer for unknown 659ac27a0ecSDave Kleikamp * errors, textualise them now. Else we just return 660ac27a0ecSDave Kleikamp * NULL. */ 661ac27a0ecSDave Kleikamp if (nbuf) { 662ac27a0ecSDave Kleikamp /* Check for truncated error codes... */ 663ac27a0ecSDave Kleikamp if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 664ac27a0ecSDave Kleikamp errstr = nbuf; 665ac27a0ecSDave Kleikamp } 666ac27a0ecSDave Kleikamp break; 667ac27a0ecSDave Kleikamp } 668ac27a0ecSDave Kleikamp 669ac27a0ecSDave Kleikamp return errstr; 670ac27a0ecSDave Kleikamp } 671ac27a0ecSDave Kleikamp 672617ba13bSMingming Cao /* __ext4_std_error decodes expected errors from journaling functions 673ac27a0ecSDave Kleikamp * automatically and invokes the appropriate error response. */ 674ac27a0ecSDave Kleikamp 675c398eda0STheodore Ts'o void __ext4_std_error(struct super_block *sb, const char *function, 676c398eda0STheodore Ts'o unsigned int line, int errno) 677ac27a0ecSDave Kleikamp { 678ac27a0ecSDave Kleikamp char nbuf[16]; 679ac27a0ecSDave Kleikamp const char *errstr; 680ac27a0ecSDave Kleikamp 6810db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 6820db1ff22STheodore Ts'o return; 6830db1ff22STheodore Ts'o 684ac27a0ecSDave Kleikamp /* Special case: if the error is EROFS, and we're not already 685ac27a0ecSDave Kleikamp * inside a transaction, then there's really no point in logging 686ac27a0ecSDave Kleikamp * an error. */ 687bc98a42cSDavid Howells if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb)) 688ac27a0ecSDave Kleikamp return; 689ac27a0ecSDave Kleikamp 690efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 691617ba13bSMingming Cao errstr = ext4_decode_error(sb, errno, nbuf); 692c398eda0STheodore Ts'o printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 693c398eda0STheodore Ts'o sb->s_id, function, line, errstr); 694efbed4dcSTheodore Ts'o } 695ac27a0ecSDave Kleikamp 69654d3adbcSTheodore Ts'o save_error_info(sb, -errno, 0, 0, function, line); 697617ba13bSMingming Cao ext4_handle_error(sb); 698ac27a0ecSDave Kleikamp } 699ac27a0ecSDave Kleikamp 700ac27a0ecSDave Kleikamp /* 701617ba13bSMingming Cao * ext4_abort is a much stronger failure handler than ext4_error. The 702ac27a0ecSDave Kleikamp * abort function may be used to deal with unrecoverable failures such 703ac27a0ecSDave Kleikamp * as journal IO errors or ENOMEM at a critical moment in log management. 704ac27a0ecSDave Kleikamp * 705ac27a0ecSDave Kleikamp * We unconditionally force the filesystem into an ABORT|READONLY state, 706ac27a0ecSDave Kleikamp * unless the error response on the fs has been set to panic in which 707ac27a0ecSDave Kleikamp * case we take the easy way out and panic immediately. 708ac27a0ecSDave Kleikamp */ 709ac27a0ecSDave Kleikamp 710c67d859eSTheodore Ts'o void __ext4_abort(struct super_block *sb, const char *function, 71154d3adbcSTheodore Ts'o unsigned int line, int error, const char *fmt, ...) 712ac27a0ecSDave Kleikamp { 713651e1c3bSJoe Perches struct va_format vaf; 714ac27a0ecSDave Kleikamp va_list args; 715ac27a0ecSDave Kleikamp 7160db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 7170db1ff22STheodore Ts'o return; 7180db1ff22STheodore Ts'o 71954d3adbcSTheodore Ts'o save_error_info(sb, error, 0, 0, function, line); 720ac27a0ecSDave Kleikamp va_start(args, fmt); 721651e1c3bSJoe Perches vaf.fmt = fmt; 722651e1c3bSJoe Perches vaf.va = &args; 723651e1c3bSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n", 724651e1c3bSJoe Perches sb->s_id, function, line, &vaf); 725ac27a0ecSDave Kleikamp va_end(args); 726ac27a0ecSDave Kleikamp 727bc98a42cSDavid Howells if (sb_rdonly(sb) == 0) { 728b31e1552SEric Sandeen ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 7294ab2f15bSTheodore Ts'o EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 7304418e141SDmitry Monakhov /* 7314418e141SDmitry Monakhov * Make sure updated value of ->s_mount_flags will be visible 7324418e141SDmitry Monakhov * before ->s_flags update 7334418e141SDmitry Monakhov */ 7344418e141SDmitry Monakhov smp_wmb(); 7351751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 736ef2cabf7SHidehiro Kawai if (EXT4_SB(sb)->s_journal) 737dab291afSMingming Cao jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); 7381c13d5c0STheodore Ts'o } 7392c1d0e36SJan Kara if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) { 7404327ba52SDaeho Jeong if (EXT4_SB(sb)->s_journal && 7414327ba52SDaeho Jeong !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) 7424327ba52SDaeho Jeong return; 7431c13d5c0STheodore Ts'o panic("EXT4-fs panic from previous error\n"); 744ac27a0ecSDave Kleikamp } 7454327ba52SDaeho Jeong } 746ac27a0ecSDave Kleikamp 747e7c96e8eSJoe Perches void __ext4_msg(struct super_block *sb, 748e7c96e8eSJoe Perches const char *prefix, const char *fmt, ...) 749b31e1552SEric Sandeen { 7500ff2ea7dSJoe Perches struct va_format vaf; 751b31e1552SEric Sandeen va_list args; 752b31e1552SEric Sandeen 753efbed4dcSTheodore Ts'o if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs")) 754efbed4dcSTheodore Ts'o return; 755efbed4dcSTheodore Ts'o 756b31e1552SEric Sandeen va_start(args, fmt); 7570ff2ea7dSJoe Perches vaf.fmt = fmt; 7580ff2ea7dSJoe Perches vaf.va = &args; 7590ff2ea7dSJoe Perches printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 760b31e1552SEric Sandeen va_end(args); 761b31e1552SEric Sandeen } 762b31e1552SEric Sandeen 763b03a2f7eSAndreas Dilger #define ext4_warning_ratelimit(sb) \ 764b03a2f7eSAndreas Dilger ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \ 765b03a2f7eSAndreas Dilger "EXT4-fs warning") 766b03a2f7eSAndreas Dilger 76712062dddSEric Sandeen void __ext4_warning(struct super_block *sb, const char *function, 768c398eda0STheodore Ts'o unsigned int line, const char *fmt, ...) 769ac27a0ecSDave Kleikamp { 7700ff2ea7dSJoe Perches struct va_format vaf; 771ac27a0ecSDave Kleikamp va_list args; 772ac27a0ecSDave Kleikamp 773b03a2f7eSAndreas Dilger if (!ext4_warning_ratelimit(sb)) 774efbed4dcSTheodore Ts'o return; 775efbed4dcSTheodore Ts'o 776ac27a0ecSDave Kleikamp va_start(args, fmt); 7770ff2ea7dSJoe Perches vaf.fmt = fmt; 7780ff2ea7dSJoe Perches vaf.va = &args; 7790ff2ea7dSJoe Perches printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 7800ff2ea7dSJoe Perches sb->s_id, function, line, &vaf); 781ac27a0ecSDave Kleikamp va_end(args); 782ac27a0ecSDave Kleikamp } 783ac27a0ecSDave Kleikamp 784b03a2f7eSAndreas Dilger void __ext4_warning_inode(const struct inode *inode, const char *function, 785b03a2f7eSAndreas Dilger unsigned int line, const char *fmt, ...) 786b03a2f7eSAndreas Dilger { 787b03a2f7eSAndreas Dilger struct va_format vaf; 788b03a2f7eSAndreas Dilger va_list args; 789b03a2f7eSAndreas Dilger 790b03a2f7eSAndreas Dilger if (!ext4_warning_ratelimit(inode->i_sb)) 791b03a2f7eSAndreas Dilger return; 792b03a2f7eSAndreas Dilger 793b03a2f7eSAndreas Dilger va_start(args, fmt); 794b03a2f7eSAndreas Dilger vaf.fmt = fmt; 795b03a2f7eSAndreas Dilger vaf.va = &args; 796b03a2f7eSAndreas Dilger printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " 797b03a2f7eSAndreas Dilger "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, 798b03a2f7eSAndreas Dilger function, line, inode->i_ino, current->comm, &vaf); 799b03a2f7eSAndreas Dilger va_end(args); 800b03a2f7eSAndreas Dilger } 801b03a2f7eSAndreas Dilger 802e29136f8STheodore Ts'o void __ext4_grp_locked_error(const char *function, unsigned int line, 803e29136f8STheodore Ts'o struct super_block *sb, ext4_group_t grp, 804e29136f8STheodore Ts'o unsigned long ino, ext4_fsblk_t block, 805e29136f8STheodore Ts'o const char *fmt, ...) 8065d1b1b3fSAneesh Kumar K.V __releases(bitlock) 8075d1b1b3fSAneesh Kumar K.V __acquires(bitlock) 8085d1b1b3fSAneesh Kumar K.V { 8090ff2ea7dSJoe Perches struct va_format vaf; 8105d1b1b3fSAneesh Kumar K.V va_list args; 8115d1b1b3fSAneesh Kumar K.V 8120db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 8130db1ff22STheodore Ts'o return; 8140db1ff22STheodore Ts'o 815ccf0f32aSTheodore Ts'o trace_ext4_error(sb, function, line); 81654d3adbcSTheodore Ts'o __save_error_info(sb, EFSCORRUPTED, ino, block, function, line); 8170ff2ea7dSJoe Perches 818efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 8195d1b1b3fSAneesh Kumar K.V va_start(args, fmt); 8200ff2ea7dSJoe Perches vaf.fmt = fmt; 8210ff2ea7dSJoe Perches vaf.va = &args; 82221149d61SRobin Dong printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 823e29136f8STheodore Ts'o sb->s_id, function, line, grp); 824e29136f8STheodore Ts'o if (ino) 8250ff2ea7dSJoe Perches printk(KERN_CONT "inode %lu: ", ino); 826e29136f8STheodore Ts'o if (block) 827efbed4dcSTheodore Ts'o printk(KERN_CONT "block %llu:", 828efbed4dcSTheodore Ts'o (unsigned long long) block); 8290ff2ea7dSJoe Perches printk(KERN_CONT "%pV\n", &vaf); 8305d1b1b3fSAneesh Kumar K.V va_end(args); 831efbed4dcSTheodore Ts'o } 8325d1b1b3fSAneesh Kumar K.V 833327eaf73STheodore Ts'o if (test_opt(sb, WARN_ON_ERROR)) 834327eaf73STheodore Ts'o WARN_ON_ONCE(1); 835327eaf73STheodore Ts'o 8365d1b1b3fSAneesh Kumar K.V if (test_opt(sb, ERRORS_CONT)) { 837e2d67052STheodore Ts'o ext4_commit_super(sb, 0); 8385d1b1b3fSAneesh Kumar K.V return; 8395d1b1b3fSAneesh Kumar K.V } 8401c13d5c0STheodore Ts'o 8415d1b1b3fSAneesh Kumar K.V ext4_unlock_group(sb, grp); 84206f29cc8SZhouyi Zhou ext4_commit_super(sb, 1); 8435d1b1b3fSAneesh Kumar K.V ext4_handle_error(sb); 8445d1b1b3fSAneesh Kumar K.V /* 8455d1b1b3fSAneesh Kumar K.V * We only get here in the ERRORS_RO case; relocking the group 8465d1b1b3fSAneesh Kumar K.V * may be dangerous, but nothing bad will happen since the 8475d1b1b3fSAneesh Kumar K.V * filesystem will have already been marked read/only and the 8485d1b1b3fSAneesh Kumar K.V * journal has been aborted. We return 1 as a hint to callers 8495d1b1b3fSAneesh Kumar K.V * who might what to use the return value from 85025985edcSLucas De Marchi * ext4_grp_locked_error() to distinguish between the 8515d1b1b3fSAneesh Kumar K.V * ERRORS_CONT and ERRORS_RO case, and perhaps return more 8525d1b1b3fSAneesh Kumar K.V * aggressively from the ext4 function in question, with a 8535d1b1b3fSAneesh Kumar K.V * more appropriate error code. 8545d1b1b3fSAneesh Kumar K.V */ 8555d1b1b3fSAneesh Kumar K.V ext4_lock_group(sb, grp); 8565d1b1b3fSAneesh Kumar K.V return; 8575d1b1b3fSAneesh Kumar K.V } 8585d1b1b3fSAneesh Kumar K.V 859db79e6d1SWang Shilong void ext4_mark_group_bitmap_corrupted(struct super_block *sb, 860db79e6d1SWang Shilong ext4_group_t group, 861db79e6d1SWang Shilong unsigned int flags) 862db79e6d1SWang Shilong { 863db79e6d1SWang Shilong struct ext4_sb_info *sbi = EXT4_SB(sb); 864db79e6d1SWang Shilong struct ext4_group_info *grp = ext4_get_group_info(sb, group); 865db79e6d1SWang Shilong struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); 8669af0b3d1SWang Shilong int ret; 867db79e6d1SWang Shilong 8689af0b3d1SWang Shilong if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) { 8699af0b3d1SWang Shilong ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, 8709af0b3d1SWang Shilong &grp->bb_state); 8719af0b3d1SWang Shilong if (!ret) 872db79e6d1SWang Shilong percpu_counter_sub(&sbi->s_freeclusters_counter, 873db79e6d1SWang Shilong grp->bb_free); 874db79e6d1SWang Shilong } 875db79e6d1SWang Shilong 8769af0b3d1SWang Shilong if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) { 8779af0b3d1SWang Shilong ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, 8789af0b3d1SWang Shilong &grp->bb_state); 8799af0b3d1SWang Shilong if (!ret && gdp) { 880db79e6d1SWang Shilong int count; 881db79e6d1SWang Shilong 882db79e6d1SWang Shilong count = ext4_free_inodes_count(sb, gdp); 883db79e6d1SWang Shilong percpu_counter_sub(&sbi->s_freeinodes_counter, 884db79e6d1SWang Shilong count); 885db79e6d1SWang Shilong } 886db79e6d1SWang Shilong } 887db79e6d1SWang Shilong } 888db79e6d1SWang Shilong 889617ba13bSMingming Cao void ext4_update_dynamic_rev(struct super_block *sb) 890ac27a0ecSDave Kleikamp { 891617ba13bSMingming Cao struct ext4_super_block *es = EXT4_SB(sb)->s_es; 892ac27a0ecSDave Kleikamp 893617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 894ac27a0ecSDave Kleikamp return; 895ac27a0ecSDave Kleikamp 89612062dddSEric Sandeen ext4_warning(sb, 897ac27a0ecSDave Kleikamp "updating to rev %d because of new feature flag, " 898ac27a0ecSDave Kleikamp "running e2fsck is recommended", 899617ba13bSMingming Cao EXT4_DYNAMIC_REV); 900ac27a0ecSDave Kleikamp 901617ba13bSMingming Cao es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 902617ba13bSMingming Cao es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 903617ba13bSMingming Cao es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 904ac27a0ecSDave Kleikamp /* leave es->s_feature_*compat flags alone */ 905ac27a0ecSDave Kleikamp /* es->s_uuid will be set by e2fsck if empty */ 906ac27a0ecSDave Kleikamp 907ac27a0ecSDave Kleikamp /* 908ac27a0ecSDave Kleikamp * The rest of the superblock fields should be zero, and if not it 909ac27a0ecSDave Kleikamp * means they are likely already in use, so leave them alone. We 910ac27a0ecSDave Kleikamp * can leave it up to e2fsck to clean up any inconsistencies there. 911ac27a0ecSDave Kleikamp */ 912ac27a0ecSDave Kleikamp } 913ac27a0ecSDave Kleikamp 914ac27a0ecSDave Kleikamp /* 915ac27a0ecSDave Kleikamp * Open the external journal device 916ac27a0ecSDave Kleikamp */ 917b31e1552SEric Sandeen static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 918ac27a0ecSDave Kleikamp { 919ac27a0ecSDave Kleikamp struct block_device *bdev; 920ac27a0ecSDave Kleikamp 921d4d77629STejun Heo bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 922ac27a0ecSDave Kleikamp if (IS_ERR(bdev)) 923ac27a0ecSDave Kleikamp goto fail; 924ac27a0ecSDave Kleikamp return bdev; 925ac27a0ecSDave Kleikamp 926ac27a0ecSDave Kleikamp fail: 927ea3edd4dSChristoph Hellwig ext4_msg(sb, KERN_ERR, 928ea3edd4dSChristoph Hellwig "failed to open journal device unknown-block(%u,%u) %ld", 929ea3edd4dSChristoph Hellwig MAJOR(dev), MINOR(dev), PTR_ERR(bdev)); 930ac27a0ecSDave Kleikamp return NULL; 931ac27a0ecSDave Kleikamp } 932ac27a0ecSDave Kleikamp 933ac27a0ecSDave Kleikamp /* 934ac27a0ecSDave Kleikamp * Release the journal device 935ac27a0ecSDave Kleikamp */ 9364385bab1SAl Viro static void ext4_blkdev_put(struct block_device *bdev) 937ac27a0ecSDave Kleikamp { 9384385bab1SAl Viro blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 939ac27a0ecSDave Kleikamp } 940ac27a0ecSDave Kleikamp 9414385bab1SAl Viro static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 942ac27a0ecSDave Kleikamp { 943ac27a0ecSDave Kleikamp struct block_device *bdev; 944ac27a0ecSDave Kleikamp bdev = sbi->journal_bdev; 945ac27a0ecSDave Kleikamp if (bdev) { 9464385bab1SAl Viro ext4_blkdev_put(bdev); 947ac27a0ecSDave Kleikamp sbi->journal_bdev = NULL; 948ac27a0ecSDave Kleikamp } 949ac27a0ecSDave Kleikamp } 950ac27a0ecSDave Kleikamp 951ac27a0ecSDave Kleikamp static inline struct inode *orphan_list_entry(struct list_head *l) 952ac27a0ecSDave Kleikamp { 953617ba13bSMingming Cao return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 954ac27a0ecSDave Kleikamp } 955ac27a0ecSDave Kleikamp 956617ba13bSMingming Cao static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 957ac27a0ecSDave Kleikamp { 958ac27a0ecSDave Kleikamp struct list_head *l; 959ac27a0ecSDave Kleikamp 960b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 961ac27a0ecSDave Kleikamp le32_to_cpu(sbi->s_es->s_last_orphan)); 962ac27a0ecSDave Kleikamp 963ac27a0ecSDave Kleikamp printk(KERN_ERR "sb_info orphan list:\n"); 964ac27a0ecSDave Kleikamp list_for_each(l, &sbi->s_orphan) { 965ac27a0ecSDave Kleikamp struct inode *inode = orphan_list_entry(l); 966ac27a0ecSDave Kleikamp printk(KERN_ERR " " 967ac27a0ecSDave Kleikamp "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 968ac27a0ecSDave Kleikamp inode->i_sb->s_id, inode->i_ino, inode, 969ac27a0ecSDave Kleikamp inode->i_mode, inode->i_nlink, 970ac27a0ecSDave Kleikamp NEXT_ORPHAN(inode)); 971ac27a0ecSDave Kleikamp } 972ac27a0ecSDave Kleikamp } 973ac27a0ecSDave Kleikamp 974957153fcSJan Kara #ifdef CONFIG_QUOTA 975957153fcSJan Kara static int ext4_quota_off(struct super_block *sb, int type); 976957153fcSJan Kara 977957153fcSJan Kara static inline void ext4_quota_off_umount(struct super_block *sb) 978957153fcSJan Kara { 979957153fcSJan Kara int type; 980957153fcSJan Kara 981957153fcSJan Kara /* Use our quota_off function to clear inode flags etc. */ 982957153fcSJan Kara for (type = 0; type < EXT4_MAXQUOTAS; type++) 983957153fcSJan Kara ext4_quota_off(sb, type); 984957153fcSJan Kara } 98533458eabSTheodore Ts'o 98633458eabSTheodore Ts'o /* 98733458eabSTheodore Ts'o * This is a helper function which is used in the mount/remount 98833458eabSTheodore Ts'o * codepaths (which holds s_umount) to fetch the quota file name. 98933458eabSTheodore Ts'o */ 99033458eabSTheodore Ts'o static inline char *get_qf_name(struct super_block *sb, 99133458eabSTheodore Ts'o struct ext4_sb_info *sbi, 99233458eabSTheodore Ts'o int type) 99333458eabSTheodore Ts'o { 99433458eabSTheodore Ts'o return rcu_dereference_protected(sbi->s_qf_names[type], 99533458eabSTheodore Ts'o lockdep_is_held(&sb->s_umount)); 99633458eabSTheodore Ts'o } 997957153fcSJan Kara #else 998957153fcSJan Kara static inline void ext4_quota_off_umount(struct super_block *sb) 999957153fcSJan Kara { 1000957153fcSJan Kara } 1001957153fcSJan Kara #endif 1002957153fcSJan Kara 1003617ba13bSMingming Cao static void ext4_put_super(struct super_block *sb) 1004ac27a0ecSDave Kleikamp { 1005617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 1006617ba13bSMingming Cao struct ext4_super_block *es = sbi->s_es; 10071d0c3924STheodore Ts'o struct buffer_head **group_desc; 10087c990728SSuraj Jitindar Singh struct flex_groups **flex_groups; 100997abd7d4STheodore Ts'o int aborted = 0; 1010ef2cabf7SHidehiro Kawai int i, err; 1011ac27a0ecSDave Kleikamp 1012857ac889SLukas Czerner ext4_unregister_li_request(sb); 1013957153fcSJan Kara ext4_quota_off_umount(sb); 1014e0ccfd95SChristoph Hellwig 10152e8fa54eSJan Kara destroy_workqueue(sbi->rsv_conversion_wq); 10164c0425ffSMingming Cao 10175e47868fSRitesh Harjani /* 10185e47868fSRitesh Harjani * Unregister sysfs before destroying jbd2 journal. 10195e47868fSRitesh Harjani * Since we could still access attr_journal_task attribute via sysfs 10205e47868fSRitesh Harjani * path which could have sbi->s_journal->j_task as NULL 10215e47868fSRitesh Harjani */ 10225e47868fSRitesh Harjani ext4_unregister_sysfs(sb); 10235e47868fSRitesh Harjani 10240390131bSFrank Mayhar if (sbi->s_journal) { 102597abd7d4STheodore Ts'o aborted = is_journal_aborted(sbi->s_journal); 1026ef2cabf7SHidehiro Kawai err = jbd2_journal_destroy(sbi->s_journal); 102747b4a50bSJan Kara sbi->s_journal = NULL; 1028878520acSTheodore Ts'o if ((err < 0) && !aborted) { 102954d3adbcSTheodore Ts'o ext4_abort(sb, -err, "Couldn't clean up the journal"); 10300390131bSFrank Mayhar } 1031878520acSTheodore Ts'o } 1032d4edac31SJosef Bacik 1033d3922a77SZheng Liu ext4_es_unregister_shrinker(sbi); 10349105bb14SAl Viro del_timer_sync(&sbi->s_err_report); 1035d4edac31SJosef Bacik ext4_release_system_zone(sb); 1036d4edac31SJosef Bacik ext4_mb_release(sb); 1037d4edac31SJosef Bacik ext4_ext_release(sb); 1038d4edac31SJosef Bacik 1039bc98a42cSDavid Howells if (!sb_rdonly(sb) && !aborted) { 1040e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 1041ac27a0ecSDave Kleikamp es->s_state = cpu_to_le16(sbi->s_mount_state); 1042ac27a0ecSDave Kleikamp } 1043bc98a42cSDavid Howells if (!sb_rdonly(sb)) 1044a8e25a83SArtem Bityutskiy ext4_commit_super(sb, 1); 1045a8e25a83SArtem Bityutskiy 10461d0c3924STheodore Ts'o rcu_read_lock(); 10471d0c3924STheodore Ts'o group_desc = rcu_dereference(sbi->s_group_desc); 1048ac27a0ecSDave Kleikamp for (i = 0; i < sbi->s_gdb_count; i++) 10491d0c3924STheodore Ts'o brelse(group_desc[i]); 10501d0c3924STheodore Ts'o kvfree(group_desc); 10517c990728SSuraj Jitindar Singh flex_groups = rcu_dereference(sbi->s_flex_groups); 10527c990728SSuraj Jitindar Singh if (flex_groups) { 10537c990728SSuraj Jitindar Singh for (i = 0; i < sbi->s_flex_groups_allocated; i++) 10547c990728SSuraj Jitindar Singh kvfree(flex_groups[i]); 10557c990728SSuraj Jitindar Singh kvfree(flex_groups); 10567c990728SSuraj Jitindar Singh } 10571d0c3924STheodore Ts'o rcu_read_unlock(); 105857042651STheodore Ts'o percpu_counter_destroy(&sbi->s_freeclusters_counter); 1059ac27a0ecSDave Kleikamp percpu_counter_destroy(&sbi->s_freeinodes_counter); 1060ac27a0ecSDave Kleikamp percpu_counter_destroy(&sbi->s_dirs_counter); 106157042651STheodore Ts'o percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 1062bbd55937SEric Biggers percpu_free_rwsem(&sbi->s_writepages_rwsem); 1063ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1064a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 106533458eabSTheodore Ts'o kfree(get_qf_name(sb, sbi, i)); 1066ac27a0ecSDave Kleikamp #endif 1067ac27a0ecSDave Kleikamp 1068ac27a0ecSDave Kleikamp /* Debugging code just in case the in-memory inode orphan list 1069ac27a0ecSDave Kleikamp * isn't empty. The on-disk one can be non-empty if we've 1070ac27a0ecSDave Kleikamp * detected an error and taken the fs readonly, but the 1071ac27a0ecSDave Kleikamp * in-memory list had better be clean by this point. */ 1072ac27a0ecSDave Kleikamp if (!list_empty(&sbi->s_orphan)) 1073ac27a0ecSDave Kleikamp dump_orphan_list(sb, sbi); 1074ac27a0ecSDave Kleikamp J_ASSERT(list_empty(&sbi->s_orphan)); 1075ac27a0ecSDave Kleikamp 107689d96a6fSTheodore Ts'o sync_blockdev(sb->s_bdev); 1077f98393a6SPeter Zijlstra invalidate_bdev(sb->s_bdev); 1078ac27a0ecSDave Kleikamp if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { 1079ac27a0ecSDave Kleikamp /* 1080ac27a0ecSDave Kleikamp * Invalidate the journal device's buffers. We don't want them 1081ac27a0ecSDave Kleikamp * floating about in memory - the physical journal device may 1082ac27a0ecSDave Kleikamp * hotswapped, and it breaks the `ro-after' testing code. 1083ac27a0ecSDave Kleikamp */ 1084ac27a0ecSDave Kleikamp sync_blockdev(sbi->journal_bdev); 1085f98393a6SPeter Zijlstra invalidate_bdev(sbi->journal_bdev); 1086617ba13bSMingming Cao ext4_blkdev_remove(sbi); 1087ac27a0ecSDave Kleikamp } 108850c15df6SChengguang Xu 1089dec214d0STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 1090dec214d0STahsin Erdogan sbi->s_ea_inode_cache = NULL; 109150c15df6SChengguang Xu 109247387409STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 109347387409STahsin Erdogan sbi->s_ea_block_cache = NULL; 109450c15df6SChengguang Xu 1095c5e06d10SJohann Lombardi if (sbi->s_mmp_tsk) 1096c5e06d10SJohann Lombardi kthread_stop(sbi->s_mmp_tsk); 10979060dd2cSEric Sandeen brelse(sbi->s_sbh); 1098ac27a0ecSDave Kleikamp sb->s_fs_info = NULL; 10993197ebdbSTheodore Ts'o /* 11003197ebdbSTheodore Ts'o * Now that we are completely done shutting down the 11013197ebdbSTheodore Ts'o * superblock, we need to actually destroy the kobject. 11023197ebdbSTheodore Ts'o */ 11033197ebdbSTheodore Ts'o kobject_put(&sbi->s_kobj); 11043197ebdbSTheodore Ts'o wait_for_completion(&sbi->s_kobj_unregister); 11050441984aSDarrick J. Wong if (sbi->s_chksum_driver) 11060441984aSDarrick J. Wong crypto_free_shash(sbi->s_chksum_driver); 1107705895b6SPekka Enberg kfree(sbi->s_blockgroup_lock); 11085e405595SDan Williams fs_put_dax(sbi->s_daxdev); 1109c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 1110c83ad55eSGabriel Krisman Bertazi utf8_unload(sbi->s_encoding); 1111c83ad55eSGabriel Krisman Bertazi #endif 1112ac27a0ecSDave Kleikamp kfree(sbi); 1113ac27a0ecSDave Kleikamp } 1114ac27a0ecSDave Kleikamp 1115e18b890bSChristoph Lameter static struct kmem_cache *ext4_inode_cachep; 1116ac27a0ecSDave Kleikamp 1117ac27a0ecSDave Kleikamp /* 1118ac27a0ecSDave Kleikamp * Called inside transaction, so use GFP_NOFS 1119ac27a0ecSDave Kleikamp */ 1120617ba13bSMingming Cao static struct inode *ext4_alloc_inode(struct super_block *sb) 1121ac27a0ecSDave Kleikamp { 1122617ba13bSMingming Cao struct ext4_inode_info *ei; 1123ac27a0ecSDave Kleikamp 1124e6b4f8daSChristoph Lameter ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 1125ac27a0ecSDave Kleikamp if (!ei) 1126ac27a0ecSDave Kleikamp return NULL; 11270b8e58a1SAndreas Dilger 1128ee73f9a5SJeff Layton inode_set_iversion(&ei->vfs_inode, 1); 1129202ee5dfSTheodore Ts'o spin_lock_init(&ei->i_raw_lock); 1130c9de560dSAlex Tomas INIT_LIST_HEAD(&ei->i_prealloc_list); 1131c9de560dSAlex Tomas spin_lock_init(&ei->i_prealloc_lock); 11329a26b661SZheng Liu ext4_es_init_tree(&ei->i_es_tree); 11339a26b661SZheng Liu rwlock_init(&ei->i_es_lock); 1134edaa53caSZheng Liu INIT_LIST_HEAD(&ei->i_es_list); 1135eb68d0e2SZheng Liu ei->i_es_all_nr = 0; 1136edaa53caSZheng Liu ei->i_es_shk_nr = 0; 1137dd475925SJan Kara ei->i_es_shrink_lblk = 0; 1138d2a17637SMingming Cao ei->i_reserved_data_blocks = 0; 1139d2a17637SMingming Cao spin_lock_init(&(ei->i_block_reservation_lock)); 11401dc0aa46SEric Whitney ext4_init_pending_tree(&ei->i_pending_tree); 1141a9e7f447SDmitry Monakhov #ifdef CONFIG_QUOTA 1142a9e7f447SDmitry Monakhov ei->i_reserved_quota = 0; 114396c7e0d9SJan Kara memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 1144a9e7f447SDmitry Monakhov #endif 11458aefcd55STheodore Ts'o ei->jinode = NULL; 11462e8fa54eSJan Kara INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 1147744692dcSJiaying Zhang spin_lock_init(&ei->i_completed_io_lock); 1148b436b9beSJan Kara ei->i_sync_tid = 0; 1149b436b9beSJan Kara ei->i_datasync_tid = 0; 1150e27f41e1SDmitry Monakhov atomic_set(&ei->i_unwritten, 0); 11512e8fa54eSJan Kara INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 1152ac27a0ecSDave Kleikamp return &ei->vfs_inode; 1153ac27a0ecSDave Kleikamp } 1154ac27a0ecSDave Kleikamp 11557ff9c073STheodore Ts'o static int ext4_drop_inode(struct inode *inode) 11567ff9c073STheodore Ts'o { 11577ff9c073STheodore Ts'o int drop = generic_drop_inode(inode); 11587ff9c073STheodore Ts'o 115929b3692eSEric Biggers if (!drop) 116029b3692eSEric Biggers drop = fscrypt_drop_inode(inode); 116129b3692eSEric Biggers 11627ff9c073STheodore Ts'o trace_ext4_drop_inode(inode, drop); 11637ff9c073STheodore Ts'o return drop; 11647ff9c073STheodore Ts'o } 11657ff9c073STheodore Ts'o 116694053139SAl Viro static void ext4_free_in_core_inode(struct inode *inode) 1167fa0d7e3dSNick Piggin { 11682c58d548SEric Biggers fscrypt_free_inode(inode); 1169fa0d7e3dSNick Piggin kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 1170fa0d7e3dSNick Piggin } 1171fa0d7e3dSNick Piggin 1172617ba13bSMingming Cao static void ext4_destroy_inode(struct inode *inode) 1173ac27a0ecSDave Kleikamp { 11749f7dd93dSVasily Averin if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 1175b31e1552SEric Sandeen ext4_msg(inode->i_sb, KERN_ERR, 1176b31e1552SEric Sandeen "Inode %lu (%p): orphan list check failed!", 1177b31e1552SEric Sandeen inode->i_ino, EXT4_I(inode)); 11789f7dd93dSVasily Averin print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 11799f7dd93dSVasily Averin EXT4_I(inode), sizeof(struct ext4_inode_info), 11809f7dd93dSVasily Averin true); 11819f7dd93dSVasily Averin dump_stack(); 11829f7dd93dSVasily Averin } 1183ac27a0ecSDave Kleikamp } 1184ac27a0ecSDave Kleikamp 118551cc5068SAlexey Dobriyan static void init_once(void *foo) 1186ac27a0ecSDave Kleikamp { 1187617ba13bSMingming Cao struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 1188ac27a0ecSDave Kleikamp 1189ac27a0ecSDave Kleikamp INIT_LIST_HEAD(&ei->i_orphan); 1190ac27a0ecSDave Kleikamp init_rwsem(&ei->xattr_sem); 11910e855ac8SAneesh Kumar K.V init_rwsem(&ei->i_data_sem); 1192ea3d7209SJan Kara init_rwsem(&ei->i_mmap_sem); 1193ac27a0ecSDave Kleikamp inode_init_once(&ei->vfs_inode); 1194ac27a0ecSDave Kleikamp } 1195ac27a0ecSDave Kleikamp 1196e67bc2b3SFabian Frederick static int __init init_inodecache(void) 1197ac27a0ecSDave Kleikamp { 1198f8dd7c70SDavid Windsor ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache", 1199f8dd7c70SDavid Windsor sizeof(struct ext4_inode_info), 0, 1200f8dd7c70SDavid Windsor (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD| 1201f8dd7c70SDavid Windsor SLAB_ACCOUNT), 1202f8dd7c70SDavid Windsor offsetof(struct ext4_inode_info, i_data), 1203f8dd7c70SDavid Windsor sizeof_field(struct ext4_inode_info, i_data), 120420c2df83SPaul Mundt init_once); 1205617ba13bSMingming Cao if (ext4_inode_cachep == NULL) 1206ac27a0ecSDave Kleikamp return -ENOMEM; 1207ac27a0ecSDave Kleikamp return 0; 1208ac27a0ecSDave Kleikamp } 1209ac27a0ecSDave Kleikamp 1210ac27a0ecSDave Kleikamp static void destroy_inodecache(void) 1211ac27a0ecSDave Kleikamp { 12128c0a8537SKirill A. Shutemov /* 12138c0a8537SKirill A. Shutemov * Make sure all delayed rcu free inodes are flushed before we 12148c0a8537SKirill A. Shutemov * destroy cache. 12158c0a8537SKirill A. Shutemov */ 12168c0a8537SKirill A. Shutemov rcu_barrier(); 1217617ba13bSMingming Cao kmem_cache_destroy(ext4_inode_cachep); 1218ac27a0ecSDave Kleikamp } 1219ac27a0ecSDave Kleikamp 12200930fcc1SAl Viro void ext4_clear_inode(struct inode *inode) 1221ac27a0ecSDave Kleikamp { 12220930fcc1SAl Viro invalidate_inode_buffers(inode); 1223dbd5768fSJan Kara clear_inode(inode); 1224c2ea3fdeSTheodore Ts'o ext4_discard_preallocations(inode); 122551865fdaSZheng Liu ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1226f4c2d372SJan Kara dquot_drop(inode); 12278aefcd55STheodore Ts'o if (EXT4_I(inode)->jinode) { 12288aefcd55STheodore Ts'o jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 12298aefcd55STheodore Ts'o EXT4_I(inode)->jinode); 12308aefcd55STheodore Ts'o jbd2_free_inode(EXT4_I(inode)->jinode); 12318aefcd55STheodore Ts'o EXT4_I(inode)->jinode = NULL; 12328aefcd55STheodore Ts'o } 12333d204e24SEric Biggers fscrypt_put_encryption_info(inode); 1234c93d8f88SEric Biggers fsverity_cleanup_inode(inode); 1235ac27a0ecSDave Kleikamp } 1236ac27a0ecSDave Kleikamp 12371b961ac0SChristoph Hellwig static struct inode *ext4_nfs_get_inode(struct super_block *sb, 12381b961ac0SChristoph Hellwig u64 ino, u32 generation) 1239ac27a0ecSDave Kleikamp { 1240ac27a0ecSDave Kleikamp struct inode *inode; 1241ac27a0ecSDave Kleikamp 12428a363970STheodore Ts'o /* 1243ac27a0ecSDave Kleikamp * Currently we don't know the generation for parent directory, so 1244ac27a0ecSDave Kleikamp * a generation of 0 means "accept any" 1245ac27a0ecSDave Kleikamp */ 12468a363970STheodore Ts'o inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); 12471d1fe1eeSDavid Howells if (IS_ERR(inode)) 12481d1fe1eeSDavid Howells return ERR_CAST(inode); 12491d1fe1eeSDavid Howells if (generation && inode->i_generation != generation) { 1250ac27a0ecSDave Kleikamp iput(inode); 1251ac27a0ecSDave Kleikamp return ERR_PTR(-ESTALE); 1252ac27a0ecSDave Kleikamp } 12531b961ac0SChristoph Hellwig 12541b961ac0SChristoph Hellwig return inode; 1255ac27a0ecSDave Kleikamp } 12561b961ac0SChristoph Hellwig 12571b961ac0SChristoph Hellwig static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 12581b961ac0SChristoph Hellwig int fh_len, int fh_type) 12591b961ac0SChristoph Hellwig { 12601b961ac0SChristoph Hellwig return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 12611b961ac0SChristoph Hellwig ext4_nfs_get_inode); 12621b961ac0SChristoph Hellwig } 12631b961ac0SChristoph Hellwig 12641b961ac0SChristoph Hellwig static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 12651b961ac0SChristoph Hellwig int fh_len, int fh_type) 12661b961ac0SChristoph Hellwig { 12671b961ac0SChristoph Hellwig return generic_fh_to_parent(sb, fid, fh_len, fh_type, 12681b961ac0SChristoph Hellwig ext4_nfs_get_inode); 1269ac27a0ecSDave Kleikamp } 1270ac27a0ecSDave Kleikamp 1271fde87268STheodore Ts'o static int ext4_nfs_commit_metadata(struct inode *inode) 1272fde87268STheodore Ts'o { 1273fde87268STheodore Ts'o struct writeback_control wbc = { 1274fde87268STheodore Ts'o .sync_mode = WB_SYNC_ALL 1275fde87268STheodore Ts'o }; 1276fde87268STheodore Ts'o 1277fde87268STheodore Ts'o trace_ext4_nfs_commit_metadata(inode); 1278fde87268STheodore Ts'o return ext4_write_inode(inode, &wbc); 1279fde87268STheodore Ts'o } 1280fde87268STheodore Ts'o 1281c39a7f84SToshiyuki Okajima /* 1282c39a7f84SToshiyuki Okajima * Try to release metadata pages (indirect blocks, directories) which are 1283c39a7f84SToshiyuki Okajima * mapped via the block device. Since these pages could have journal heads 1284c39a7f84SToshiyuki Okajima * which would prevent try_to_free_buffers() from freeing them, we must use 1285c39a7f84SToshiyuki Okajima * jbd2 layer's try_to_free_buffers() function to release them. 1286c39a7f84SToshiyuki Okajima */ 12870b8e58a1SAndreas Dilger static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 12880b8e58a1SAndreas Dilger gfp_t wait) 1289c39a7f84SToshiyuki Okajima { 1290c39a7f84SToshiyuki Okajima journal_t *journal = EXT4_SB(sb)->s_journal; 1291c39a7f84SToshiyuki Okajima 1292c39a7f84SToshiyuki Okajima WARN_ON(PageChecked(page)); 1293c39a7f84SToshiyuki Okajima if (!page_has_buffers(page)) 1294c39a7f84SToshiyuki Okajima return 0; 1295c39a7f84SToshiyuki Okajima if (journal) 1296c39a7f84SToshiyuki Okajima return jbd2_journal_try_to_free_buffers(journal, page, 1297d0164adcSMel Gorman wait & ~__GFP_DIRECT_RECLAIM); 1298c39a7f84SToshiyuki Okajima return try_to_free_buffers(page); 1299c39a7f84SToshiyuki Okajima } 1300c39a7f84SToshiyuki Okajima 1301643fa961SChandan Rajendra #ifdef CONFIG_FS_ENCRYPTION 1302a7550b30SJaegeuk Kim static int ext4_get_context(struct inode *inode, void *ctx, size_t len) 1303a7550b30SJaegeuk Kim { 1304a7550b30SJaegeuk Kim return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION, 1305a7550b30SJaegeuk Kim EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len); 1306a7550b30SJaegeuk Kim } 1307a7550b30SJaegeuk Kim 1308a7550b30SJaegeuk Kim static int ext4_set_context(struct inode *inode, const void *ctx, size_t len, 1309a7550b30SJaegeuk Kim void *fs_data) 1310a7550b30SJaegeuk Kim { 13112f8f5e76SEric Biggers handle_t *handle = fs_data; 1312c1a5d5f6STahsin Erdogan int res, res2, credits, retries = 0; 1313a7550b30SJaegeuk Kim 13149ce0151aSEric Biggers /* 13159ce0151aSEric Biggers * Encrypting the root directory is not allowed because e2fsck expects 13169ce0151aSEric Biggers * lost+found to exist and be unencrypted, and encrypting the root 13179ce0151aSEric Biggers * directory would imply encrypting the lost+found directory as well as 13189ce0151aSEric Biggers * the filename "lost+found" itself. 13199ce0151aSEric Biggers */ 13209ce0151aSEric Biggers if (inode->i_ino == EXT4_ROOT_INO) 13219ce0151aSEric Biggers return -EPERM; 1322a7550b30SJaegeuk Kim 13237d3e06a8SRoss Zwisler if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode))) 13247d3e06a8SRoss Zwisler return -EINVAL; 13257d3e06a8SRoss Zwisler 132694840e3cSEric Biggers res = ext4_convert_inline_data(inode); 132794840e3cSEric Biggers if (res) 132894840e3cSEric Biggers return res; 132994840e3cSEric Biggers 13302f8f5e76SEric Biggers /* 13312f8f5e76SEric Biggers * If a journal handle was specified, then the encryption context is 13322f8f5e76SEric Biggers * being set on a new inode via inheritance and is part of a larger 13332f8f5e76SEric Biggers * transaction to create the inode. Otherwise the encryption context is 13342f8f5e76SEric Biggers * being set on an existing inode in its own transaction. Only in the 13352f8f5e76SEric Biggers * latter case should the "retry on ENOSPC" logic be used. 13362f8f5e76SEric Biggers */ 13372f8f5e76SEric Biggers 13382f8f5e76SEric Biggers if (handle) { 13392f8f5e76SEric Biggers res = ext4_xattr_set_handle(handle, inode, 13402f8f5e76SEric Biggers EXT4_XATTR_INDEX_ENCRYPTION, 13412f8f5e76SEric Biggers EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 13422f8f5e76SEric Biggers ctx, len, 0); 1343a7550b30SJaegeuk Kim if (!res) { 1344a7550b30SJaegeuk Kim ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1345a7550b30SJaegeuk Kim ext4_clear_inode_state(inode, 1346a7550b30SJaegeuk Kim EXT4_STATE_MAY_INLINE_DATA); 1347a3caa24bSJan Kara /* 13482ee6a576SEric Biggers * Update inode->i_flags - S_ENCRYPTED will be enabled, 13492ee6a576SEric Biggers * S_DAX may be disabled 1350a3caa24bSJan Kara */ 1351043546e4SIra Weiny ext4_set_inode_flags(inode, false); 1352a7550b30SJaegeuk Kim } 1353a7550b30SJaegeuk Kim return res; 1354a7550b30SJaegeuk Kim } 1355a7550b30SJaegeuk Kim 1356b8cb5a54STahsin Erdogan res = dquot_initialize(inode); 1357b8cb5a54STahsin Erdogan if (res) 1358b8cb5a54STahsin Erdogan return res; 13592f8f5e76SEric Biggers retry: 1360af65207cSTahsin Erdogan res = ext4_xattr_set_credits(inode, len, false /* is_create */, 1361af65207cSTahsin Erdogan &credits); 1362dec214d0STahsin Erdogan if (res) 1363dec214d0STahsin Erdogan return res; 1364dec214d0STahsin Erdogan 1365c1a5d5f6STahsin Erdogan handle = ext4_journal_start(inode, EXT4_HT_MISC, credits); 1366a7550b30SJaegeuk Kim if (IS_ERR(handle)) 1367a7550b30SJaegeuk Kim return PTR_ERR(handle); 1368a7550b30SJaegeuk Kim 13692f8f5e76SEric Biggers res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION, 13702f8f5e76SEric Biggers EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 13712f8f5e76SEric Biggers ctx, len, 0); 1372a7550b30SJaegeuk Kim if (!res) { 1373a7550b30SJaegeuk Kim ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 13742ee6a576SEric Biggers /* 13752ee6a576SEric Biggers * Update inode->i_flags - S_ENCRYPTED will be enabled, 13762ee6a576SEric Biggers * S_DAX may be disabled 13772ee6a576SEric Biggers */ 1378043546e4SIra Weiny ext4_set_inode_flags(inode, false); 1379a7550b30SJaegeuk Kim res = ext4_mark_inode_dirty(handle, inode); 1380a7550b30SJaegeuk Kim if (res) 1381a7550b30SJaegeuk Kim EXT4_ERROR_INODE(inode, "Failed to mark inode dirty"); 1382a7550b30SJaegeuk Kim } 1383a7550b30SJaegeuk Kim res2 = ext4_journal_stop(handle); 13842f8f5e76SEric Biggers 13852f8f5e76SEric Biggers if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) 13862f8f5e76SEric Biggers goto retry; 1387a7550b30SJaegeuk Kim if (!res) 1388a7550b30SJaegeuk Kim res = res2; 1389a7550b30SJaegeuk Kim return res; 1390a7550b30SJaegeuk Kim } 1391a7550b30SJaegeuk Kim 1392c250b7ddSEric Biggers static bool ext4_dummy_context(struct inode *inode) 1393a7550b30SJaegeuk Kim { 1394a7550b30SJaegeuk Kim return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb)); 1395a7550b30SJaegeuk Kim } 1396a7550b30SJaegeuk Kim 1397b925acb8SEric Biggers static bool ext4_has_stable_inodes(struct super_block *sb) 1398b925acb8SEric Biggers { 1399b925acb8SEric Biggers return ext4_has_feature_stable_inodes(sb); 1400b925acb8SEric Biggers } 1401b925acb8SEric Biggers 1402b925acb8SEric Biggers static void ext4_get_ino_and_lblk_bits(struct super_block *sb, 1403b925acb8SEric Biggers int *ino_bits_ret, int *lblk_bits_ret) 1404b925acb8SEric Biggers { 1405b925acb8SEric Biggers *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count); 1406b925acb8SEric Biggers *lblk_bits_ret = 8 * sizeof(ext4_lblk_t); 1407b925acb8SEric Biggers } 1408b925acb8SEric Biggers 14096f69f0edSEric Biggers static const struct fscrypt_operations ext4_cryptops = { 1410a5d431efSEric Biggers .key_prefix = "ext4:", 1411a7550b30SJaegeuk Kim .get_context = ext4_get_context, 1412a7550b30SJaegeuk Kim .set_context = ext4_set_context, 1413a7550b30SJaegeuk Kim .dummy_context = ext4_dummy_context, 1414a7550b30SJaegeuk Kim .empty_dir = ext4_empty_dir, 1415e12ee683SEric Biggers .max_namelen = EXT4_NAME_LEN, 1416b925acb8SEric Biggers .has_stable_inodes = ext4_has_stable_inodes, 1417b925acb8SEric Biggers .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits, 1418a7550b30SJaegeuk Kim }; 1419a7550b30SJaegeuk Kim #endif 1420a7550b30SJaegeuk Kim 1421ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1422d6006186SEric Biggers static const char * const quotatypes[] = INITQFNAMES; 1423689c958cSLi Xi #define QTYPE2NAME(t) (quotatypes[t]) 1424ac27a0ecSDave Kleikamp 1425617ba13bSMingming Cao static int ext4_write_dquot(struct dquot *dquot); 1426617ba13bSMingming Cao static int ext4_acquire_dquot(struct dquot *dquot); 1427617ba13bSMingming Cao static int ext4_release_dquot(struct dquot *dquot); 1428617ba13bSMingming Cao static int ext4_mark_dquot_dirty(struct dquot *dquot); 1429617ba13bSMingming Cao static int ext4_write_info(struct super_block *sb, int type); 14306f28e087SJan Kara static int ext4_quota_on(struct super_block *sb, int type, int format_id, 14318c54ca9cSAl Viro const struct path *path); 1432617ba13bSMingming Cao static int ext4_quota_on_mount(struct super_block *sb, int type); 1433617ba13bSMingming Cao static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1434ac27a0ecSDave Kleikamp size_t len, loff_t off); 1435617ba13bSMingming Cao static ssize_t ext4_quota_write(struct super_block *sb, int type, 1436ac27a0ecSDave Kleikamp const char *data, size_t len, loff_t off); 14377c319d32SAditya Kali static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 14387c319d32SAditya Kali unsigned int flags); 14397c319d32SAditya Kali static int ext4_enable_quotas(struct super_block *sb); 1440ac27a0ecSDave Kleikamp 144196c7e0d9SJan Kara static struct dquot **ext4_get_dquots(struct inode *inode) 144296c7e0d9SJan Kara { 144396c7e0d9SJan Kara return EXT4_I(inode)->i_dquot; 144496c7e0d9SJan Kara } 144596c7e0d9SJan Kara 144661e225dcSAlexey Dobriyan static const struct dquot_operations ext4_quota_operations = { 144760e58e0fSMingming Cao .get_reserved_space = ext4_get_reserved_space, 1448617ba13bSMingming Cao .write_dquot = ext4_write_dquot, 1449617ba13bSMingming Cao .acquire_dquot = ext4_acquire_dquot, 1450617ba13bSMingming Cao .release_dquot = ext4_release_dquot, 1451617ba13bSMingming Cao .mark_dirty = ext4_mark_dquot_dirty, 1452a5b5ee32SJan Kara .write_info = ext4_write_info, 1453a5b5ee32SJan Kara .alloc_dquot = dquot_alloc, 1454a5b5ee32SJan Kara .destroy_dquot = dquot_destroy, 1455040cb378SLi Xi .get_projid = ext4_get_projid, 14567a9ca53aSTahsin Erdogan .get_inode_usage = ext4_get_inode_usage, 1457ebc11f7bSChengguang Xu .get_next_id = dquot_get_next_id, 1458ac27a0ecSDave Kleikamp }; 1459ac27a0ecSDave Kleikamp 14600d54b217SAlexey Dobriyan static const struct quotactl_ops ext4_qctl_operations = { 1461617ba13bSMingming Cao .quota_on = ext4_quota_on, 1462ca0e05e4SDmitry Monakhov .quota_off = ext4_quota_off, 1463287a8095SChristoph Hellwig .quota_sync = dquot_quota_sync, 14640a240339SJan Kara .get_state = dquot_get_state, 1465287a8095SChristoph Hellwig .set_info = dquot_set_dqinfo, 1466287a8095SChristoph Hellwig .get_dqblk = dquot_get_dqblk, 14676332b9b5SEric Sandeen .set_dqblk = dquot_set_dqblk, 14686332b9b5SEric Sandeen .get_nextdqblk = dquot_get_next_dqblk, 1469ac27a0ecSDave Kleikamp }; 1470ac27a0ecSDave Kleikamp #endif 1471ac27a0ecSDave Kleikamp 1472ee9b6d61SJosef 'Jeff' Sipek static const struct super_operations ext4_sops = { 1473617ba13bSMingming Cao .alloc_inode = ext4_alloc_inode, 147494053139SAl Viro .free_inode = ext4_free_in_core_inode, 1475617ba13bSMingming Cao .destroy_inode = ext4_destroy_inode, 1476617ba13bSMingming Cao .write_inode = ext4_write_inode, 1477617ba13bSMingming Cao .dirty_inode = ext4_dirty_inode, 14787ff9c073STheodore Ts'o .drop_inode = ext4_drop_inode, 14790930fcc1SAl Viro .evict_inode = ext4_evict_inode, 1480617ba13bSMingming Cao .put_super = ext4_put_super, 1481617ba13bSMingming Cao .sync_fs = ext4_sync_fs, 1482c4be0c1dSTakashi Sato .freeze_fs = ext4_freeze, 1483c4be0c1dSTakashi Sato .unfreeze_fs = ext4_unfreeze, 1484617ba13bSMingming Cao .statfs = ext4_statfs, 1485617ba13bSMingming Cao .remount_fs = ext4_remount, 1486617ba13bSMingming Cao .show_options = ext4_show_options, 1487ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1488617ba13bSMingming Cao .quota_read = ext4_quota_read, 1489617ba13bSMingming Cao .quota_write = ext4_quota_write, 149096c7e0d9SJan Kara .get_dquots = ext4_get_dquots, 1491ac27a0ecSDave Kleikamp #endif 1492c39a7f84SToshiyuki Okajima .bdev_try_to_free_page = bdev_try_to_free_page, 1493ac27a0ecSDave Kleikamp }; 1494ac27a0ecSDave Kleikamp 149539655164SChristoph Hellwig static const struct export_operations ext4_export_ops = { 14961b961ac0SChristoph Hellwig .fh_to_dentry = ext4_fh_to_dentry, 14971b961ac0SChristoph Hellwig .fh_to_parent = ext4_fh_to_parent, 1498617ba13bSMingming Cao .get_parent = ext4_get_parent, 1499fde87268STheodore Ts'o .commit_metadata = ext4_nfs_commit_metadata, 1500ac27a0ecSDave Kleikamp }; 1501ac27a0ecSDave Kleikamp 1502ac27a0ecSDave Kleikamp enum { 1503ac27a0ecSDave Kleikamp Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1504ac27a0ecSDave Kleikamp Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 150572578c33STheodore Ts'o Opt_nouid32, Opt_debug, Opt_removed, 1506ac27a0ecSDave Kleikamp Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 150772578c33STheodore Ts'o Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1508ad4eec61SEric Sandeen Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1509ad4eec61SEric Sandeen Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1510ac27a0ecSDave Kleikamp Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 15116ddb2447STheodore Ts'o Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 1512ac27a0ecSDave Kleikamp Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 15135a20bdfcSJan Kara Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1514ee4a3fcdSTheodore Ts'o Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1515*9cb20f94SIra Weiny Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, 1516*9cb20f94SIra Weiny Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never, 1517327eaf73STheodore Ts'o Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1518327eaf73STheodore Ts'o Opt_nowarn_on_error, Opt_mblk_io_submit, 1519670e9875STheodore Ts'o Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize, 15201449032bSTheodore Ts'o Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 15215328e635SEric Sandeen Opt_inode_readahead_blks, Opt_journal_ioprio, 1522744692dcSJiaying Zhang Opt_dioread_nolock, Opt_dioread_lock, 1523fc6cb1cdSTheodore Ts'o Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1524cdb7ee4cSTahsin Erdogan Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 1525ac27a0ecSDave Kleikamp }; 1526ac27a0ecSDave Kleikamp 1527a447c093SSteven Whitehouse static const match_table_t tokens = { 1528ac27a0ecSDave Kleikamp {Opt_bsd_df, "bsddf"}, 1529ac27a0ecSDave Kleikamp {Opt_minix_df, "minixdf"}, 1530ac27a0ecSDave Kleikamp {Opt_grpid, "grpid"}, 1531ac27a0ecSDave Kleikamp {Opt_grpid, "bsdgroups"}, 1532ac27a0ecSDave Kleikamp {Opt_nogrpid, "nogrpid"}, 1533ac27a0ecSDave Kleikamp {Opt_nogrpid, "sysvgroups"}, 1534ac27a0ecSDave Kleikamp {Opt_resgid, "resgid=%u"}, 1535ac27a0ecSDave Kleikamp {Opt_resuid, "resuid=%u"}, 1536ac27a0ecSDave Kleikamp {Opt_sb, "sb=%u"}, 1537ac27a0ecSDave Kleikamp {Opt_err_cont, "errors=continue"}, 1538ac27a0ecSDave Kleikamp {Opt_err_panic, "errors=panic"}, 1539ac27a0ecSDave Kleikamp {Opt_err_ro, "errors=remount-ro"}, 1540ac27a0ecSDave Kleikamp {Opt_nouid32, "nouid32"}, 1541ac27a0ecSDave Kleikamp {Opt_debug, "debug"}, 154272578c33STheodore Ts'o {Opt_removed, "oldalloc"}, 154372578c33STheodore Ts'o {Opt_removed, "orlov"}, 1544ac27a0ecSDave Kleikamp {Opt_user_xattr, "user_xattr"}, 1545ac27a0ecSDave Kleikamp {Opt_nouser_xattr, "nouser_xattr"}, 1546ac27a0ecSDave Kleikamp {Opt_acl, "acl"}, 1547ac27a0ecSDave Kleikamp {Opt_noacl, "noacl"}, 1548e3bb52aeSEric Sandeen {Opt_noload, "norecovery"}, 15495a916be1STheodore Ts'o {Opt_noload, "noload"}, 155072578c33STheodore Ts'o {Opt_removed, "nobh"}, 155172578c33STheodore Ts'o {Opt_removed, "bh"}, 1552ac27a0ecSDave Kleikamp {Opt_commit, "commit=%u"}, 155330773840STheodore Ts'o {Opt_min_batch_time, "min_batch_time=%u"}, 155430773840STheodore Ts'o {Opt_max_batch_time, "max_batch_time=%u"}, 1555ac27a0ecSDave Kleikamp {Opt_journal_dev, "journal_dev=%u"}, 1556ad4eec61SEric Sandeen {Opt_journal_path, "journal_path=%s"}, 1557818d276cSGirish Shilamkar {Opt_journal_checksum, "journal_checksum"}, 1558c6d3d56dSDarrick J. Wong {Opt_nojournal_checksum, "nojournal_checksum"}, 1559818d276cSGirish Shilamkar {Opt_journal_async_commit, "journal_async_commit"}, 1560ac27a0ecSDave Kleikamp {Opt_abort, "abort"}, 1561ac27a0ecSDave Kleikamp {Opt_data_journal, "data=journal"}, 1562ac27a0ecSDave Kleikamp {Opt_data_ordered, "data=ordered"}, 1563ac27a0ecSDave Kleikamp {Opt_data_writeback, "data=writeback"}, 15645bf5683aSHidehiro Kawai {Opt_data_err_abort, "data_err=abort"}, 15655bf5683aSHidehiro Kawai {Opt_data_err_ignore, "data_err=ignore"}, 1566ac27a0ecSDave Kleikamp {Opt_offusrjquota, "usrjquota="}, 1567ac27a0ecSDave Kleikamp {Opt_usrjquota, "usrjquota=%s"}, 1568ac27a0ecSDave Kleikamp {Opt_offgrpjquota, "grpjquota="}, 1569ac27a0ecSDave Kleikamp {Opt_grpjquota, "grpjquota=%s"}, 1570ac27a0ecSDave Kleikamp {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1571ac27a0ecSDave Kleikamp {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 15725a20bdfcSJan Kara {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1573ac27a0ecSDave Kleikamp {Opt_grpquota, "grpquota"}, 1574ac27a0ecSDave Kleikamp {Opt_noquota, "noquota"}, 1575ac27a0ecSDave Kleikamp {Opt_quota, "quota"}, 1576ac27a0ecSDave Kleikamp {Opt_usrquota, "usrquota"}, 157749da9392SJan Kara {Opt_prjquota, "prjquota"}, 1578ac27a0ecSDave Kleikamp {Opt_barrier, "barrier=%u"}, 157906705bffSTheodore Ts'o {Opt_barrier, "barrier"}, 158006705bffSTheodore Ts'o {Opt_nobarrier, "nobarrier"}, 158125ec56b5SJean Noel Cordenner {Opt_i_version, "i_version"}, 1582923ae0ffSRoss Zwisler {Opt_dax, "dax"}, 1583*9cb20f94SIra Weiny {Opt_dax_always, "dax=always"}, 1584*9cb20f94SIra Weiny {Opt_dax_inode, "dax=inode"}, 1585*9cb20f94SIra Weiny {Opt_dax_never, "dax=never"}, 1586c9de560dSAlex Tomas {Opt_stripe, "stripe=%u"}, 158764769240SAlex Tomas {Opt_delalloc, "delalloc"}, 1588327eaf73STheodore Ts'o {Opt_warn_on_error, "warn_on_error"}, 1589327eaf73STheodore Ts'o {Opt_nowarn_on_error, "nowarn_on_error"}, 1590a26f4992STheodore Ts'o {Opt_lazytime, "lazytime"}, 1591a26f4992STheodore Ts'o {Opt_nolazytime, "nolazytime"}, 1592670e9875STheodore Ts'o {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"}, 1593dd919b98SAneesh Kumar K.V {Opt_nodelalloc, "nodelalloc"}, 159436ade451SJan Kara {Opt_removed, "mblk_io_submit"}, 159536ade451SJan Kara {Opt_removed, "nomblk_io_submit"}, 15966fd058f7STheodore Ts'o {Opt_block_validity, "block_validity"}, 15976fd058f7STheodore Ts'o {Opt_noblock_validity, "noblock_validity"}, 1598240799cdSTheodore Ts'o {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1599b3881f74STheodore Ts'o {Opt_journal_ioprio, "journal_ioprio=%u"}, 1600afd4672dSTheodore Ts'o {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 160106705bffSTheodore Ts'o {Opt_auto_da_alloc, "auto_da_alloc"}, 160206705bffSTheodore Ts'o {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1603744692dcSJiaying Zhang {Opt_dioread_nolock, "dioread_nolock"}, 1604244adf64STheodore Ts'o {Opt_dioread_lock, "nodioread_nolock"}, 1605744692dcSJiaying Zhang {Opt_dioread_lock, "dioread_lock"}, 16065328e635SEric Sandeen {Opt_discard, "discard"}, 16075328e635SEric Sandeen {Opt_nodiscard, "nodiscard"}, 1608fc6cb1cdSTheodore Ts'o {Opt_init_itable, "init_itable=%u"}, 1609fc6cb1cdSTheodore Ts'o {Opt_init_itable, "init_itable"}, 1610fc6cb1cdSTheodore Ts'o {Opt_noinit_itable, "noinit_itable"}, 1611df981d03STheodore Ts'o {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 16126ddb2447STheodore Ts'o {Opt_test_dummy_encryption, "test_dummy_encryption"}, 1613cdb7ee4cSTahsin Erdogan {Opt_nombcache, "nombcache"}, 1614cdb7ee4cSTahsin Erdogan {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */ 1615c7198b9cSTheodore Ts'o {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1616c7198b9cSTheodore Ts'o {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1617c7198b9cSTheodore Ts'o {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1618c7198b9cSTheodore Ts'o {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1619c7198b9cSTheodore Ts'o {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1620f3f12faaSJosef Bacik {Opt_err, NULL}, 1621ac27a0ecSDave Kleikamp }; 1622ac27a0ecSDave Kleikamp 1623617ba13bSMingming Cao static ext4_fsblk_t get_sb_block(void **data) 1624ac27a0ecSDave Kleikamp { 1625617ba13bSMingming Cao ext4_fsblk_t sb_block; 1626ac27a0ecSDave Kleikamp char *options = (char *) *data; 1627ac27a0ecSDave Kleikamp 1628ac27a0ecSDave Kleikamp if (!options || strncmp(options, "sb=", 3) != 0) 1629ac27a0ecSDave Kleikamp return 1; /* Default location */ 16300b8e58a1SAndreas Dilger 1631ac27a0ecSDave Kleikamp options += 3; 16320b8e58a1SAndreas Dilger /* TODO: use simple_strtoll with >32bit ext4 */ 1633ac27a0ecSDave Kleikamp sb_block = simple_strtoul(options, &options, 0); 1634ac27a0ecSDave Kleikamp if (*options && *options != ',') { 16354776004fSTheodore Ts'o printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1636ac27a0ecSDave Kleikamp (char *) *data); 1637ac27a0ecSDave Kleikamp return 1; 1638ac27a0ecSDave Kleikamp } 1639ac27a0ecSDave Kleikamp if (*options == ',') 1640ac27a0ecSDave Kleikamp options++; 1641ac27a0ecSDave Kleikamp *data = (void *) options; 16420b8e58a1SAndreas Dilger 1643ac27a0ecSDave Kleikamp return sb_block; 1644ac27a0ecSDave Kleikamp } 1645ac27a0ecSDave Kleikamp 1646b3881f74STheodore Ts'o #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1647d6006186SEric Biggers static const char deprecated_msg[] = 1648d6006186SEric Biggers "Mount option \"%s\" will be removed by %s\n" 1649437ca0fdSDmitry Monakhov "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1650b3881f74STheodore Ts'o 165156c50f11SDmitry Monakhov #ifdef CONFIG_QUOTA 165256c50f11SDmitry Monakhov static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 165356c50f11SDmitry Monakhov { 165456c50f11SDmitry Monakhov struct ext4_sb_info *sbi = EXT4_SB(sb); 165533458eabSTheodore Ts'o char *qname, *old_qname = get_qf_name(sb, sbi, qtype); 165603dafb5fSChen Gang int ret = -1; 165756c50f11SDmitry Monakhov 165833458eabSTheodore Ts'o if (sb_any_quota_loaded(sb) && !old_qname) { 165956c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 166056c50f11SDmitry Monakhov "Cannot change journaled " 166156c50f11SDmitry Monakhov "quota options when quota turned on"); 166257f73c2cSTheodore Ts'o return -1; 166356c50f11SDmitry Monakhov } 1664e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) { 1665c325a67cSTheodore Ts'o ext4_msg(sb, KERN_INFO, "Journaled quota options " 1666c325a67cSTheodore Ts'o "ignored when QUOTA feature is enabled"); 1667c325a67cSTheodore Ts'o return 1; 1668262b4662SJan Kara } 166956c50f11SDmitry Monakhov qname = match_strdup(args); 167056c50f11SDmitry Monakhov if (!qname) { 167156c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 167256c50f11SDmitry Monakhov "Not enough memory for storing quotafile name"); 167357f73c2cSTheodore Ts'o return -1; 167456c50f11SDmitry Monakhov } 167533458eabSTheodore Ts'o if (old_qname) { 167633458eabSTheodore Ts'o if (strcmp(old_qname, qname) == 0) 167703dafb5fSChen Gang ret = 1; 167803dafb5fSChen Gang else 167956c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 168003dafb5fSChen Gang "%s quota file already specified", 168103dafb5fSChen Gang QTYPE2NAME(qtype)); 168203dafb5fSChen Gang goto errout; 168356c50f11SDmitry Monakhov } 168403dafb5fSChen Gang if (strchr(qname, '/')) { 168556c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 168656c50f11SDmitry Monakhov "quotafile must be on filesystem root"); 168703dafb5fSChen Gang goto errout; 168856c50f11SDmitry Monakhov } 168933458eabSTheodore Ts'o rcu_assign_pointer(sbi->s_qf_names[qtype], qname); 1690fd8c37ecSTheodore Ts'o set_opt(sb, QUOTA); 169156c50f11SDmitry Monakhov return 1; 169203dafb5fSChen Gang errout: 169303dafb5fSChen Gang kfree(qname); 169403dafb5fSChen Gang return ret; 169556c50f11SDmitry Monakhov } 169656c50f11SDmitry Monakhov 169756c50f11SDmitry Monakhov static int clear_qf_name(struct super_block *sb, int qtype) 169856c50f11SDmitry Monakhov { 169956c50f11SDmitry Monakhov 170056c50f11SDmitry Monakhov struct ext4_sb_info *sbi = EXT4_SB(sb); 170133458eabSTheodore Ts'o char *old_qname = get_qf_name(sb, sbi, qtype); 170256c50f11SDmitry Monakhov 170333458eabSTheodore Ts'o if (sb_any_quota_loaded(sb) && old_qname) { 170456c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 170556c50f11SDmitry Monakhov " when quota turned on"); 170657f73c2cSTheodore Ts'o return -1; 170756c50f11SDmitry Monakhov } 170833458eabSTheodore Ts'o rcu_assign_pointer(sbi->s_qf_names[qtype], NULL); 170933458eabSTheodore Ts'o synchronize_rcu(); 171033458eabSTheodore Ts'o kfree(old_qname); 171156c50f11SDmitry Monakhov return 1; 171256c50f11SDmitry Monakhov } 171356c50f11SDmitry Monakhov #endif 171456c50f11SDmitry Monakhov 171526092bf5STheodore Ts'o #define MOPT_SET 0x0001 171626092bf5STheodore Ts'o #define MOPT_CLEAR 0x0002 171726092bf5STheodore Ts'o #define MOPT_NOSUPPORT 0x0004 171826092bf5STheodore Ts'o #define MOPT_EXPLICIT 0x0008 171926092bf5STheodore Ts'o #define MOPT_CLEAR_ERR 0x0010 172026092bf5STheodore Ts'o #define MOPT_GTE0 0x0020 172126092bf5STheodore Ts'o #ifdef CONFIG_QUOTA 172226092bf5STheodore Ts'o #define MOPT_Q 0 172326092bf5STheodore Ts'o #define MOPT_QFMT 0x0040 172426092bf5STheodore Ts'o #else 172526092bf5STheodore Ts'o #define MOPT_Q MOPT_NOSUPPORT 172626092bf5STheodore Ts'o #define MOPT_QFMT MOPT_NOSUPPORT 172726092bf5STheodore Ts'o #endif 172826092bf5STheodore Ts'o #define MOPT_DATAJ 0x0080 17298dc0aa8cSTheodore Ts'o #define MOPT_NO_EXT2 0x0100 17308dc0aa8cSTheodore Ts'o #define MOPT_NO_EXT3 0x0200 17318dc0aa8cSTheodore Ts'o #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1732ad4eec61SEric Sandeen #define MOPT_STRING 0x0400 1733*9cb20f94SIra Weiny #define MOPT_SKIP 0x0800 173426092bf5STheodore Ts'o 173526092bf5STheodore Ts'o static const struct mount_opts { 173626092bf5STheodore Ts'o int token; 173726092bf5STheodore Ts'o int mount_opt; 173826092bf5STheodore Ts'o int flags; 173926092bf5STheodore Ts'o } ext4_mount_opts[] = { 174026092bf5STheodore Ts'o {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 174126092bf5STheodore Ts'o {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 174226092bf5STheodore Ts'o {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 174326092bf5STheodore Ts'o {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 174426092bf5STheodore Ts'o {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 174526092bf5STheodore Ts'o {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 17468dc0aa8cSTheodore Ts'o {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 17478dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_SET}, 17488dc0aa8cSTheodore Ts'o {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 17498dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_CLEAR}, 175026092bf5STheodore Ts'o {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 175126092bf5STheodore Ts'o {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 17528dc0aa8cSTheodore Ts'o {Opt_delalloc, EXT4_MOUNT_DELALLOC, 17538dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 17548dc0aa8cSTheodore Ts'o {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 175559d9fa5cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_CLEAR}, 1756327eaf73STheodore Ts'o {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1757327eaf73STheodore Ts'o {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1758c6d3d56dSDarrick J. Wong {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1759c6d3d56dSDarrick J. Wong MOPT_EXT4_ONLY | MOPT_CLEAR}, 17608dc0aa8cSTheodore Ts'o {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 17611e381f60SDmitry Monakhov MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 176226092bf5STheodore Ts'o {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 17638dc0aa8cSTheodore Ts'o EXT4_MOUNT_JOURNAL_CHECKSUM), 17641e381f60SDmitry Monakhov MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 17658dc0aa8cSTheodore Ts'o {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 176626092bf5STheodore Ts'o {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 176726092bf5STheodore Ts'o {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 176826092bf5STheodore Ts'o {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 17698dc0aa8cSTheodore Ts'o {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 17707915a861SAles Novak MOPT_NO_EXT2}, 17718dc0aa8cSTheodore Ts'o {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 17727915a861SAles Novak MOPT_NO_EXT2}, 177326092bf5STheodore Ts'o {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 177426092bf5STheodore Ts'o {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 177526092bf5STheodore Ts'o {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 177626092bf5STheodore Ts'o {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 177726092bf5STheodore Ts'o {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 177826092bf5STheodore Ts'o {Opt_commit, 0, MOPT_GTE0}, 177926092bf5STheodore Ts'o {Opt_max_batch_time, 0, MOPT_GTE0}, 178026092bf5STheodore Ts'o {Opt_min_batch_time, 0, MOPT_GTE0}, 178126092bf5STheodore Ts'o {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 178226092bf5STheodore Ts'o {Opt_init_itable, 0, MOPT_GTE0}, 1783*9cb20f94SIra Weiny {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP}, 1784*9cb20f94SIra Weiny {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS, 1785*9cb20f94SIra Weiny MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 1786*9cb20f94SIra Weiny {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE, 1787*9cb20f94SIra Weiny MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 1788*9cb20f94SIra Weiny {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER, 1789*9cb20f94SIra Weiny MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 179026092bf5STheodore Ts'o {Opt_stripe, 0, MOPT_GTE0}, 17910efb3b23SJan Kara {Opt_resuid, 0, MOPT_GTE0}, 17920efb3b23SJan Kara {Opt_resgid, 0, MOPT_GTE0}, 17935ba92bcfSCarlos Maiolino {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 17945ba92bcfSCarlos Maiolino {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING}, 17955ba92bcfSCarlos Maiolino {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 17968dc0aa8cSTheodore Ts'o {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 17978dc0aa8cSTheodore Ts'o {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 17988dc0aa8cSTheodore Ts'o {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 17998dc0aa8cSTheodore Ts'o MOPT_NO_EXT2 | MOPT_DATAJ}, 180026092bf5STheodore Ts'o {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 180126092bf5STheodore Ts'o {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 180226092bf5STheodore Ts'o #ifdef CONFIG_EXT4_FS_POSIX_ACL 180326092bf5STheodore Ts'o {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 180426092bf5STheodore Ts'o {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 180526092bf5STheodore Ts'o #else 180626092bf5STheodore Ts'o {Opt_acl, 0, MOPT_NOSUPPORT}, 180726092bf5STheodore Ts'o {Opt_noacl, 0, MOPT_NOSUPPORT}, 180826092bf5STheodore Ts'o #endif 180926092bf5STheodore Ts'o {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 181026092bf5STheodore Ts'o {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1811670e9875STheodore Ts'o {Opt_debug_want_extra_isize, 0, MOPT_GTE0}, 181226092bf5STheodore Ts'o {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 181326092bf5STheodore Ts'o {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 181426092bf5STheodore Ts'o MOPT_SET | MOPT_Q}, 181526092bf5STheodore Ts'o {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 181626092bf5STheodore Ts'o MOPT_SET | MOPT_Q}, 181749da9392SJan Kara {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 181849da9392SJan Kara MOPT_SET | MOPT_Q}, 181926092bf5STheodore Ts'o {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 182049da9392SJan Kara EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 182149da9392SJan Kara MOPT_CLEAR | MOPT_Q}, 182226092bf5STheodore Ts'o {Opt_usrjquota, 0, MOPT_Q}, 182326092bf5STheodore Ts'o {Opt_grpjquota, 0, MOPT_Q}, 182426092bf5STheodore Ts'o {Opt_offusrjquota, 0, MOPT_Q}, 182526092bf5STheodore Ts'o {Opt_offgrpjquota, 0, MOPT_Q}, 182626092bf5STheodore Ts'o {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 182726092bf5STheodore Ts'o {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 182826092bf5STheodore Ts'o {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 1829df981d03STheodore Ts'o {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 18306ddb2447STheodore Ts'o {Opt_test_dummy_encryption, 0, MOPT_GTE0}, 1831cdb7ee4cSTahsin Erdogan {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 183226092bf5STheodore Ts'o {Opt_err, 0, 0} 183326092bf5STheodore Ts'o }; 183426092bf5STheodore Ts'o 1835c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 1836c83ad55eSGabriel Krisman Bertazi static const struct ext4_sb_encodings { 1837c83ad55eSGabriel Krisman Bertazi __u16 magic; 1838c83ad55eSGabriel Krisman Bertazi char *name; 1839c83ad55eSGabriel Krisman Bertazi char *version; 1840c83ad55eSGabriel Krisman Bertazi } ext4_sb_encoding_map[] = { 1841c83ad55eSGabriel Krisman Bertazi {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"}, 1842c83ad55eSGabriel Krisman Bertazi }; 1843c83ad55eSGabriel Krisman Bertazi 1844c83ad55eSGabriel Krisman Bertazi static int ext4_sb_read_encoding(const struct ext4_super_block *es, 1845c83ad55eSGabriel Krisman Bertazi const struct ext4_sb_encodings **encoding, 1846c83ad55eSGabriel Krisman Bertazi __u16 *flags) 1847c83ad55eSGabriel Krisman Bertazi { 1848c83ad55eSGabriel Krisman Bertazi __u16 magic = le16_to_cpu(es->s_encoding); 1849c83ad55eSGabriel Krisman Bertazi int i; 1850c83ad55eSGabriel Krisman Bertazi 1851c83ad55eSGabriel Krisman Bertazi for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) 1852c83ad55eSGabriel Krisman Bertazi if (magic == ext4_sb_encoding_map[i].magic) 1853c83ad55eSGabriel Krisman Bertazi break; 1854c83ad55eSGabriel Krisman Bertazi 1855c83ad55eSGabriel Krisman Bertazi if (i >= ARRAY_SIZE(ext4_sb_encoding_map)) 1856c83ad55eSGabriel Krisman Bertazi return -EINVAL; 1857c83ad55eSGabriel Krisman Bertazi 1858c83ad55eSGabriel Krisman Bertazi *encoding = &ext4_sb_encoding_map[i]; 1859c83ad55eSGabriel Krisman Bertazi *flags = le16_to_cpu(es->s_encoding_flags); 1860c83ad55eSGabriel Krisman Bertazi 1861c83ad55eSGabriel Krisman Bertazi return 0; 1862c83ad55eSGabriel Krisman Bertazi } 1863c83ad55eSGabriel Krisman Bertazi #endif 1864c83ad55eSGabriel Krisman Bertazi 186526092bf5STheodore Ts'o static int handle_mount_opt(struct super_block *sb, char *opt, int token, 186626092bf5STheodore Ts'o substring_t *args, unsigned long *journal_devnum, 186726092bf5STheodore Ts'o unsigned int *journal_ioprio, int is_remount) 186826092bf5STheodore Ts'o { 186926092bf5STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 187026092bf5STheodore Ts'o const struct mount_opts *m; 187108cefc7aSEric W. Biederman kuid_t uid; 187208cefc7aSEric W. Biederman kgid_t gid; 187326092bf5STheodore Ts'o int arg = 0; 187426092bf5STheodore Ts'o 187557f73c2cSTheodore Ts'o #ifdef CONFIG_QUOTA 187657f73c2cSTheodore Ts'o if (token == Opt_usrjquota) 187757f73c2cSTheodore Ts'o return set_qf_name(sb, USRQUOTA, &args[0]); 187857f73c2cSTheodore Ts'o else if (token == Opt_grpjquota) 187957f73c2cSTheodore Ts'o return set_qf_name(sb, GRPQUOTA, &args[0]); 188057f73c2cSTheodore Ts'o else if (token == Opt_offusrjquota) 188157f73c2cSTheodore Ts'o return clear_qf_name(sb, USRQUOTA); 188257f73c2cSTheodore Ts'o else if (token == Opt_offgrpjquota) 188357f73c2cSTheodore Ts'o return clear_qf_name(sb, GRPQUOTA); 188457f73c2cSTheodore Ts'o #endif 188526092bf5STheodore Ts'o switch (token) { 1886f7048605STheodore Ts'o case Opt_noacl: 1887f7048605STheodore Ts'o case Opt_nouser_xattr: 1888f7048605STheodore Ts'o ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 1889f7048605STheodore Ts'o break; 189026092bf5STheodore Ts'o case Opt_sb: 189126092bf5STheodore Ts'o return 1; /* handled by get_sb_block() */ 189226092bf5STheodore Ts'o case Opt_removed: 18935f3633e3SJan Kara ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 189426092bf5STheodore Ts'o return 1; 189526092bf5STheodore Ts'o case Opt_abort: 189626092bf5STheodore Ts'o sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 189726092bf5STheodore Ts'o return 1; 189826092bf5STheodore Ts'o case Opt_i_version: 1899357fdad0SMatthew Garrett sb->s_flags |= SB_I_VERSION; 190026092bf5STheodore Ts'o return 1; 1901a26f4992STheodore Ts'o case Opt_lazytime: 19021751e8a6SLinus Torvalds sb->s_flags |= SB_LAZYTIME; 1903a26f4992STheodore Ts'o return 1; 1904a26f4992STheodore Ts'o case Opt_nolazytime: 19051751e8a6SLinus Torvalds sb->s_flags &= ~SB_LAZYTIME; 1906a26f4992STheodore Ts'o return 1; 190726092bf5STheodore Ts'o } 190826092bf5STheodore Ts'o 19095f3633e3SJan Kara for (m = ext4_mount_opts; m->token != Opt_err; m++) 19105f3633e3SJan Kara if (token == m->token) 19115f3633e3SJan Kara break; 19125f3633e3SJan Kara 19135f3633e3SJan Kara if (m->token == Opt_err) { 19145f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 19155f3633e3SJan Kara "or missing value", opt); 19165f3633e3SJan Kara return -1; 19175f3633e3SJan Kara } 19185f3633e3SJan Kara 19198dc0aa8cSTheodore Ts'o if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 19208dc0aa8cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 19218dc0aa8cSTheodore Ts'o "Mount option \"%s\" incompatible with ext2", opt); 19228dc0aa8cSTheodore Ts'o return -1; 19238dc0aa8cSTheodore Ts'o } 19248dc0aa8cSTheodore Ts'o if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 19258dc0aa8cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 19268dc0aa8cSTheodore Ts'o "Mount option \"%s\" incompatible with ext3", opt); 19278dc0aa8cSTheodore Ts'o return -1; 19288dc0aa8cSTheodore Ts'o } 19298dc0aa8cSTheodore Ts'o 1930ad4eec61SEric Sandeen if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) 19310efb3b23SJan Kara return -1; 193226092bf5STheodore Ts'o if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 193326092bf5STheodore Ts'o return -1; 1934c93cf2d7SDmitry Monakhov if (m->flags & MOPT_EXPLICIT) { 1935c93cf2d7SDmitry Monakhov if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 193626092bf5STheodore Ts'o set_opt2(sb, EXPLICIT_DELALLOC); 19371e381f60SDmitry Monakhov } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 19381e381f60SDmitry Monakhov set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM); 1939c93cf2d7SDmitry Monakhov } else 1940c93cf2d7SDmitry Monakhov return -1; 1941c93cf2d7SDmitry Monakhov } 194226092bf5STheodore Ts'o if (m->flags & MOPT_CLEAR_ERR) 194326092bf5STheodore Ts'o clear_opt(sb, ERRORS_MASK); 194426092bf5STheodore Ts'o if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 194526092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "Cannot change quota " 194626092bf5STheodore Ts'o "options when quota turned on"); 194726092bf5STheodore Ts'o return -1; 194826092bf5STheodore Ts'o } 194926092bf5STheodore Ts'o 195026092bf5STheodore Ts'o if (m->flags & MOPT_NOSUPPORT) { 195126092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 195226092bf5STheodore Ts'o } else if (token == Opt_commit) { 195326092bf5STheodore Ts'o if (arg == 0) 195426092bf5STheodore Ts'o arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 19559ba55543Szhangyi (F) else if (arg > INT_MAX / HZ) { 19569ba55543Szhangyi (F) ext4_msg(sb, KERN_ERR, 19579ba55543Szhangyi (F) "Invalid commit interval %d, " 19589ba55543Szhangyi (F) "must be smaller than %d", 19599ba55543Szhangyi (F) arg, INT_MAX / HZ); 19609ba55543Szhangyi (F) return -1; 19619ba55543Szhangyi (F) } 196226092bf5STheodore Ts'o sbi->s_commit_interval = HZ * arg; 1963670e9875STheodore Ts'o } else if (token == Opt_debug_want_extra_isize) { 19649803387cSTheodore Ts'o if ((arg & 1) || 19659803387cSTheodore Ts'o (arg < 4) || 19669803387cSTheodore Ts'o (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) { 19679803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 19689803387cSTheodore Ts'o "Invalid want_extra_isize %d", arg); 19699803387cSTheodore Ts'o return -1; 19709803387cSTheodore Ts'o } 1971670e9875STheodore Ts'o sbi->s_want_extra_isize = arg; 197226092bf5STheodore Ts'o } else if (token == Opt_max_batch_time) { 197326092bf5STheodore Ts'o sbi->s_max_batch_time = arg; 197426092bf5STheodore Ts'o } else if (token == Opt_min_batch_time) { 197526092bf5STheodore Ts'o sbi->s_min_batch_time = arg; 197626092bf5STheodore Ts'o } else if (token == Opt_inode_readahead_blks) { 1977e33e60eaSJan Kara if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 1978e33e60eaSJan Kara ext4_msg(sb, KERN_ERR, 1979e33e60eaSJan Kara "EXT4-fs: inode_readahead_blks must be " 1980e33e60eaSJan Kara "0 or a power of 2 smaller than 2^31"); 198126092bf5STheodore Ts'o return -1; 198226092bf5STheodore Ts'o } 198326092bf5STheodore Ts'o sbi->s_inode_readahead_blks = arg; 198426092bf5STheodore Ts'o } else if (token == Opt_init_itable) { 198526092bf5STheodore Ts'o set_opt(sb, INIT_INODE_TABLE); 198626092bf5STheodore Ts'o if (!args->from) 198726092bf5STheodore Ts'o arg = EXT4_DEF_LI_WAIT_MULT; 198826092bf5STheodore Ts'o sbi->s_li_wait_mult = arg; 1989df981d03STheodore Ts'o } else if (token == Opt_max_dir_size_kb) { 1990df981d03STheodore Ts'o sbi->s_max_dir_size_kb = arg; 199126092bf5STheodore Ts'o } else if (token == Opt_stripe) { 199226092bf5STheodore Ts'o sbi->s_stripe = arg; 19930efb3b23SJan Kara } else if (token == Opt_resuid) { 19940efb3b23SJan Kara uid = make_kuid(current_user_ns(), arg); 19950efb3b23SJan Kara if (!uid_valid(uid)) { 19965f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 19970efb3b23SJan Kara return -1; 19980efb3b23SJan Kara } 19990efb3b23SJan Kara sbi->s_resuid = uid; 20000efb3b23SJan Kara } else if (token == Opt_resgid) { 20010efb3b23SJan Kara gid = make_kgid(current_user_ns(), arg); 20020efb3b23SJan Kara if (!gid_valid(gid)) { 20035f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 20040efb3b23SJan Kara return -1; 20050efb3b23SJan Kara } 20060efb3b23SJan Kara sbi->s_resgid = gid; 20070efb3b23SJan Kara } else if (token == Opt_journal_dev) { 20080efb3b23SJan Kara if (is_remount) { 20090efb3b23SJan Kara ext4_msg(sb, KERN_ERR, 20100efb3b23SJan Kara "Cannot specify journal on remount"); 20110efb3b23SJan Kara return -1; 20120efb3b23SJan Kara } 20130efb3b23SJan Kara *journal_devnum = arg; 2014ad4eec61SEric Sandeen } else if (token == Opt_journal_path) { 2015ad4eec61SEric Sandeen char *journal_path; 2016ad4eec61SEric Sandeen struct inode *journal_inode; 2017ad4eec61SEric Sandeen struct path path; 2018ad4eec61SEric Sandeen int error; 2019ad4eec61SEric Sandeen 2020ad4eec61SEric Sandeen if (is_remount) { 2021ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, 2022ad4eec61SEric Sandeen "Cannot specify journal on remount"); 2023ad4eec61SEric Sandeen return -1; 2024ad4eec61SEric Sandeen } 2025ad4eec61SEric Sandeen journal_path = match_strdup(&args[0]); 2026ad4eec61SEric Sandeen if (!journal_path) { 2027ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: could not dup " 2028ad4eec61SEric Sandeen "journal device string"); 2029ad4eec61SEric Sandeen return -1; 2030ad4eec61SEric Sandeen } 2031ad4eec61SEric Sandeen 2032ad4eec61SEric Sandeen error = kern_path(journal_path, LOOKUP_FOLLOW, &path); 2033ad4eec61SEric Sandeen if (error) { 2034ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: could not find " 2035ad4eec61SEric Sandeen "journal device path: error %d", error); 2036ad4eec61SEric Sandeen kfree(journal_path); 2037ad4eec61SEric Sandeen return -1; 2038ad4eec61SEric Sandeen } 2039ad4eec61SEric Sandeen 20402b0143b5SDavid Howells journal_inode = d_inode(path.dentry); 2041ad4eec61SEric Sandeen if (!S_ISBLK(journal_inode->i_mode)) { 2042ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: journal path %s " 2043ad4eec61SEric Sandeen "is not a block device", journal_path); 2044ad4eec61SEric Sandeen path_put(&path); 2045ad4eec61SEric Sandeen kfree(journal_path); 2046ad4eec61SEric Sandeen return -1; 2047ad4eec61SEric Sandeen } 2048ad4eec61SEric Sandeen 2049ad4eec61SEric Sandeen *journal_devnum = new_encode_dev(journal_inode->i_rdev); 2050ad4eec61SEric Sandeen path_put(&path); 2051ad4eec61SEric Sandeen kfree(journal_path); 20520efb3b23SJan Kara } else if (token == Opt_journal_ioprio) { 20530efb3b23SJan Kara if (arg > 7) { 20545f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 20550efb3b23SJan Kara " (must be 0-7)"); 20560efb3b23SJan Kara return -1; 20570efb3b23SJan Kara } 20580efb3b23SJan Kara *journal_ioprio = 20590efb3b23SJan Kara IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 20606ddb2447STheodore Ts'o } else if (token == Opt_test_dummy_encryption) { 2061643fa961SChandan Rajendra #ifdef CONFIG_FS_ENCRYPTION 20626ddb2447STheodore Ts'o sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION; 20636ddb2447STheodore Ts'o ext4_msg(sb, KERN_WARNING, 20646ddb2447STheodore Ts'o "Test dummy encryption mode enabled"); 20656ddb2447STheodore Ts'o #else 20666ddb2447STheodore Ts'o ext4_msg(sb, KERN_WARNING, 20676ddb2447STheodore Ts'o "Test dummy encryption mount option ignored"); 20686ddb2447STheodore Ts'o #endif 206926092bf5STheodore Ts'o } else if (m->flags & MOPT_DATAJ) { 207026092bf5STheodore Ts'o if (is_remount) { 207126092bf5STheodore Ts'o if (!sbi->s_journal) 207226092bf5STheodore Ts'o ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 20735f3633e3SJan Kara else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 207426092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, 207526092bf5STheodore Ts'o "Cannot change data mode on remount"); 207626092bf5STheodore Ts'o return -1; 207726092bf5STheodore Ts'o } 207826092bf5STheodore Ts'o } else { 207926092bf5STheodore Ts'o clear_opt(sb, DATA_FLAGS); 208026092bf5STheodore Ts'o sbi->s_mount_opt |= m->mount_opt; 208126092bf5STheodore Ts'o } 208226092bf5STheodore Ts'o #ifdef CONFIG_QUOTA 208326092bf5STheodore Ts'o } else if (m->flags & MOPT_QFMT) { 208426092bf5STheodore Ts'o if (sb_any_quota_loaded(sb) && 208526092bf5STheodore Ts'o sbi->s_jquota_fmt != m->mount_opt) { 20865f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Cannot change journaled " 20875f3633e3SJan Kara "quota options when quota turned on"); 208826092bf5STheodore Ts'o return -1; 208926092bf5STheodore Ts'o } 2090e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) { 2091c325a67cSTheodore Ts'o ext4_msg(sb, KERN_INFO, 2092c325a67cSTheodore Ts'o "Quota format mount options ignored " 2093262b4662SJan Kara "when QUOTA feature is enabled"); 2094c325a67cSTheodore Ts'o return 1; 2095262b4662SJan Kara } 209626092bf5STheodore Ts'o sbi->s_jquota_fmt = m->mount_opt; 209726092bf5STheodore Ts'o #endif 2098*9cb20f94SIra Weiny } else if (token == Opt_dax || token == Opt_dax_always || 2099*9cb20f94SIra Weiny token == Opt_dax_inode || token == Opt_dax_never) { 2100ef83b6e8SDan Williams #ifdef CONFIG_FS_DAX 2101*9cb20f94SIra Weiny switch (token) { 2102*9cb20f94SIra Weiny case Opt_dax: 2103*9cb20f94SIra Weiny case Opt_dax_always: 2104ef83b6e8SDan Williams ext4_msg(sb, KERN_WARNING, 2105ef83b6e8SDan Williams "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); 2106*9cb20f94SIra Weiny sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS; 2107*9cb20f94SIra Weiny sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER; 2108*9cb20f94SIra Weiny break; 2109*9cb20f94SIra Weiny case Opt_dax_never: 2110*9cb20f94SIra Weiny sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER; 2111*9cb20f94SIra Weiny sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 2112*9cb20f94SIra Weiny break; 2113*9cb20f94SIra Weiny case Opt_dax_inode: 2114*9cb20f94SIra Weiny sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 2115*9cb20f94SIra Weiny sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER; 2116*9cb20f94SIra Weiny /* Strictly for printing options */ 2117*9cb20f94SIra Weiny sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE; 2118*9cb20f94SIra Weiny break; 2119*9cb20f94SIra Weiny } 2120ef83b6e8SDan Williams #else 2121923ae0ffSRoss Zwisler ext4_msg(sb, KERN_INFO, "dax option not supported"); 2122*9cb20f94SIra Weiny sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER; 2123*9cb20f94SIra Weiny sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 2124923ae0ffSRoss Zwisler return -1; 2125923ae0ffSRoss Zwisler #endif 21267915a861SAles Novak } else if (token == Opt_data_err_abort) { 21277915a861SAles Novak sbi->s_mount_opt |= m->mount_opt; 21287915a861SAles Novak } else if (token == Opt_data_err_ignore) { 21297915a861SAles Novak sbi->s_mount_opt &= ~m->mount_opt; 213026092bf5STheodore Ts'o } else { 213126092bf5STheodore Ts'o if (!args->from) 213226092bf5STheodore Ts'o arg = 1; 213326092bf5STheodore Ts'o if (m->flags & MOPT_CLEAR) 213426092bf5STheodore Ts'o arg = !arg; 213526092bf5STheodore Ts'o else if (unlikely(!(m->flags & MOPT_SET))) { 213626092bf5STheodore Ts'o ext4_msg(sb, KERN_WARNING, 213726092bf5STheodore Ts'o "buggy handling of option %s", opt); 213826092bf5STheodore Ts'o WARN_ON(1); 213926092bf5STheodore Ts'o return -1; 214026092bf5STheodore Ts'o } 214126092bf5STheodore Ts'o if (arg != 0) 214226092bf5STheodore Ts'o sbi->s_mount_opt |= m->mount_opt; 214326092bf5STheodore Ts'o else 214426092bf5STheodore Ts'o sbi->s_mount_opt &= ~m->mount_opt; 214526092bf5STheodore Ts'o } 214626092bf5STheodore Ts'o return 1; 214726092bf5STheodore Ts'o } 214826092bf5STheodore Ts'o 2149ac27a0ecSDave Kleikamp static int parse_options(char *options, struct super_block *sb, 2150c3191067STheodore Ts'o unsigned long *journal_devnum, 2151b3881f74STheodore Ts'o unsigned int *journal_ioprio, 2152661aa520SEric Sandeen int is_remount) 2153ac27a0ecSDave Kleikamp { 21541e1a76edSOlof Johansson struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb); 215533458eabSTheodore Ts'o char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name; 2156ac27a0ecSDave Kleikamp substring_t args[MAX_OPT_ARGS]; 215726092bf5STheodore Ts'o int token; 2158ac27a0ecSDave Kleikamp 2159ac27a0ecSDave Kleikamp if (!options) 2160ac27a0ecSDave Kleikamp return 1; 2161ac27a0ecSDave Kleikamp 2162ac27a0ecSDave Kleikamp while ((p = strsep(&options, ",")) != NULL) { 2163ac27a0ecSDave Kleikamp if (!*p) 2164ac27a0ecSDave Kleikamp continue; 216515121c18SEric Sandeen /* 216615121c18SEric Sandeen * Initialize args struct so we know whether arg was 216715121c18SEric Sandeen * found; some options take optional arguments. 216815121c18SEric Sandeen */ 2169caecd0afSSachin Kamat args[0].to = args[0].from = NULL; 2170ac27a0ecSDave Kleikamp token = match_token(p, tokens, args); 217126092bf5STheodore Ts'o if (handle_mount_opt(sb, p, token, args, journal_devnum, 217226092bf5STheodore Ts'o journal_ioprio, is_remount) < 0) 2173ac27a0ecSDave Kleikamp return 0; 2174ac27a0ecSDave Kleikamp } 2175ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 217649da9392SJan Kara /* 217749da9392SJan Kara * We do the test below only for project quotas. 'usrquota' and 217849da9392SJan Kara * 'grpquota' mount options are allowed even without quota feature 217949da9392SJan Kara * to support legacy quotas in quota files. 218049da9392SJan Kara */ 218149da9392SJan Kara if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) { 218249da9392SJan Kara ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. " 218349da9392SJan Kara "Cannot enable project quota enforcement."); 218449da9392SJan Kara return 0; 218549da9392SJan Kara } 218633458eabSTheodore Ts'o usr_qf_name = get_qf_name(sb, sbi, USRQUOTA); 218733458eabSTheodore Ts'o grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA); 218833458eabSTheodore Ts'o if (usr_qf_name || grp_qf_name) { 218933458eabSTheodore Ts'o if (test_opt(sb, USRQUOTA) && usr_qf_name) 2190fd8c37ecSTheodore Ts'o clear_opt(sb, USRQUOTA); 2191ac27a0ecSDave Kleikamp 219233458eabSTheodore Ts'o if (test_opt(sb, GRPQUOTA) && grp_qf_name) 2193fd8c37ecSTheodore Ts'o clear_opt(sb, GRPQUOTA); 2194ac27a0ecSDave Kleikamp 219556c50f11SDmitry Monakhov if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 2196b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "old and new quota " 2197b31e1552SEric Sandeen "format mixing"); 2198ac27a0ecSDave Kleikamp return 0; 2199ac27a0ecSDave Kleikamp } 2200ac27a0ecSDave Kleikamp 2201ac27a0ecSDave Kleikamp if (!sbi->s_jquota_fmt) { 2202b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journaled quota format " 2203b31e1552SEric Sandeen "not specified"); 2204ac27a0ecSDave Kleikamp return 0; 2205ac27a0ecSDave Kleikamp } 2206ac27a0ecSDave Kleikamp } 2207ac27a0ecSDave Kleikamp #endif 2208626b035bSRitesh Harjani if (test_opt(sb, DIOREAD_NOLOCK)) { 2209626b035bSRitesh Harjani int blocksize = 2210626b035bSRitesh Harjani BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 2211626b035bSRitesh Harjani if (blocksize < PAGE_SIZE) 2212626b035bSRitesh Harjani ext4_msg(sb, KERN_WARNING, "Warning: mounting with an " 2213626b035bSRitesh Harjani "experimental mount option 'dioread_nolock' " 2214626b035bSRitesh Harjani "for blocksize < PAGE_SIZE"); 2215626b035bSRitesh Harjani } 2216ac27a0ecSDave Kleikamp return 1; 2217ac27a0ecSDave Kleikamp } 2218ac27a0ecSDave Kleikamp 22192adf6da8STheodore Ts'o static inline void ext4_show_quota_options(struct seq_file *seq, 22202adf6da8STheodore Ts'o struct super_block *sb) 22212adf6da8STheodore Ts'o { 22222adf6da8STheodore Ts'o #if defined(CONFIG_QUOTA) 22232adf6da8STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 222433458eabSTheodore Ts'o char *usr_qf_name, *grp_qf_name; 22252adf6da8STheodore Ts'o 22262adf6da8STheodore Ts'o if (sbi->s_jquota_fmt) { 22272adf6da8STheodore Ts'o char *fmtname = ""; 22282adf6da8STheodore Ts'o 22292adf6da8STheodore Ts'o switch (sbi->s_jquota_fmt) { 22302adf6da8STheodore Ts'o case QFMT_VFS_OLD: 22312adf6da8STheodore Ts'o fmtname = "vfsold"; 22322adf6da8STheodore Ts'o break; 22332adf6da8STheodore Ts'o case QFMT_VFS_V0: 22342adf6da8STheodore Ts'o fmtname = "vfsv0"; 22352adf6da8STheodore Ts'o break; 22362adf6da8STheodore Ts'o case QFMT_VFS_V1: 22372adf6da8STheodore Ts'o fmtname = "vfsv1"; 22382adf6da8STheodore Ts'o break; 22392adf6da8STheodore Ts'o } 22402adf6da8STheodore Ts'o seq_printf(seq, ",jqfmt=%s", fmtname); 22412adf6da8STheodore Ts'o } 22422adf6da8STheodore Ts'o 224333458eabSTheodore Ts'o rcu_read_lock(); 224433458eabSTheodore Ts'o usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 224533458eabSTheodore Ts'o grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 224633458eabSTheodore Ts'o if (usr_qf_name) 224733458eabSTheodore Ts'o seq_show_option(seq, "usrjquota", usr_qf_name); 224833458eabSTheodore Ts'o if (grp_qf_name) 224933458eabSTheodore Ts'o seq_show_option(seq, "grpjquota", grp_qf_name); 225033458eabSTheodore Ts'o rcu_read_unlock(); 22512adf6da8STheodore Ts'o #endif 22522adf6da8STheodore Ts'o } 22532adf6da8STheodore Ts'o 22545a916be1STheodore Ts'o static const char *token2str(int token) 22555a916be1STheodore Ts'o { 225650df9fd5SHerton Ronaldo Krzesinski const struct match_token *t; 22575a916be1STheodore Ts'o 22585a916be1STheodore Ts'o for (t = tokens; t->token != Opt_err; t++) 22595a916be1STheodore Ts'o if (t->token == token && !strchr(t->pattern, '=')) 22605a916be1STheodore Ts'o break; 22615a916be1STheodore Ts'o return t->pattern; 22625a916be1STheodore Ts'o } 22635a916be1STheodore Ts'o 22642adf6da8STheodore Ts'o /* 22652adf6da8STheodore Ts'o * Show an option if 22662adf6da8STheodore Ts'o * - it's set to a non-default value OR 22672adf6da8STheodore Ts'o * - if the per-sb default is different from the global default 22682adf6da8STheodore Ts'o */ 226966acdcf4STheodore Ts'o static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 227066acdcf4STheodore Ts'o int nodefs) 22712adf6da8STheodore Ts'o { 22722adf6da8STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 22732adf6da8STheodore Ts'o struct ext4_super_block *es = sbi->s_es; 227468afa7e0STyson Nottingham int def_errors, def_mount_opt = sbi->s_def_mount_opt; 22755a916be1STheodore Ts'o const struct mount_opts *m; 227666acdcf4STheodore Ts'o char sep = nodefs ? '\n' : ','; 22772adf6da8STheodore Ts'o 227866acdcf4STheodore Ts'o #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 227966acdcf4STheodore Ts'o #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 22802adf6da8STheodore Ts'o 22812adf6da8STheodore Ts'o if (sbi->s_sb_block != 1) 22825a916be1STheodore Ts'o SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 22835a916be1STheodore Ts'o 22845a916be1STheodore Ts'o for (m = ext4_mount_opts; m->token != Opt_err; m++) { 22855a916be1STheodore Ts'o int want_set = m->flags & MOPT_SET; 22865a916be1STheodore Ts'o if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 2287*9cb20f94SIra Weiny (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP) 22885a916be1STheodore Ts'o continue; 228968afa7e0STyson Nottingham if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 22905a916be1STheodore Ts'o continue; /* skip if same as the default */ 22915a916be1STheodore Ts'o if ((want_set && 22925a916be1STheodore Ts'o (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 22935a916be1STheodore Ts'o (!want_set && (sbi->s_mount_opt & m->mount_opt))) 22945a916be1STheodore Ts'o continue; /* select Opt_noFoo vs Opt_Foo */ 22955a916be1STheodore Ts'o SEQ_OPTS_PRINT("%s", token2str(m->token)); 22965a916be1STheodore Ts'o } 22975a916be1STheodore Ts'o 229808cefc7aSEric W. Biederman if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 22995a916be1STheodore Ts'o le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 230008cefc7aSEric W. Biederman SEQ_OPTS_PRINT("resuid=%u", 230108cefc7aSEric W. Biederman from_kuid_munged(&init_user_ns, sbi->s_resuid)); 230208cefc7aSEric W. Biederman if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 23035a916be1STheodore Ts'o le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 230408cefc7aSEric W. Biederman SEQ_OPTS_PRINT("resgid=%u", 230508cefc7aSEric W. Biederman from_kgid_munged(&init_user_ns, sbi->s_resgid)); 230666acdcf4STheodore Ts'o def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 23075a916be1STheodore Ts'o if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 23085a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=remount-ro"); 23092adf6da8STheodore Ts'o if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 23105a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=continue"); 23112adf6da8STheodore Ts'o if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 23125a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=panic"); 231366acdcf4STheodore Ts'o if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 23145a916be1STheodore Ts'o SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 231566acdcf4STheodore Ts'o if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 23165a916be1STheodore Ts'o SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 231766acdcf4STheodore Ts'o if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 23185a916be1STheodore Ts'o SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 2319357fdad0SMatthew Garrett if (sb->s_flags & SB_I_VERSION) 23205a916be1STheodore Ts'o SEQ_OPTS_PUTS("i_version"); 232166acdcf4STheodore Ts'o if (nodefs || sbi->s_stripe) 23225a916be1STheodore Ts'o SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 232368afa7e0STyson Nottingham if (nodefs || EXT4_MOUNT_DATA_FLAGS & 232468afa7e0STyson Nottingham (sbi->s_mount_opt ^ def_mount_opt)) { 23252adf6da8STheodore Ts'o if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 23265a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=journal"); 23272adf6da8STheodore Ts'o else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 23285a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=ordered"); 23292adf6da8STheodore Ts'o else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 23305a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=writeback"); 23315a916be1STheodore Ts'o } 233266acdcf4STheodore Ts'o if (nodefs || 233366acdcf4STheodore Ts'o sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 23345a916be1STheodore Ts'o SEQ_OPTS_PRINT("inode_readahead_blks=%u", 23352adf6da8STheodore Ts'o sbi->s_inode_readahead_blks); 23362adf6da8STheodore Ts'o 2337ceec0376STyson Nottingham if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 233866acdcf4STheodore Ts'o (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 23395a916be1STheodore Ts'o SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 2340df981d03STheodore Ts'o if (nodefs || sbi->s_max_dir_size_kb) 2341df981d03STheodore Ts'o SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 23427915a861SAles Novak if (test_opt(sb, DATA_ERR_ABORT)) 23437915a861SAles Novak SEQ_OPTS_PUTS("data_err=abort"); 2344338affb5SEric Biggers if (DUMMY_ENCRYPTION_ENABLED(sbi)) 2345338affb5SEric Biggers SEQ_OPTS_PUTS("test_dummy_encryption"); 23462adf6da8STheodore Ts'o 2347*9cb20f94SIra Weiny if (test_opt(sb, DAX_ALWAYS)) { 2348*9cb20f94SIra Weiny if (IS_EXT2_SB(sb)) 2349*9cb20f94SIra Weiny SEQ_OPTS_PUTS("dax"); 2350*9cb20f94SIra Weiny else 2351*9cb20f94SIra Weiny SEQ_OPTS_PUTS("dax=always"); 2352*9cb20f94SIra Weiny } else if (test_opt2(sb, DAX_NEVER)) { 2353*9cb20f94SIra Weiny SEQ_OPTS_PUTS("dax=never"); 2354*9cb20f94SIra Weiny } else if (test_opt2(sb, DAX_INODE)) { 2355*9cb20f94SIra Weiny SEQ_OPTS_PUTS("dax=inode"); 2356*9cb20f94SIra Weiny } 2357*9cb20f94SIra Weiny 23582adf6da8STheodore Ts'o ext4_show_quota_options(seq, sb); 23592adf6da8STheodore Ts'o return 0; 23602adf6da8STheodore Ts'o } 23612adf6da8STheodore Ts'o 236266acdcf4STheodore Ts'o static int ext4_show_options(struct seq_file *seq, struct dentry *root) 236366acdcf4STheodore Ts'o { 236466acdcf4STheodore Ts'o return _ext4_show_options(seq, root->d_sb, 0); 236566acdcf4STheodore Ts'o } 236666acdcf4STheodore Ts'o 2367ebd173beSTheodore Ts'o int ext4_seq_options_show(struct seq_file *seq, void *offset) 236866acdcf4STheodore Ts'o { 236966acdcf4STheodore Ts'o struct super_block *sb = seq->private; 237066acdcf4STheodore Ts'o int rc; 237166acdcf4STheodore Ts'o 2372bc98a42cSDavid Howells seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 237366acdcf4STheodore Ts'o rc = _ext4_show_options(seq, sb, 1); 237466acdcf4STheodore Ts'o seq_puts(seq, "\n"); 237566acdcf4STheodore Ts'o return rc; 237666acdcf4STheodore Ts'o } 237766acdcf4STheodore Ts'o 2378617ba13bSMingming Cao static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 2379ac27a0ecSDave Kleikamp int read_only) 2380ac27a0ecSDave Kleikamp { 2381617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 2382c89128a0SJaegeuk Kim int err = 0; 2383ac27a0ecSDave Kleikamp 2384617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 2385b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "revision level too high, " 2386b31e1552SEric Sandeen "forcing read-only mode"); 2387c89128a0SJaegeuk Kim err = -EROFS; 2388ac27a0ecSDave Kleikamp } 2389ac27a0ecSDave Kleikamp if (read_only) 2390281b5995STheodore Ts'o goto done; 2391617ba13bSMingming Cao if (!(sbi->s_mount_state & EXT4_VALID_FS)) 2392b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 2393b31e1552SEric Sandeen "running e2fsck is recommended"); 2394c8b459f4SLukas Czerner else if (sbi->s_mount_state & EXT4_ERROR_FS) 2395b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 2396b31e1552SEric Sandeen "warning: mounting fs with errors, " 2397b31e1552SEric Sandeen "running e2fsck is recommended"); 2398ed3ce80aSTao Ma else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 2399ac27a0ecSDave Kleikamp le16_to_cpu(es->s_mnt_count) >= 2400ac27a0ecSDave Kleikamp (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 2401b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 2402b31e1552SEric Sandeen "warning: maximal mount count reached, " 2403b31e1552SEric Sandeen "running e2fsck is recommended"); 2404ac27a0ecSDave Kleikamp else if (le32_to_cpu(es->s_checkinterval) && 24056a0678a7SArnd Bergmann (ext4_get_tstamp(es, s_lastcheck) + 24066a0678a7SArnd Bergmann le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 2407b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 2408b31e1552SEric Sandeen "warning: checktime reached, " 2409b31e1552SEric Sandeen "running e2fsck is recommended"); 24100390131bSFrank Mayhar if (!sbi->s_journal) 2411216c34b2SMarcin Slusarz es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 2412ac27a0ecSDave Kleikamp if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 2413617ba13bSMingming Cao es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 2414e8546d06SMarcin Slusarz le16_add_cpu(&es->s_mnt_count, 1); 24156a0678a7SArnd Bergmann ext4_update_tstamp(es, s_mtime); 24160390131bSFrank Mayhar if (sbi->s_journal) 2417e2b911c5SDarrick J. Wong ext4_set_feature_journal_needs_recovery(sb); 2418ac27a0ecSDave Kleikamp 2419c89128a0SJaegeuk Kim err = ext4_commit_super(sb, 1); 2420281b5995STheodore Ts'o done: 2421ac27a0ecSDave Kleikamp if (test_opt(sb, DEBUG)) 2422a9df9a49STheodore Ts'o printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 2423a2595b8aSTheodore Ts'o "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 2424ac27a0ecSDave Kleikamp sb->s_blocksize, 2425ac27a0ecSDave Kleikamp sbi->s_groups_count, 2426617ba13bSMingming Cao EXT4_BLOCKS_PER_GROUP(sb), 2427617ba13bSMingming Cao EXT4_INODES_PER_GROUP(sb), 2428a2595b8aSTheodore Ts'o sbi->s_mount_opt, sbi->s_mount_opt2); 2429ac27a0ecSDave Kleikamp 24307abc52c2SDan Magenheimer cleancache_init_fs(sb); 2431c89128a0SJaegeuk Kim return err; 2432ac27a0ecSDave Kleikamp } 2433ac27a0ecSDave Kleikamp 2434117fff10STheodore Ts'o int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 2435117fff10STheodore Ts'o { 2436117fff10STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 24377c990728SSuraj Jitindar Singh struct flex_groups **old_groups, **new_groups; 243837b0b6b8SDan Carpenter int size, i, j; 2439117fff10STheodore Ts'o 2440117fff10STheodore Ts'o if (!sbi->s_log_groups_per_flex) 2441117fff10STheodore Ts'o return 0; 2442117fff10STheodore Ts'o 2443117fff10STheodore Ts'o size = ext4_flex_group(sbi, ngroup - 1) + 1; 2444117fff10STheodore Ts'o if (size <= sbi->s_flex_groups_allocated) 2445117fff10STheodore Ts'o return 0; 2446117fff10STheodore Ts'o 24477c990728SSuraj Jitindar Singh new_groups = kvzalloc(roundup_pow_of_two(size * 24487c990728SSuraj Jitindar Singh sizeof(*sbi->s_flex_groups)), GFP_KERNEL); 2449117fff10STheodore Ts'o if (!new_groups) { 24507c990728SSuraj Jitindar Singh ext4_msg(sb, KERN_ERR, 24517c990728SSuraj Jitindar Singh "not enough memory for %d flex group pointers", size); 2452117fff10STheodore Ts'o return -ENOMEM; 2453117fff10STheodore Ts'o } 24547c990728SSuraj Jitindar Singh for (i = sbi->s_flex_groups_allocated; i < size; i++) { 24557c990728SSuraj Jitindar Singh new_groups[i] = kvzalloc(roundup_pow_of_two( 24567c990728SSuraj Jitindar Singh sizeof(struct flex_groups)), 24577c990728SSuraj Jitindar Singh GFP_KERNEL); 24587c990728SSuraj Jitindar Singh if (!new_groups[i]) { 245937b0b6b8SDan Carpenter for (j = sbi->s_flex_groups_allocated; j < i; j++) 246037b0b6b8SDan Carpenter kvfree(new_groups[j]); 24617c990728SSuraj Jitindar Singh kvfree(new_groups); 24627c990728SSuraj Jitindar Singh ext4_msg(sb, KERN_ERR, 24637c990728SSuraj Jitindar Singh "not enough memory for %d flex groups", size); 24647c990728SSuraj Jitindar Singh return -ENOMEM; 2465117fff10STheodore Ts'o } 24667c990728SSuraj Jitindar Singh } 24677c990728SSuraj Jitindar Singh rcu_read_lock(); 24687c990728SSuraj Jitindar Singh old_groups = rcu_dereference(sbi->s_flex_groups); 24697c990728SSuraj Jitindar Singh if (old_groups) 24707c990728SSuraj Jitindar Singh memcpy(new_groups, old_groups, 24717c990728SSuraj Jitindar Singh (sbi->s_flex_groups_allocated * 24727c990728SSuraj Jitindar Singh sizeof(struct flex_groups *))); 24737c990728SSuraj Jitindar Singh rcu_read_unlock(); 24747c990728SSuraj Jitindar Singh rcu_assign_pointer(sbi->s_flex_groups, new_groups); 24757c990728SSuraj Jitindar Singh sbi->s_flex_groups_allocated = size; 24767c990728SSuraj Jitindar Singh if (old_groups) 24777c990728SSuraj Jitindar Singh ext4_kvfree_array_rcu(old_groups); 2478117fff10STheodore Ts'o return 0; 2479117fff10STheodore Ts'o } 2480117fff10STheodore Ts'o 2481772cb7c8SJose R. Santos static int ext4_fill_flex_info(struct super_block *sb) 2482772cb7c8SJose R. Santos { 2483772cb7c8SJose R. Santos struct ext4_sb_info *sbi = EXT4_SB(sb); 2484772cb7c8SJose R. Santos struct ext4_group_desc *gdp = NULL; 24857c990728SSuraj Jitindar Singh struct flex_groups *fg; 2486772cb7c8SJose R. Santos ext4_group_t flex_group; 2487117fff10STheodore Ts'o int i, err; 2488772cb7c8SJose R. Santos 2489503358aeSTheodore Ts'o sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 2490d50f2ab6SXi Wang if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 2491772cb7c8SJose R. Santos sbi->s_log_groups_per_flex = 0; 2492772cb7c8SJose R. Santos return 1; 2493772cb7c8SJose R. Santos } 2494772cb7c8SJose R. Santos 2495117fff10STheodore Ts'o err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 2496117fff10STheodore Ts'o if (err) 2497772cb7c8SJose R. Santos goto failed; 2498772cb7c8SJose R. Santos 2499772cb7c8SJose R. Santos for (i = 0; i < sbi->s_groups_count; i++) { 250088b6edd1STheodore Ts'o gdp = ext4_get_group_desc(sb, i, NULL); 2501772cb7c8SJose R. Santos 2502772cb7c8SJose R. Santos flex_group = ext4_flex_group(sbi, i); 25037c990728SSuraj Jitindar Singh fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group); 25047c990728SSuraj Jitindar Singh atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes); 250590ba983fSTheodore Ts'o atomic64_add(ext4_free_group_clusters(sb, gdp), 25067c990728SSuraj Jitindar Singh &fg->free_clusters); 25077c990728SSuraj Jitindar Singh atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs); 2508772cb7c8SJose R. Santos } 2509772cb7c8SJose R. Santos 2510772cb7c8SJose R. Santos return 1; 2511772cb7c8SJose R. Santos failed: 2512772cb7c8SJose R. Santos return 0; 2513772cb7c8SJose R. Santos } 2514772cb7c8SJose R. Santos 2515e2b911c5SDarrick J. Wong static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 2516717d50e4SAndreas Dilger struct ext4_group_desc *gdp) 2517717d50e4SAndreas Dilger { 2518b47820edSDaeho Jeong int offset = offsetof(struct ext4_group_desc, bg_checksum); 2519717d50e4SAndreas Dilger __u16 crc = 0; 2520717d50e4SAndreas Dilger __le32 le_group = cpu_to_le32(block_group); 2521e2b911c5SDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 2522717d50e4SAndreas Dilger 25239aa5d32bSDmitry Monakhov if (ext4_has_metadata_csum(sbi->s_sb)) { 2524feb0ab32SDarrick J. Wong /* Use new metadata_csum algorithm */ 2525feb0ab32SDarrick J. Wong __u32 csum32; 2526b47820edSDaeho Jeong __u16 dummy_csum = 0; 2527feb0ab32SDarrick J. Wong 2528feb0ab32SDarrick J. Wong csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 2529feb0ab32SDarrick J. Wong sizeof(le_group)); 2530b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 2531b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 2532b47820edSDaeho Jeong sizeof(dummy_csum)); 2533b47820edSDaeho Jeong offset += sizeof(dummy_csum); 2534b47820edSDaeho Jeong if (offset < sbi->s_desc_size) 2535b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 2536b47820edSDaeho Jeong sbi->s_desc_size - offset); 2537feb0ab32SDarrick J. Wong 2538feb0ab32SDarrick J. Wong crc = csum32 & 0xFFFF; 2539feb0ab32SDarrick J. Wong goto out; 2540feb0ab32SDarrick J. Wong } 2541feb0ab32SDarrick J. Wong 2542feb0ab32SDarrick J. Wong /* old crc16 code */ 2543e2b911c5SDarrick J. Wong if (!ext4_has_feature_gdt_csum(sb)) 2544813d32f9SDarrick J. Wong return 0; 2545813d32f9SDarrick J. Wong 2546717d50e4SAndreas Dilger crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2547717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2548717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)gdp, offset); 2549717d50e4SAndreas Dilger offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2550717d50e4SAndreas Dilger /* for checksum of struct ext4_group_desc do the rest...*/ 2551e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb) && 2552717d50e4SAndreas Dilger offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2553717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)gdp + offset, 2554717d50e4SAndreas Dilger le16_to_cpu(sbi->s_es->s_desc_size) - 2555717d50e4SAndreas Dilger offset); 2556717d50e4SAndreas Dilger 2557feb0ab32SDarrick J. Wong out: 2558717d50e4SAndreas Dilger return cpu_to_le16(crc); 2559717d50e4SAndreas Dilger } 2560717d50e4SAndreas Dilger 2561feb0ab32SDarrick J. Wong int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2562717d50e4SAndreas Dilger struct ext4_group_desc *gdp) 2563717d50e4SAndreas Dilger { 2564feb0ab32SDarrick J. Wong if (ext4_has_group_desc_csum(sb) && 2565e2b911c5SDarrick J. Wong (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 2566717d50e4SAndreas Dilger return 0; 2567717d50e4SAndreas Dilger 2568717d50e4SAndreas Dilger return 1; 2569717d50e4SAndreas Dilger } 2570717d50e4SAndreas Dilger 2571feb0ab32SDarrick J. Wong void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2572feb0ab32SDarrick J. Wong struct ext4_group_desc *gdp) 2573feb0ab32SDarrick J. Wong { 2574feb0ab32SDarrick J. Wong if (!ext4_has_group_desc_csum(sb)) 2575feb0ab32SDarrick J. Wong return; 2576e2b911c5SDarrick J. Wong gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 2577feb0ab32SDarrick J. Wong } 2578feb0ab32SDarrick J. Wong 2579ac27a0ecSDave Kleikamp /* Called at mount-time, super-block is locked */ 2580bfff6873SLukas Czerner static int ext4_check_descriptors(struct super_block *sb, 2581829fa70dSTheodore Ts'o ext4_fsblk_t sb_block, 2582bfff6873SLukas Czerner ext4_group_t *first_not_zeroed) 2583ac27a0ecSDave Kleikamp { 2584617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 2585617ba13bSMingming Cao ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2586617ba13bSMingming Cao ext4_fsblk_t last_block; 258744de022cSTheodore Ts'o ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 2588bd81d8eeSLaurent Vivier ext4_fsblk_t block_bitmap; 2589bd81d8eeSLaurent Vivier ext4_fsblk_t inode_bitmap; 2590bd81d8eeSLaurent Vivier ext4_fsblk_t inode_table; 2591ce421581SJose R. Santos int flexbg_flag = 0; 2592bfff6873SLukas Czerner ext4_group_t i, grp = sbi->s_groups_count; 2593ac27a0ecSDave Kleikamp 2594e2b911c5SDarrick J. Wong if (ext4_has_feature_flex_bg(sb)) 2595ce421581SJose R. Santos flexbg_flag = 1; 2596ce421581SJose R. Santos 2597617ba13bSMingming Cao ext4_debug("Checking group descriptors"); 2598ac27a0ecSDave Kleikamp 2599197cd65aSAkinobu Mita for (i = 0; i < sbi->s_groups_count; i++) { 2600197cd65aSAkinobu Mita struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2601197cd65aSAkinobu Mita 2602ce421581SJose R. Santos if (i == sbi->s_groups_count - 1 || flexbg_flag) 2603bd81d8eeSLaurent Vivier last_block = ext4_blocks_count(sbi->s_es) - 1; 2604ac27a0ecSDave Kleikamp else 2605ac27a0ecSDave Kleikamp last_block = first_block + 2606617ba13bSMingming Cao (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2607ac27a0ecSDave Kleikamp 2608bfff6873SLukas Czerner if ((grp == sbi->s_groups_count) && 2609bfff6873SLukas Czerner !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2610bfff6873SLukas Czerner grp = i; 2611bfff6873SLukas Czerner 26128fadc143SAlexandre Ratchov block_bitmap = ext4_block_bitmap(sb, gdp); 2613829fa70dSTheodore Ts'o if (block_bitmap == sb_block) { 2614829fa70dSTheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2615829fa70dSTheodore Ts'o "Block bitmap for group %u overlaps " 2616829fa70dSTheodore Ts'o "superblock", i); 261718db4b4eSTheodore Ts'o if (!sb_rdonly(sb)) 261818db4b4eSTheodore Ts'o return 0; 2619829fa70dSTheodore Ts'o } 262077260807STheodore Ts'o if (block_bitmap >= sb_block + 1 && 262177260807STheodore Ts'o block_bitmap <= last_bg_block) { 262277260807STheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 262377260807STheodore Ts'o "Block bitmap for group %u overlaps " 262477260807STheodore Ts'o "block group descriptors", i); 262577260807STheodore Ts'o if (!sb_rdonly(sb)) 262677260807STheodore Ts'o return 0; 262777260807STheodore Ts'o } 26282b2d6d01STheodore Ts'o if (block_bitmap < first_block || block_bitmap > last_block) { 2629b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2630a9df9a49STheodore Ts'o "Block bitmap for group %u not in group " 2631b31e1552SEric Sandeen "(block %llu)!", i, block_bitmap); 2632ac27a0ecSDave Kleikamp return 0; 2633ac27a0ecSDave Kleikamp } 26348fadc143SAlexandre Ratchov inode_bitmap = ext4_inode_bitmap(sb, gdp); 2635829fa70dSTheodore Ts'o if (inode_bitmap == sb_block) { 2636829fa70dSTheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2637829fa70dSTheodore Ts'o "Inode bitmap for group %u overlaps " 2638829fa70dSTheodore Ts'o "superblock", i); 263918db4b4eSTheodore Ts'o if (!sb_rdonly(sb)) 264018db4b4eSTheodore Ts'o return 0; 2641829fa70dSTheodore Ts'o } 264277260807STheodore Ts'o if (inode_bitmap >= sb_block + 1 && 264377260807STheodore Ts'o inode_bitmap <= last_bg_block) { 264477260807STheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 264577260807STheodore Ts'o "Inode bitmap for group %u overlaps " 264677260807STheodore Ts'o "block group descriptors", i); 264777260807STheodore Ts'o if (!sb_rdonly(sb)) 264877260807STheodore Ts'o return 0; 264977260807STheodore Ts'o } 26502b2d6d01STheodore Ts'o if (inode_bitmap < first_block || inode_bitmap > last_block) { 2651b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2652a9df9a49STheodore Ts'o "Inode bitmap for group %u not in group " 2653b31e1552SEric Sandeen "(block %llu)!", i, inode_bitmap); 2654ac27a0ecSDave Kleikamp return 0; 2655ac27a0ecSDave Kleikamp } 26568fadc143SAlexandre Ratchov inode_table = ext4_inode_table(sb, gdp); 2657829fa70dSTheodore Ts'o if (inode_table == sb_block) { 2658829fa70dSTheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2659829fa70dSTheodore Ts'o "Inode table for group %u overlaps " 2660829fa70dSTheodore Ts'o "superblock", i); 266118db4b4eSTheodore Ts'o if (!sb_rdonly(sb)) 266218db4b4eSTheodore Ts'o return 0; 2663829fa70dSTheodore Ts'o } 266477260807STheodore Ts'o if (inode_table >= sb_block + 1 && 266577260807STheodore Ts'o inode_table <= last_bg_block) { 266677260807STheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 266777260807STheodore Ts'o "Inode table for group %u overlaps " 266877260807STheodore Ts'o "block group descriptors", i); 266977260807STheodore Ts'o if (!sb_rdonly(sb)) 267077260807STheodore Ts'o return 0; 267177260807STheodore Ts'o } 2672bd81d8eeSLaurent Vivier if (inode_table < first_block || 26732b2d6d01STheodore Ts'o inode_table + sbi->s_itb_per_group - 1 > last_block) { 2674b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2675a9df9a49STheodore Ts'o "Inode table for group %u not in group " 2676b31e1552SEric Sandeen "(block %llu)!", i, inode_table); 2677ac27a0ecSDave Kleikamp return 0; 2678ac27a0ecSDave Kleikamp } 2679955ce5f5SAneesh Kumar K.V ext4_lock_group(sb, i); 2680feb0ab32SDarrick J. Wong if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2681b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2682b31e1552SEric Sandeen "Checksum for group %u failed (%u!=%u)", 2683e2b911c5SDarrick J. Wong i, le16_to_cpu(ext4_group_desc_csum(sb, i, 2684fd2d4291SAvantika Mathur gdp)), le16_to_cpu(gdp->bg_checksum)); 2685bc98a42cSDavid Howells if (!sb_rdonly(sb)) { 2686955ce5f5SAneesh Kumar K.V ext4_unlock_group(sb, i); 2687717d50e4SAndreas Dilger return 0; 2688717d50e4SAndreas Dilger } 26897ee1ec4cSLi Zefan } 2690955ce5f5SAneesh Kumar K.V ext4_unlock_group(sb, i); 2691ce421581SJose R. Santos if (!flexbg_flag) 2692617ba13bSMingming Cao first_block += EXT4_BLOCKS_PER_GROUP(sb); 2693ac27a0ecSDave Kleikamp } 2694bfff6873SLukas Czerner if (NULL != first_not_zeroed) 2695bfff6873SLukas Czerner *first_not_zeroed = grp; 2696ac27a0ecSDave Kleikamp return 1; 2697ac27a0ecSDave Kleikamp } 2698ac27a0ecSDave Kleikamp 2699617ba13bSMingming Cao /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2700ac27a0ecSDave Kleikamp * the superblock) which were deleted from all directories, but held open by 2701ac27a0ecSDave Kleikamp * a process at the time of a crash. We walk the list and try to delete these 2702ac27a0ecSDave Kleikamp * inodes at recovery time (only with a read-write filesystem). 2703ac27a0ecSDave Kleikamp * 2704ac27a0ecSDave Kleikamp * In order to keep the orphan inode chain consistent during traversal (in 2705ac27a0ecSDave Kleikamp * case of crash during recovery), we link each inode into the superblock 2706ac27a0ecSDave Kleikamp * orphan list_head and handle it the same way as an inode deletion during 2707ac27a0ecSDave Kleikamp * normal operation (which journals the operations for us). 2708ac27a0ecSDave Kleikamp * 2709ac27a0ecSDave Kleikamp * We only do an iget() and an iput() on each inode, which is very safe if we 2710ac27a0ecSDave Kleikamp * accidentally point at an in-use or already deleted inode. The worst that 2711ac27a0ecSDave Kleikamp * can happen in this case is that we get a "bit already cleared" message from 2712617ba13bSMingming Cao * ext4_free_inode(). The only reason we would point at a wrong inode is if 2713ac27a0ecSDave Kleikamp * e2fsck was run on this filesystem, and it must have already done the orphan 2714ac27a0ecSDave Kleikamp * inode cleanup for us, so we can safely abort without any further action. 2715ac27a0ecSDave Kleikamp */ 2716617ba13bSMingming Cao static void ext4_orphan_cleanup(struct super_block *sb, 2717617ba13bSMingming Cao struct ext4_super_block *es) 2718ac27a0ecSDave Kleikamp { 2719ac27a0ecSDave Kleikamp unsigned int s_flags = sb->s_flags; 27202c98eb5eSTheodore Ts'o int ret, nr_orphans = 0, nr_truncates = 0; 2721ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 272295f1fda4Szhangyi (F) int quota_update = 0; 2723ac27a0ecSDave Kleikamp int i; 2724ac27a0ecSDave Kleikamp #endif 2725ac27a0ecSDave Kleikamp if (!es->s_last_orphan) { 2726ac27a0ecSDave Kleikamp jbd_debug(4, "no orphan inodes to clean up\n"); 2727ac27a0ecSDave Kleikamp return; 2728ac27a0ecSDave Kleikamp } 2729ac27a0ecSDave Kleikamp 2730a8f48a95SEric Sandeen if (bdev_read_only(sb->s_bdev)) { 2731b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "write access " 2732b31e1552SEric Sandeen "unavailable, skipping orphan cleanup"); 2733a8f48a95SEric Sandeen return; 2734a8f48a95SEric Sandeen } 2735a8f48a95SEric Sandeen 2736d39195c3SAmir Goldstein /* Check if feature set would not allow a r/w mount */ 2737d39195c3SAmir Goldstein if (!ext4_feature_set_ok(sb, 0)) { 2738d39195c3SAmir Goldstein ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2739d39195c3SAmir Goldstein "unknown ROCOMPAT features"); 2740d39195c3SAmir Goldstein return; 2741d39195c3SAmir Goldstein } 2742d39195c3SAmir Goldstein 2743617ba13bSMingming Cao if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2744c25f9bc6SEric Sandeen /* don't clear list on RO mount w/ errors */ 27451751e8a6SLinus Torvalds if (es->s_last_orphan && !(s_flags & SB_RDONLY)) { 274684474976SDmitry Monakhov ext4_msg(sb, KERN_INFO, "Errors on filesystem, " 2747ac27a0ecSDave Kleikamp "clearing orphan list.\n"); 2748ac27a0ecSDave Kleikamp es->s_last_orphan = 0; 2749c25f9bc6SEric Sandeen } 2750ac27a0ecSDave Kleikamp jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2751ac27a0ecSDave Kleikamp return; 2752ac27a0ecSDave Kleikamp } 2753ac27a0ecSDave Kleikamp 27541751e8a6SLinus Torvalds if (s_flags & SB_RDONLY) { 2755b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 27561751e8a6SLinus Torvalds sb->s_flags &= ~SB_RDONLY; 2757ac27a0ecSDave Kleikamp } 2758ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 2759ac27a0ecSDave Kleikamp /* Needed for iput() to work correctly and not trash data */ 27601751e8a6SLinus Torvalds sb->s_flags |= SB_ACTIVE; 276195f1fda4Szhangyi (F) 276295f1fda4Szhangyi (F) /* 276395f1fda4Szhangyi (F) * Turn on quotas which were not enabled for read-only mounts if 276495f1fda4Szhangyi (F) * filesystem has quota feature, so that they are updated correctly. 276595f1fda4Szhangyi (F) */ 27661751e8a6SLinus Torvalds if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) { 276795f1fda4Szhangyi (F) int ret = ext4_enable_quotas(sb); 276895f1fda4Szhangyi (F) 276995f1fda4Szhangyi (F) if (!ret) 277095f1fda4Szhangyi (F) quota_update = 1; 277195f1fda4Szhangyi (F) else 277295f1fda4Szhangyi (F) ext4_msg(sb, KERN_ERR, 277395f1fda4Szhangyi (F) "Cannot turn on quotas: error %d", ret); 277495f1fda4Szhangyi (F) } 277595f1fda4Szhangyi (F) 277695f1fda4Szhangyi (F) /* Turn on journaled quotas used for old sytle */ 2777a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2778617ba13bSMingming Cao if (EXT4_SB(sb)->s_qf_names[i]) { 2779617ba13bSMingming Cao int ret = ext4_quota_on_mount(sb, i); 278095f1fda4Szhangyi (F) 278195f1fda4Szhangyi (F) if (!ret) 278295f1fda4Szhangyi (F) quota_update = 1; 278395f1fda4Szhangyi (F) else 2784b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 2785b31e1552SEric Sandeen "Cannot turn on journaled " 278695f1fda4Szhangyi (F) "quota: type %d: error %d", i, ret); 2787ac27a0ecSDave Kleikamp } 2788ac27a0ecSDave Kleikamp } 2789ac27a0ecSDave Kleikamp #endif 2790ac27a0ecSDave Kleikamp 2791ac27a0ecSDave Kleikamp while (es->s_last_orphan) { 2792ac27a0ecSDave Kleikamp struct inode *inode; 2793ac27a0ecSDave Kleikamp 2794c65d5c6cSVegard Nossum /* 2795c65d5c6cSVegard Nossum * We may have encountered an error during cleanup; if 2796c65d5c6cSVegard Nossum * so, skip the rest. 2797c65d5c6cSVegard Nossum */ 2798c65d5c6cSVegard Nossum if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2799c65d5c6cSVegard Nossum jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2800c65d5c6cSVegard Nossum es->s_last_orphan = 0; 2801c65d5c6cSVegard Nossum break; 2802c65d5c6cSVegard Nossum } 2803c65d5c6cSVegard Nossum 280497bd42b9SJosef Bacik inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 280597bd42b9SJosef Bacik if (IS_ERR(inode)) { 2806ac27a0ecSDave Kleikamp es->s_last_orphan = 0; 2807ac27a0ecSDave Kleikamp break; 2808ac27a0ecSDave Kleikamp } 2809ac27a0ecSDave Kleikamp 2810617ba13bSMingming Cao list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2811871a2931SChristoph Hellwig dquot_initialize(inode); 2812ac27a0ecSDave Kleikamp if (inode->i_nlink) { 2813566370a2SPaul Taysom if (test_opt(sb, DEBUG)) 2814b31e1552SEric Sandeen ext4_msg(sb, KERN_DEBUG, 2815b31e1552SEric Sandeen "%s: truncating inode %lu to %lld bytes", 281646e665e9SHarvey Harrison __func__, inode->i_ino, inode->i_size); 2817e5f8eab8STheodore Ts'o jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2818ac27a0ecSDave Kleikamp inode->i_ino, inode->i_size); 28195955102cSAl Viro inode_lock(inode); 282055f252c9SLukas Czerner truncate_inode_pages(inode->i_mapping, inode->i_size); 28212c98eb5eSTheodore Ts'o ret = ext4_truncate(inode); 28222c98eb5eSTheodore Ts'o if (ret) 28232c98eb5eSTheodore Ts'o ext4_std_error(inode->i_sb, ret); 28245955102cSAl Viro inode_unlock(inode); 2825ac27a0ecSDave Kleikamp nr_truncates++; 2826ac27a0ecSDave Kleikamp } else { 2827566370a2SPaul Taysom if (test_opt(sb, DEBUG)) 2828b31e1552SEric Sandeen ext4_msg(sb, KERN_DEBUG, 2829b31e1552SEric Sandeen "%s: deleting unreferenced inode %lu", 283046e665e9SHarvey Harrison __func__, inode->i_ino); 2831ac27a0ecSDave Kleikamp jbd_debug(2, "deleting unreferenced inode %lu\n", 2832ac27a0ecSDave Kleikamp inode->i_ino); 2833ac27a0ecSDave Kleikamp nr_orphans++; 2834ac27a0ecSDave Kleikamp } 2835ac27a0ecSDave Kleikamp iput(inode); /* The delete magic happens here! */ 2836ac27a0ecSDave Kleikamp } 2837ac27a0ecSDave Kleikamp 2838ac27a0ecSDave Kleikamp #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 2839ac27a0ecSDave Kleikamp 2840ac27a0ecSDave Kleikamp if (nr_orphans) 2841b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 2842b31e1552SEric Sandeen PLURAL(nr_orphans)); 2843ac27a0ecSDave Kleikamp if (nr_truncates) 2844b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 2845b31e1552SEric Sandeen PLURAL(nr_truncates)); 2846ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 284795f1fda4Szhangyi (F) /* Turn off quotas if they were enabled for orphan cleanup */ 284895f1fda4Szhangyi (F) if (quota_update) { 2849a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 285095f1fda4Szhangyi (F) if (sb_dqopt(sb)->files[i]) 2851287a8095SChristoph Hellwig dquot_quota_off(sb, i); 2852ac27a0ecSDave Kleikamp } 285395f1fda4Szhangyi (F) } 2854ac27a0ecSDave Kleikamp #endif 28551751e8a6SLinus Torvalds sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2856ac27a0ecSDave Kleikamp } 28570b8e58a1SAndreas Dilger 2858cd2291a4SEric Sandeen /* 2859cd2291a4SEric Sandeen * Maximal extent format file size. 2860cd2291a4SEric Sandeen * Resulting logical blkno at s_maxbytes must fit in our on-disk 2861cd2291a4SEric Sandeen * extent format containers, within a sector_t, and within i_blocks 2862cd2291a4SEric Sandeen * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2863cd2291a4SEric Sandeen * so that won't be a limiting factor. 2864cd2291a4SEric Sandeen * 2865f17722f9SLukas Czerner * However there is other limiting factor. We do store extents in the form 2866f17722f9SLukas Czerner * of starting block and length, hence the resulting length of the extent 2867f17722f9SLukas Czerner * covering maximum file size must fit into on-disk format containers as 2868f17722f9SLukas Czerner * well. Given that length is always by 1 unit bigger than max unit (because 2869f17722f9SLukas Czerner * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2870f17722f9SLukas Czerner * 2871cd2291a4SEric Sandeen * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2872cd2291a4SEric Sandeen */ 2873f287a1a5STheodore Ts'o static loff_t ext4_max_size(int blkbits, int has_huge_files) 2874cd2291a4SEric Sandeen { 2875cd2291a4SEric Sandeen loff_t res; 2876cd2291a4SEric Sandeen loff_t upper_limit = MAX_LFS_FILESIZE; 2877cd2291a4SEric Sandeen 287872deb455SChristoph Hellwig BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); 287972deb455SChristoph Hellwig 288072deb455SChristoph Hellwig if (!has_huge_files) { 2881cd2291a4SEric Sandeen upper_limit = (1LL << 32) - 1; 2882cd2291a4SEric Sandeen 2883cd2291a4SEric Sandeen /* total blocks in file system block size */ 2884cd2291a4SEric Sandeen upper_limit >>= (blkbits - 9); 2885cd2291a4SEric Sandeen upper_limit <<= blkbits; 2886cd2291a4SEric Sandeen } 2887cd2291a4SEric Sandeen 2888f17722f9SLukas Czerner /* 2889f17722f9SLukas Czerner * 32-bit extent-start container, ee_block. We lower the maxbytes 2890f17722f9SLukas Czerner * by one fs block, so ee_len can cover the extent of maximum file 2891f17722f9SLukas Czerner * size 2892f17722f9SLukas Czerner */ 2893f17722f9SLukas Czerner res = (1LL << 32) - 1; 2894cd2291a4SEric Sandeen res <<= blkbits; 2895cd2291a4SEric Sandeen 2896cd2291a4SEric Sandeen /* Sanity check against vm- & vfs- imposed limits */ 2897cd2291a4SEric Sandeen if (res > upper_limit) 2898cd2291a4SEric Sandeen res = upper_limit; 2899cd2291a4SEric Sandeen 2900cd2291a4SEric Sandeen return res; 2901cd2291a4SEric Sandeen } 2902ac27a0ecSDave Kleikamp 2903ac27a0ecSDave Kleikamp /* 2904cd2291a4SEric Sandeen * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 29050fc1b451SAneesh Kumar K.V * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 29060fc1b451SAneesh Kumar K.V * We need to be 1 filesystem block less than the 2^48 sector limit. 2907ac27a0ecSDave Kleikamp */ 2908f287a1a5STheodore Ts'o static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 2909ac27a0ecSDave Kleikamp { 2910617ba13bSMingming Cao loff_t res = EXT4_NDIR_BLOCKS; 29110fc1b451SAneesh Kumar K.V int meta_blocks; 29120fc1b451SAneesh Kumar K.V loff_t upper_limit; 29130b8e58a1SAndreas Dilger /* This is calculated to be the largest file size for a dense, block 29140b8e58a1SAndreas Dilger * mapped file such that the file's total number of 512-byte sectors, 29150b8e58a1SAndreas Dilger * including data and all indirect blocks, does not exceed (2^48 - 1). 29160b8e58a1SAndreas Dilger * 29170b8e58a1SAndreas Dilger * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 29180b8e58a1SAndreas Dilger * number of 512-byte sectors of the file. 29190fc1b451SAneesh Kumar K.V */ 29200fc1b451SAneesh Kumar K.V 292172deb455SChristoph Hellwig if (!has_huge_files) { 29220fc1b451SAneesh Kumar K.V /* 292372deb455SChristoph Hellwig * !has_huge_files or implies that the inode i_block field 292472deb455SChristoph Hellwig * represents total file blocks in 2^32 512-byte sectors == 292572deb455SChristoph Hellwig * size of vfs inode i_blocks * 8 29260fc1b451SAneesh Kumar K.V */ 29270fc1b451SAneesh Kumar K.V upper_limit = (1LL << 32) - 1; 29280fc1b451SAneesh Kumar K.V 29290fc1b451SAneesh Kumar K.V /* total blocks in file system block size */ 29300fc1b451SAneesh Kumar K.V upper_limit >>= (bits - 9); 29310fc1b451SAneesh Kumar K.V 29320fc1b451SAneesh Kumar K.V } else { 29338180a562SAneesh Kumar K.V /* 29348180a562SAneesh Kumar K.V * We use 48 bit ext4_inode i_blocks 29358180a562SAneesh Kumar K.V * With EXT4_HUGE_FILE_FL set the i_blocks 29368180a562SAneesh Kumar K.V * represent total number of blocks in 29378180a562SAneesh Kumar K.V * file system block size 29388180a562SAneesh Kumar K.V */ 29390fc1b451SAneesh Kumar K.V upper_limit = (1LL << 48) - 1; 29400fc1b451SAneesh Kumar K.V 29410fc1b451SAneesh Kumar K.V } 29420fc1b451SAneesh Kumar K.V 29430fc1b451SAneesh Kumar K.V /* indirect blocks */ 29440fc1b451SAneesh Kumar K.V meta_blocks = 1; 29450fc1b451SAneesh Kumar K.V /* double indirect blocks */ 29460fc1b451SAneesh Kumar K.V meta_blocks += 1 + (1LL << (bits-2)); 29470fc1b451SAneesh Kumar K.V /* tripple indirect blocks */ 29480fc1b451SAneesh Kumar K.V meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 29490fc1b451SAneesh Kumar K.V 29500fc1b451SAneesh Kumar K.V upper_limit -= meta_blocks; 29510fc1b451SAneesh Kumar K.V upper_limit <<= bits; 2952ac27a0ecSDave Kleikamp 2953ac27a0ecSDave Kleikamp res += 1LL << (bits-2); 2954ac27a0ecSDave Kleikamp res += 1LL << (2*(bits-2)); 2955ac27a0ecSDave Kleikamp res += 1LL << (3*(bits-2)); 2956ac27a0ecSDave Kleikamp res <<= bits; 2957ac27a0ecSDave Kleikamp if (res > upper_limit) 2958ac27a0ecSDave Kleikamp res = upper_limit; 29590fc1b451SAneesh Kumar K.V 29600fc1b451SAneesh Kumar K.V if (res > MAX_LFS_FILESIZE) 29610fc1b451SAneesh Kumar K.V res = MAX_LFS_FILESIZE; 29620fc1b451SAneesh Kumar K.V 2963ac27a0ecSDave Kleikamp return res; 2964ac27a0ecSDave Kleikamp } 2965ac27a0ecSDave Kleikamp 2966617ba13bSMingming Cao static ext4_fsblk_t descriptor_loc(struct super_block *sb, 296770bbb3e0SAndrew Morton ext4_fsblk_t logical_sb_block, int nr) 2968ac27a0ecSDave Kleikamp { 2969617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 2970fd2d4291SAvantika Mathur ext4_group_t bg, first_meta_bg; 2971ac27a0ecSDave Kleikamp int has_super = 0; 2972ac27a0ecSDave Kleikamp 2973ac27a0ecSDave Kleikamp first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 2974ac27a0ecSDave Kleikamp 2975e2b911c5SDarrick J. Wong if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 297670bbb3e0SAndrew Morton return logical_sb_block + nr + 1; 2977ac27a0ecSDave Kleikamp bg = sbi->s_desc_per_block * nr; 2978617ba13bSMingming Cao if (ext4_bg_has_super(sb, bg)) 2979ac27a0ecSDave Kleikamp has_super = 1; 29800b8e58a1SAndreas Dilger 2981bd63f6b0SDarrick J. Wong /* 2982bd63f6b0SDarrick J. Wong * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 2983bd63f6b0SDarrick J. Wong * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 2984bd63f6b0SDarrick J. Wong * on modern mke2fs or blksize > 1k on older mke2fs) then we must 2985bd63f6b0SDarrick J. Wong * compensate. 2986bd63f6b0SDarrick J. Wong */ 2987bd63f6b0SDarrick J. Wong if (sb->s_blocksize == 1024 && nr == 0 && 298849598e04SJun Piao le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 2989bd63f6b0SDarrick J. Wong has_super++; 2990bd63f6b0SDarrick J. Wong 2991617ba13bSMingming Cao return (has_super + ext4_group_first_block_no(sb, bg)); 2992ac27a0ecSDave Kleikamp } 2993ac27a0ecSDave Kleikamp 2994c9de560dSAlex Tomas /** 2995c9de560dSAlex Tomas * ext4_get_stripe_size: Get the stripe size. 2996c9de560dSAlex Tomas * @sbi: In memory super block info 2997c9de560dSAlex Tomas * 2998c9de560dSAlex Tomas * If we have specified it via mount option, then 2999c9de560dSAlex Tomas * use the mount option value. If the value specified at mount time is 3000c9de560dSAlex Tomas * greater than the blocks per group use the super block value. 3001c9de560dSAlex Tomas * If the super block value is greater than blocks per group return 0. 3002c9de560dSAlex Tomas * Allocator needs it be less than blocks per group. 3003c9de560dSAlex Tomas * 3004c9de560dSAlex Tomas */ 3005c9de560dSAlex Tomas static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 3006c9de560dSAlex Tomas { 3007c9de560dSAlex Tomas unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 3008c9de560dSAlex Tomas unsigned long stripe_width = 3009c9de560dSAlex Tomas le32_to_cpu(sbi->s_es->s_raid_stripe_width); 30103eb08658SDan Ehrenberg int ret; 3011c9de560dSAlex Tomas 3012c9de560dSAlex Tomas if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 30133eb08658SDan Ehrenberg ret = sbi->s_stripe; 30145469d7c3SJan Kara else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 30153eb08658SDan Ehrenberg ret = stripe_width; 30165469d7c3SJan Kara else if (stride && stride <= sbi->s_blocks_per_group) 30173eb08658SDan Ehrenberg ret = stride; 30183eb08658SDan Ehrenberg else 30193eb08658SDan Ehrenberg ret = 0; 3020c9de560dSAlex Tomas 30213eb08658SDan Ehrenberg /* 30223eb08658SDan Ehrenberg * If the stripe width is 1, this makes no sense and 30233eb08658SDan Ehrenberg * we set it to 0 to turn off stripe handling code. 30243eb08658SDan Ehrenberg */ 30253eb08658SDan Ehrenberg if (ret <= 1) 30263eb08658SDan Ehrenberg ret = 0; 3027c9de560dSAlex Tomas 30283eb08658SDan Ehrenberg return ret; 3029c9de560dSAlex Tomas } 3030ac27a0ecSDave Kleikamp 3031a13fb1a4SEric Sandeen /* 3032a13fb1a4SEric Sandeen * Check whether this filesystem can be mounted based on 3033a13fb1a4SEric Sandeen * the features present and the RDONLY/RDWR mount requested. 3034a13fb1a4SEric Sandeen * Returns 1 if this filesystem can be mounted as requested, 3035a13fb1a4SEric Sandeen * 0 if it cannot be. 3036a13fb1a4SEric Sandeen */ 3037a13fb1a4SEric Sandeen static int ext4_feature_set_ok(struct super_block *sb, int readonly) 3038a13fb1a4SEric Sandeen { 3039e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext4_incompat_features(sb)) { 3040a13fb1a4SEric Sandeen ext4_msg(sb, KERN_ERR, 3041a13fb1a4SEric Sandeen "Couldn't mount because of " 3042a13fb1a4SEric Sandeen "unsupported optional features (%x)", 3043a13fb1a4SEric Sandeen (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 3044a13fb1a4SEric Sandeen ~EXT4_FEATURE_INCOMPAT_SUPP)); 3045a13fb1a4SEric Sandeen return 0; 3046a13fb1a4SEric Sandeen } 3047a13fb1a4SEric Sandeen 3048c83ad55eSGabriel Krisman Bertazi #ifndef CONFIG_UNICODE 3049c83ad55eSGabriel Krisman Bertazi if (ext4_has_feature_casefold(sb)) { 3050c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 3051c83ad55eSGabriel Krisman Bertazi "Filesystem with casefold feature cannot be " 3052c83ad55eSGabriel Krisman Bertazi "mounted without CONFIG_UNICODE"); 3053c83ad55eSGabriel Krisman Bertazi return 0; 3054c83ad55eSGabriel Krisman Bertazi } 3055c83ad55eSGabriel Krisman Bertazi #endif 3056c83ad55eSGabriel Krisman Bertazi 3057a13fb1a4SEric Sandeen if (readonly) 3058a13fb1a4SEric Sandeen return 1; 3059a13fb1a4SEric Sandeen 3060e2b911c5SDarrick J. Wong if (ext4_has_feature_readonly(sb)) { 30612cb5cc8bSDarrick J. Wong ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 30621751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 30632cb5cc8bSDarrick J. Wong return 1; 30642cb5cc8bSDarrick J. Wong } 30652cb5cc8bSDarrick J. Wong 3066a13fb1a4SEric Sandeen /* Check that feature set is OK for a read-write mount */ 3067e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 3068a13fb1a4SEric Sandeen ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 3069a13fb1a4SEric Sandeen "unsupported optional features (%x)", 3070a13fb1a4SEric Sandeen (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 3071a13fb1a4SEric Sandeen ~EXT4_FEATURE_RO_COMPAT_SUPP)); 3072a13fb1a4SEric Sandeen return 0; 3073a13fb1a4SEric Sandeen } 3074e2b911c5SDarrick J. Wong if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 3075bab08ab9STheodore Ts'o ext4_msg(sb, KERN_ERR, 3076bab08ab9STheodore Ts'o "Can't support bigalloc feature without " 3077bab08ab9STheodore Ts'o "extents feature\n"); 3078bab08ab9STheodore Ts'o return 0; 3079bab08ab9STheodore Ts'o } 30807c319d32SAditya Kali 30819db176bcSJan Kara #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2) 3082d65d87a0STheodore Ts'o if (!readonly && (ext4_has_feature_quota(sb) || 3083d65d87a0STheodore Ts'o ext4_has_feature_project(sb))) { 30847c319d32SAditya Kali ext4_msg(sb, KERN_ERR, 3085d65d87a0STheodore Ts'o "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2"); 3086689c958cSLi Xi return 0; 3087689c958cSLi Xi } 30887c319d32SAditya Kali #endif /* CONFIG_QUOTA */ 3089a13fb1a4SEric Sandeen return 1; 3090a13fb1a4SEric Sandeen } 3091a13fb1a4SEric Sandeen 309266e61a9eSTheodore Ts'o /* 309366e61a9eSTheodore Ts'o * This function is called once a day if we have errors logged 309466e61a9eSTheodore Ts'o * on the file system 309566e61a9eSTheodore Ts'o */ 3096235699a8SKees Cook static void print_daily_error_info(struct timer_list *t) 309766e61a9eSTheodore Ts'o { 3098235699a8SKees Cook struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report); 3099235699a8SKees Cook struct super_block *sb = sbi->s_sb; 3100235699a8SKees Cook struct ext4_super_block *es = sbi->s_es; 310166e61a9eSTheodore Ts'o 310266e61a9eSTheodore Ts'o if (es->s_error_count) 3103ae0f78deSTheodore Ts'o /* fsck newer than v1.41.13 is needed to clean this condition. */ 3104ae0f78deSTheodore Ts'o ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 310566e61a9eSTheodore Ts'o le32_to_cpu(es->s_error_count)); 310666e61a9eSTheodore Ts'o if (es->s_first_error_time) { 31076a0678a7SArnd Bergmann printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 31086a0678a7SArnd Bergmann sb->s_id, 31096a0678a7SArnd Bergmann ext4_get_tstamp(es, s_first_error_time), 311066e61a9eSTheodore Ts'o (int) sizeof(es->s_first_error_func), 311166e61a9eSTheodore Ts'o es->s_first_error_func, 311266e61a9eSTheodore Ts'o le32_to_cpu(es->s_first_error_line)); 311366e61a9eSTheodore Ts'o if (es->s_first_error_ino) 3114651e1c3bSJoe Perches printk(KERN_CONT ": inode %u", 311566e61a9eSTheodore Ts'o le32_to_cpu(es->s_first_error_ino)); 311666e61a9eSTheodore Ts'o if (es->s_first_error_block) 3117651e1c3bSJoe Perches printk(KERN_CONT ": block %llu", (unsigned long long) 311866e61a9eSTheodore Ts'o le64_to_cpu(es->s_first_error_block)); 3119651e1c3bSJoe Perches printk(KERN_CONT "\n"); 312066e61a9eSTheodore Ts'o } 312166e61a9eSTheodore Ts'o if (es->s_last_error_time) { 31226a0678a7SArnd Bergmann printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 31236a0678a7SArnd Bergmann sb->s_id, 31246a0678a7SArnd Bergmann ext4_get_tstamp(es, s_last_error_time), 312566e61a9eSTheodore Ts'o (int) sizeof(es->s_last_error_func), 312666e61a9eSTheodore Ts'o es->s_last_error_func, 312766e61a9eSTheodore Ts'o le32_to_cpu(es->s_last_error_line)); 312866e61a9eSTheodore Ts'o if (es->s_last_error_ino) 3129651e1c3bSJoe Perches printk(KERN_CONT ": inode %u", 313066e61a9eSTheodore Ts'o le32_to_cpu(es->s_last_error_ino)); 313166e61a9eSTheodore Ts'o if (es->s_last_error_block) 3132651e1c3bSJoe Perches printk(KERN_CONT ": block %llu", (unsigned long long) 313366e61a9eSTheodore Ts'o le64_to_cpu(es->s_last_error_block)); 3134651e1c3bSJoe Perches printk(KERN_CONT "\n"); 313566e61a9eSTheodore Ts'o } 313666e61a9eSTheodore Ts'o mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 313766e61a9eSTheodore Ts'o } 313866e61a9eSTheodore Ts'o 3139bfff6873SLukas Czerner /* Find next suitable group and run ext4_init_inode_table */ 3140bfff6873SLukas Czerner static int ext4_run_li_request(struct ext4_li_request *elr) 3141bfff6873SLukas Czerner { 3142bfff6873SLukas Czerner struct ext4_group_desc *gdp = NULL; 3143bfff6873SLukas Czerner ext4_group_t group, ngroups; 3144bfff6873SLukas Czerner struct super_block *sb; 3145bfff6873SLukas Czerner unsigned long timeout = 0; 3146bfff6873SLukas Czerner int ret = 0; 3147bfff6873SLukas Czerner 3148bfff6873SLukas Czerner sb = elr->lr_super; 3149bfff6873SLukas Czerner ngroups = EXT4_SB(sb)->s_groups_count; 3150bfff6873SLukas Czerner 3151bfff6873SLukas Czerner for (group = elr->lr_next_group; group < ngroups; group++) { 3152bfff6873SLukas Czerner gdp = ext4_get_group_desc(sb, group, NULL); 3153bfff6873SLukas Czerner if (!gdp) { 3154bfff6873SLukas Czerner ret = 1; 3155bfff6873SLukas Czerner break; 3156bfff6873SLukas Czerner } 3157bfff6873SLukas Czerner 3158bfff6873SLukas Czerner if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3159bfff6873SLukas Czerner break; 3160bfff6873SLukas Czerner } 3161bfff6873SLukas Czerner 31627f511862STheodore Ts'o if (group >= ngroups) 3163bfff6873SLukas Czerner ret = 1; 3164bfff6873SLukas Czerner 3165bfff6873SLukas Czerner if (!ret) { 3166bfff6873SLukas Czerner timeout = jiffies; 3167bfff6873SLukas Czerner ret = ext4_init_inode_table(sb, group, 3168bfff6873SLukas Czerner elr->lr_timeout ? 0 : 1); 3169bfff6873SLukas Czerner if (elr->lr_timeout == 0) { 317051ce6511SLukas Czerner timeout = (jiffies - timeout) * 317151ce6511SLukas Czerner elr->lr_sbi->s_li_wait_mult; 3172bfff6873SLukas Czerner elr->lr_timeout = timeout; 3173bfff6873SLukas Czerner } 3174bfff6873SLukas Czerner elr->lr_next_sched = jiffies + elr->lr_timeout; 3175bfff6873SLukas Czerner elr->lr_next_group = group + 1; 3176bfff6873SLukas Czerner } 3177bfff6873SLukas Czerner return ret; 3178bfff6873SLukas Czerner } 3179bfff6873SLukas Czerner 3180bfff6873SLukas Czerner /* 3181bfff6873SLukas Czerner * Remove lr_request from the list_request and free the 31824ed5c033SLukas Czerner * request structure. Should be called with li_list_mtx held 3183bfff6873SLukas Czerner */ 3184bfff6873SLukas Czerner static void ext4_remove_li_request(struct ext4_li_request *elr) 3185bfff6873SLukas Czerner { 3186bfff6873SLukas Czerner struct ext4_sb_info *sbi; 3187bfff6873SLukas Czerner 3188bfff6873SLukas Czerner if (!elr) 3189bfff6873SLukas Czerner return; 3190bfff6873SLukas Czerner 3191bfff6873SLukas Czerner sbi = elr->lr_sbi; 3192bfff6873SLukas Czerner 3193bfff6873SLukas Czerner list_del(&elr->lr_request); 3194bfff6873SLukas Czerner sbi->s_li_request = NULL; 3195bfff6873SLukas Czerner kfree(elr); 3196bfff6873SLukas Czerner } 3197bfff6873SLukas Czerner 3198bfff6873SLukas Czerner static void ext4_unregister_li_request(struct super_block *sb) 3199bfff6873SLukas Czerner { 32001bb933fbSLukas Czerner mutex_lock(&ext4_li_mtx); 32011bb933fbSLukas Czerner if (!ext4_li_info) { 32021bb933fbSLukas Czerner mutex_unlock(&ext4_li_mtx); 3203bfff6873SLukas Czerner return; 32041bb933fbSLukas Czerner } 3205bfff6873SLukas Czerner 3206bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 32071bb933fbSLukas Czerner ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3208bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 32091bb933fbSLukas Czerner mutex_unlock(&ext4_li_mtx); 3210bfff6873SLukas Czerner } 3211bfff6873SLukas Czerner 32128f1f7453SEric Sandeen static struct task_struct *ext4_lazyinit_task; 32138f1f7453SEric Sandeen 3214bfff6873SLukas Czerner /* 3215bfff6873SLukas Czerner * This is the function where ext4lazyinit thread lives. It walks 3216bfff6873SLukas Czerner * through the request list searching for next scheduled filesystem. 3217bfff6873SLukas Czerner * When such a fs is found, run the lazy initialization request 3218bfff6873SLukas Czerner * (ext4_rn_li_request) and keep track of the time spend in this 3219bfff6873SLukas Czerner * function. Based on that time we compute next schedule time of 3220bfff6873SLukas Czerner * the request. When walking through the list is complete, compute 3221bfff6873SLukas Czerner * next waking time and put itself into sleep. 3222bfff6873SLukas Czerner */ 3223bfff6873SLukas Czerner static int ext4_lazyinit_thread(void *arg) 3224bfff6873SLukas Czerner { 3225bfff6873SLukas Czerner struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 3226bfff6873SLukas Czerner struct list_head *pos, *n; 3227bfff6873SLukas Czerner struct ext4_li_request *elr; 32284ed5c033SLukas Czerner unsigned long next_wakeup, cur; 3229bfff6873SLukas Czerner 3230bfff6873SLukas Czerner BUG_ON(NULL == eli); 3231bfff6873SLukas Czerner 3232bfff6873SLukas Czerner cont_thread: 3233bfff6873SLukas Czerner while (true) { 3234bfff6873SLukas Czerner next_wakeup = MAX_JIFFY_OFFSET; 3235bfff6873SLukas Czerner 3236bfff6873SLukas Czerner mutex_lock(&eli->li_list_mtx); 3237bfff6873SLukas Czerner if (list_empty(&eli->li_request_list)) { 3238bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3239bfff6873SLukas Czerner goto exit_thread; 3240bfff6873SLukas Czerner } 3241bfff6873SLukas Czerner list_for_each_safe(pos, n, &eli->li_request_list) { 3242e22834f0SDmitry Monakhov int err = 0; 3243e22834f0SDmitry Monakhov int progress = 0; 3244bfff6873SLukas Czerner elr = list_entry(pos, struct ext4_li_request, 3245bfff6873SLukas Czerner lr_request); 3246bfff6873SLukas Czerner 3247e22834f0SDmitry Monakhov if (time_before(jiffies, elr->lr_next_sched)) { 3248e22834f0SDmitry Monakhov if (time_before(elr->lr_next_sched, next_wakeup)) 3249e22834f0SDmitry Monakhov next_wakeup = elr->lr_next_sched; 3250e22834f0SDmitry Monakhov continue; 3251e22834f0SDmitry Monakhov } 3252e22834f0SDmitry Monakhov if (down_read_trylock(&elr->lr_super->s_umount)) { 3253e22834f0SDmitry Monakhov if (sb_start_write_trylock(elr->lr_super)) { 3254e22834f0SDmitry Monakhov progress = 1; 3255e22834f0SDmitry Monakhov /* 3256e22834f0SDmitry Monakhov * We hold sb->s_umount, sb can not 3257e22834f0SDmitry Monakhov * be removed from the list, it is 3258e22834f0SDmitry Monakhov * now safe to drop li_list_mtx 3259e22834f0SDmitry Monakhov */ 3260e22834f0SDmitry Monakhov mutex_unlock(&eli->li_list_mtx); 3261e22834f0SDmitry Monakhov err = ext4_run_li_request(elr); 3262e22834f0SDmitry Monakhov sb_end_write(elr->lr_super); 3263e22834f0SDmitry Monakhov mutex_lock(&eli->li_list_mtx); 3264e22834f0SDmitry Monakhov n = pos->next; 3265e22834f0SDmitry Monakhov } 3266e22834f0SDmitry Monakhov up_read((&elr->lr_super->s_umount)); 3267e22834f0SDmitry Monakhov } 3268b2c78cd0STheodore Ts'o /* error, remove the lazy_init job */ 3269e22834f0SDmitry Monakhov if (err) { 3270bfff6873SLukas Czerner ext4_remove_li_request(elr); 3271bfff6873SLukas Czerner continue; 3272bfff6873SLukas Czerner } 3273e22834f0SDmitry Monakhov if (!progress) { 3274e22834f0SDmitry Monakhov elr->lr_next_sched = jiffies + 3275e22834f0SDmitry Monakhov (prandom_u32() 3276e22834f0SDmitry Monakhov % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3277b2c78cd0STheodore Ts'o } 3278bfff6873SLukas Czerner if (time_before(elr->lr_next_sched, next_wakeup)) 3279bfff6873SLukas Czerner next_wakeup = elr->lr_next_sched; 3280bfff6873SLukas Czerner } 3281bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3282bfff6873SLukas Czerner 3283a0acae0eSTejun Heo try_to_freeze(); 3284bfff6873SLukas Czerner 32854ed5c033SLukas Czerner cur = jiffies; 32864ed5c033SLukas Czerner if ((time_after_eq(cur, next_wakeup)) || 3287f4245bd4SLukas Czerner (MAX_JIFFY_OFFSET == next_wakeup)) { 3288bfff6873SLukas Czerner cond_resched(); 3289bfff6873SLukas Czerner continue; 3290bfff6873SLukas Czerner } 3291bfff6873SLukas Czerner 32924ed5c033SLukas Czerner schedule_timeout_interruptible(next_wakeup - cur); 32934ed5c033SLukas Czerner 32948f1f7453SEric Sandeen if (kthread_should_stop()) { 32958f1f7453SEric Sandeen ext4_clear_request_list(); 32968f1f7453SEric Sandeen goto exit_thread; 32978f1f7453SEric Sandeen } 3298bfff6873SLukas Czerner } 3299bfff6873SLukas Czerner 3300bfff6873SLukas Czerner exit_thread: 3301bfff6873SLukas Czerner /* 3302bfff6873SLukas Czerner * It looks like the request list is empty, but we need 3303bfff6873SLukas Czerner * to check it under the li_list_mtx lock, to prevent any 3304bfff6873SLukas Czerner * additions into it, and of course we should lock ext4_li_mtx 3305bfff6873SLukas Czerner * to atomically free the list and ext4_li_info, because at 3306bfff6873SLukas Czerner * this point another ext4 filesystem could be registering 3307bfff6873SLukas Czerner * new one. 3308bfff6873SLukas Czerner */ 3309bfff6873SLukas Czerner mutex_lock(&ext4_li_mtx); 3310bfff6873SLukas Czerner mutex_lock(&eli->li_list_mtx); 3311bfff6873SLukas Czerner if (!list_empty(&eli->li_request_list)) { 3312bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3313bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 3314bfff6873SLukas Czerner goto cont_thread; 3315bfff6873SLukas Czerner } 3316bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3317bfff6873SLukas Czerner kfree(ext4_li_info); 3318bfff6873SLukas Czerner ext4_li_info = NULL; 3319bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 3320bfff6873SLukas Czerner 3321bfff6873SLukas Czerner return 0; 3322bfff6873SLukas Czerner } 3323bfff6873SLukas Czerner 3324bfff6873SLukas Czerner static void ext4_clear_request_list(void) 3325bfff6873SLukas Czerner { 3326bfff6873SLukas Czerner struct list_head *pos, *n; 3327bfff6873SLukas Czerner struct ext4_li_request *elr; 3328bfff6873SLukas Czerner 3329bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 3330bfff6873SLukas Czerner list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3331bfff6873SLukas Czerner elr = list_entry(pos, struct ext4_li_request, 3332bfff6873SLukas Czerner lr_request); 3333bfff6873SLukas Czerner ext4_remove_li_request(elr); 3334bfff6873SLukas Czerner } 3335bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 3336bfff6873SLukas Czerner } 3337bfff6873SLukas Czerner 3338bfff6873SLukas Czerner static int ext4_run_lazyinit_thread(void) 3339bfff6873SLukas Czerner { 33408f1f7453SEric Sandeen ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 33418f1f7453SEric Sandeen ext4_li_info, "ext4lazyinit"); 33428f1f7453SEric Sandeen if (IS_ERR(ext4_lazyinit_task)) { 33438f1f7453SEric Sandeen int err = PTR_ERR(ext4_lazyinit_task); 3344bfff6873SLukas Czerner ext4_clear_request_list(); 3345bfff6873SLukas Czerner kfree(ext4_li_info); 3346bfff6873SLukas Czerner ext4_li_info = NULL; 334792b97816STheodore Ts'o printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3348bfff6873SLukas Czerner "initialization thread\n", 3349bfff6873SLukas Czerner err); 3350bfff6873SLukas Czerner return err; 3351bfff6873SLukas Czerner } 3352bfff6873SLukas Czerner ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3353bfff6873SLukas Czerner return 0; 3354bfff6873SLukas Czerner } 3355bfff6873SLukas Czerner 3356bfff6873SLukas Czerner /* 3357bfff6873SLukas Czerner * Check whether it make sense to run itable init. thread or not. 3358bfff6873SLukas Czerner * If there is at least one uninitialized inode table, return 3359bfff6873SLukas Czerner * corresponding group number, else the loop goes through all 3360bfff6873SLukas Czerner * groups and return total number of groups. 3361bfff6873SLukas Czerner */ 3362bfff6873SLukas Czerner static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3363bfff6873SLukas Czerner { 3364bfff6873SLukas Czerner ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3365bfff6873SLukas Czerner struct ext4_group_desc *gdp = NULL; 3366bfff6873SLukas Czerner 33678844618dSTheodore Ts'o if (!ext4_has_group_desc_csum(sb)) 33688844618dSTheodore Ts'o return ngroups; 33698844618dSTheodore Ts'o 3370bfff6873SLukas Czerner for (group = 0; group < ngroups; group++) { 3371bfff6873SLukas Czerner gdp = ext4_get_group_desc(sb, group, NULL); 3372bfff6873SLukas Czerner if (!gdp) 3373bfff6873SLukas Czerner continue; 3374bfff6873SLukas Czerner 337550122847STheodore Ts'o if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3376bfff6873SLukas Czerner break; 3377bfff6873SLukas Czerner } 3378bfff6873SLukas Czerner 3379bfff6873SLukas Czerner return group; 3380bfff6873SLukas Czerner } 3381bfff6873SLukas Czerner 3382bfff6873SLukas Czerner static int ext4_li_info_new(void) 3383bfff6873SLukas Czerner { 3384bfff6873SLukas Czerner struct ext4_lazy_init *eli = NULL; 3385bfff6873SLukas Czerner 3386bfff6873SLukas Czerner eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3387bfff6873SLukas Czerner if (!eli) 3388bfff6873SLukas Czerner return -ENOMEM; 3389bfff6873SLukas Czerner 3390bfff6873SLukas Czerner INIT_LIST_HEAD(&eli->li_request_list); 3391bfff6873SLukas Czerner mutex_init(&eli->li_list_mtx); 3392bfff6873SLukas Czerner 3393bfff6873SLukas Czerner eli->li_state |= EXT4_LAZYINIT_QUIT; 3394bfff6873SLukas Czerner 3395bfff6873SLukas Czerner ext4_li_info = eli; 3396bfff6873SLukas Czerner 3397bfff6873SLukas Czerner return 0; 3398bfff6873SLukas Czerner } 3399bfff6873SLukas Czerner 3400bfff6873SLukas Czerner static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3401bfff6873SLukas Czerner ext4_group_t start) 3402bfff6873SLukas Czerner { 3403bfff6873SLukas Czerner struct ext4_sb_info *sbi = EXT4_SB(sb); 3404bfff6873SLukas Czerner struct ext4_li_request *elr; 3405bfff6873SLukas Czerner 3406bfff6873SLukas Czerner elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3407bfff6873SLukas Czerner if (!elr) 3408bfff6873SLukas Czerner return NULL; 3409bfff6873SLukas Czerner 3410bfff6873SLukas Czerner elr->lr_super = sb; 3411bfff6873SLukas Czerner elr->lr_sbi = sbi; 3412bfff6873SLukas Czerner elr->lr_next_group = start; 3413bfff6873SLukas Czerner 3414bfff6873SLukas Czerner /* 3415bfff6873SLukas Czerner * Randomize first schedule time of the request to 3416bfff6873SLukas Czerner * spread the inode table initialization requests 3417bfff6873SLukas Czerner * better. 3418bfff6873SLukas Czerner */ 3419dd1f723bSTheodore Ts'o elr->lr_next_sched = jiffies + (prandom_u32() % 3420dd1f723bSTheodore Ts'o (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3421bfff6873SLukas Czerner return elr; 3422bfff6873SLukas Czerner } 3423bfff6873SLukas Czerner 34247f511862STheodore Ts'o int ext4_register_li_request(struct super_block *sb, 3425bfff6873SLukas Czerner ext4_group_t first_not_zeroed) 3426bfff6873SLukas Czerner { 3427bfff6873SLukas Czerner struct ext4_sb_info *sbi = EXT4_SB(sb); 34287f511862STheodore Ts'o struct ext4_li_request *elr = NULL; 342949598e04SJun Piao ext4_group_t ngroups = sbi->s_groups_count; 34306c5a6cb9SAndrew Morton int ret = 0; 3431bfff6873SLukas Czerner 34327f511862STheodore Ts'o mutex_lock(&ext4_li_mtx); 343351ce6511SLukas Czerner if (sbi->s_li_request != NULL) { 343451ce6511SLukas Czerner /* 343551ce6511SLukas Czerner * Reset timeout so it can be computed again, because 343651ce6511SLukas Czerner * s_li_wait_mult might have changed. 343751ce6511SLukas Czerner */ 343851ce6511SLukas Czerner sbi->s_li_request->lr_timeout = 0; 34397f511862STheodore Ts'o goto out; 344051ce6511SLukas Czerner } 3441bfff6873SLukas Czerner 3442bc98a42cSDavid Howells if (first_not_zeroed == ngroups || sb_rdonly(sb) || 344355ff3840STao Ma !test_opt(sb, INIT_INODE_TABLE)) 34447f511862STheodore Ts'o goto out; 3445bfff6873SLukas Czerner 3446bfff6873SLukas Czerner elr = ext4_li_request_new(sb, first_not_zeroed); 34477f511862STheodore Ts'o if (!elr) { 34487f511862STheodore Ts'o ret = -ENOMEM; 34497f511862STheodore Ts'o goto out; 34507f511862STheodore Ts'o } 3451bfff6873SLukas Czerner 3452bfff6873SLukas Czerner if (NULL == ext4_li_info) { 3453bfff6873SLukas Czerner ret = ext4_li_info_new(); 3454bfff6873SLukas Czerner if (ret) 3455bfff6873SLukas Czerner goto out; 3456bfff6873SLukas Czerner } 3457bfff6873SLukas Czerner 3458bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 3459bfff6873SLukas Czerner list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3460bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 3461bfff6873SLukas Czerner 3462bfff6873SLukas Czerner sbi->s_li_request = elr; 346346e4690bSTao Ma /* 346446e4690bSTao Ma * set elr to NULL here since it has been inserted to 346546e4690bSTao Ma * the request_list and the removal and free of it is 346646e4690bSTao Ma * handled by ext4_clear_request_list from now on. 346746e4690bSTao Ma */ 346846e4690bSTao Ma elr = NULL; 3469bfff6873SLukas Czerner 3470bfff6873SLukas Czerner if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3471bfff6873SLukas Czerner ret = ext4_run_lazyinit_thread(); 3472bfff6873SLukas Czerner if (ret) 3473bfff6873SLukas Czerner goto out; 3474bfff6873SLukas Czerner } 3475bfff6873SLukas Czerner out: 3476bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 3477beed5ecbSNicolas Kaiser if (ret) 3478bfff6873SLukas Czerner kfree(elr); 3479bfff6873SLukas Czerner return ret; 3480bfff6873SLukas Czerner } 3481bfff6873SLukas Czerner 3482bfff6873SLukas Czerner /* 3483bfff6873SLukas Czerner * We do not need to lock anything since this is called on 3484bfff6873SLukas Czerner * module unload. 3485bfff6873SLukas Czerner */ 3486bfff6873SLukas Czerner static void ext4_destroy_lazyinit_thread(void) 3487bfff6873SLukas Czerner { 3488bfff6873SLukas Czerner /* 3489bfff6873SLukas Czerner * If thread exited earlier 3490bfff6873SLukas Czerner * there's nothing to be done. 3491bfff6873SLukas Czerner */ 34928f1f7453SEric Sandeen if (!ext4_li_info || !ext4_lazyinit_task) 3493bfff6873SLukas Czerner return; 3494bfff6873SLukas Czerner 34958f1f7453SEric Sandeen kthread_stop(ext4_lazyinit_task); 3496bfff6873SLukas Czerner } 3497bfff6873SLukas Czerner 349825ed6e8aSDarrick J. Wong static int set_journal_csum_feature_set(struct super_block *sb) 349925ed6e8aSDarrick J. Wong { 350025ed6e8aSDarrick J. Wong int ret = 1; 350125ed6e8aSDarrick J. Wong int compat, incompat; 350225ed6e8aSDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 350325ed6e8aSDarrick J. Wong 35049aa5d32bSDmitry Monakhov if (ext4_has_metadata_csum(sb)) { 3505db9ee220SDarrick J. Wong /* journal checksum v3 */ 350625ed6e8aSDarrick J. Wong compat = 0; 3507db9ee220SDarrick J. Wong incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 350825ed6e8aSDarrick J. Wong } else { 350925ed6e8aSDarrick J. Wong /* journal checksum v1 */ 351025ed6e8aSDarrick J. Wong compat = JBD2_FEATURE_COMPAT_CHECKSUM; 351125ed6e8aSDarrick J. Wong incompat = 0; 351225ed6e8aSDarrick J. Wong } 351325ed6e8aSDarrick J. Wong 3514feb8c6d3SDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 3515feb8c6d3SDarrick J. Wong JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3516feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_CSUM_V3 | 3517feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_CSUM_V2); 351825ed6e8aSDarrick J. Wong if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 351925ed6e8aSDarrick J. Wong ret = jbd2_journal_set_features(sbi->s_journal, 352025ed6e8aSDarrick J. Wong compat, 0, 352125ed6e8aSDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 352225ed6e8aSDarrick J. Wong incompat); 352325ed6e8aSDarrick J. Wong } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 352425ed6e8aSDarrick J. Wong ret = jbd2_journal_set_features(sbi->s_journal, 352525ed6e8aSDarrick J. Wong compat, 0, 352625ed6e8aSDarrick J. Wong incompat); 352725ed6e8aSDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 0, 0, 352825ed6e8aSDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 352925ed6e8aSDarrick J. Wong } else { 3530feb8c6d3SDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3531feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 353225ed6e8aSDarrick J. Wong } 353325ed6e8aSDarrick J. Wong 353425ed6e8aSDarrick J. Wong return ret; 353525ed6e8aSDarrick J. Wong } 353625ed6e8aSDarrick J. Wong 3537952fc18eSTheodore Ts'o /* 3538952fc18eSTheodore Ts'o * Note: calculating the overhead so we can be compatible with 3539952fc18eSTheodore Ts'o * historical BSD practice is quite difficult in the face of 3540952fc18eSTheodore Ts'o * clusters/bigalloc. This is because multiple metadata blocks from 3541952fc18eSTheodore Ts'o * different block group can end up in the same allocation cluster. 3542952fc18eSTheodore Ts'o * Calculating the exact overhead in the face of clustered allocation 3543952fc18eSTheodore Ts'o * requires either O(all block bitmaps) in memory or O(number of block 3544952fc18eSTheodore Ts'o * groups**2) in time. We will still calculate the superblock for 3545952fc18eSTheodore Ts'o * older file systems --- and if we come across with a bigalloc file 3546952fc18eSTheodore Ts'o * system with zero in s_overhead_clusters the estimate will be close to 3547952fc18eSTheodore Ts'o * correct especially for very large cluster sizes --- but for newer 3548952fc18eSTheodore Ts'o * file systems, it's better to calculate this figure once at mkfs 3549952fc18eSTheodore Ts'o * time, and store it in the superblock. If the superblock value is 3550952fc18eSTheodore Ts'o * present (even for non-bigalloc file systems), we will use it. 3551952fc18eSTheodore Ts'o */ 3552952fc18eSTheodore Ts'o static int count_overhead(struct super_block *sb, ext4_group_t grp, 3553952fc18eSTheodore Ts'o char *buf) 3554952fc18eSTheodore Ts'o { 3555952fc18eSTheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 3556952fc18eSTheodore Ts'o struct ext4_group_desc *gdp; 3557952fc18eSTheodore Ts'o ext4_fsblk_t first_block, last_block, b; 3558952fc18eSTheodore Ts'o ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3559952fc18eSTheodore Ts'o int s, j, count = 0; 3560952fc18eSTheodore Ts'o 3561e2b911c5SDarrick J. Wong if (!ext4_has_feature_bigalloc(sb)) 35620548bbb8STheodore Ts'o return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 35630548bbb8STheodore Ts'o sbi->s_itb_per_group + 2); 35640548bbb8STheodore Ts'o 3565952fc18eSTheodore Ts'o first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3566952fc18eSTheodore Ts'o (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3567952fc18eSTheodore Ts'o last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3568952fc18eSTheodore Ts'o for (i = 0; i < ngroups; i++) { 3569952fc18eSTheodore Ts'o gdp = ext4_get_group_desc(sb, i, NULL); 3570952fc18eSTheodore Ts'o b = ext4_block_bitmap(sb, gdp); 3571952fc18eSTheodore Ts'o if (b >= first_block && b <= last_block) { 3572952fc18eSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3573952fc18eSTheodore Ts'o count++; 3574952fc18eSTheodore Ts'o } 3575952fc18eSTheodore Ts'o b = ext4_inode_bitmap(sb, gdp); 3576952fc18eSTheodore Ts'o if (b >= first_block && b <= last_block) { 3577952fc18eSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3578952fc18eSTheodore Ts'o count++; 3579952fc18eSTheodore Ts'o } 3580952fc18eSTheodore Ts'o b = ext4_inode_table(sb, gdp); 3581952fc18eSTheodore Ts'o if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3582952fc18eSTheodore Ts'o for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3583952fc18eSTheodore Ts'o int c = EXT4_B2C(sbi, b - first_block); 3584952fc18eSTheodore Ts'o ext4_set_bit(c, buf); 3585952fc18eSTheodore Ts'o count++; 3586952fc18eSTheodore Ts'o } 3587952fc18eSTheodore Ts'o if (i != grp) 3588952fc18eSTheodore Ts'o continue; 3589952fc18eSTheodore Ts'o s = 0; 3590952fc18eSTheodore Ts'o if (ext4_bg_has_super(sb, grp)) { 3591952fc18eSTheodore Ts'o ext4_set_bit(s++, buf); 3592952fc18eSTheodore Ts'o count++; 3593952fc18eSTheodore Ts'o } 3594c48ae41bSTheodore Ts'o j = ext4_bg_num_gdb(sb, grp); 3595c48ae41bSTheodore Ts'o if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 3596c48ae41bSTheodore Ts'o ext4_error(sb, "Invalid number of block group " 3597c48ae41bSTheodore Ts'o "descriptor blocks: %d", j); 3598c48ae41bSTheodore Ts'o j = EXT4_BLOCKS_PER_GROUP(sb) - s; 3599952fc18eSTheodore Ts'o } 3600c48ae41bSTheodore Ts'o count += j; 3601c48ae41bSTheodore Ts'o for (; j > 0; j--) 3602c48ae41bSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3603952fc18eSTheodore Ts'o } 3604952fc18eSTheodore Ts'o if (!count) 3605952fc18eSTheodore Ts'o return 0; 3606952fc18eSTheodore Ts'o return EXT4_CLUSTERS_PER_GROUP(sb) - 3607952fc18eSTheodore Ts'o ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3608952fc18eSTheodore Ts'o } 3609952fc18eSTheodore Ts'o 3610952fc18eSTheodore Ts'o /* 3611952fc18eSTheodore Ts'o * Compute the overhead and stash it in sbi->s_overhead 3612952fc18eSTheodore Ts'o */ 3613952fc18eSTheodore Ts'o int ext4_calculate_overhead(struct super_block *sb) 3614952fc18eSTheodore Ts'o { 3615952fc18eSTheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 3616952fc18eSTheodore Ts'o struct ext4_super_block *es = sbi->s_es; 36173c816dedSEric Whitney struct inode *j_inode; 36183c816dedSEric Whitney unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 3619952fc18eSTheodore Ts'o ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3620952fc18eSTheodore Ts'o ext4_fsblk_t overhead = 0; 36214fdb5543SDmitry Monakhov char *buf = (char *) get_zeroed_page(GFP_NOFS); 3622952fc18eSTheodore Ts'o 3623952fc18eSTheodore Ts'o if (!buf) 3624952fc18eSTheodore Ts'o return -ENOMEM; 3625952fc18eSTheodore Ts'o 3626952fc18eSTheodore Ts'o /* 3627952fc18eSTheodore Ts'o * Compute the overhead (FS structures). This is constant 3628952fc18eSTheodore Ts'o * for a given filesystem unless the number of block groups 3629952fc18eSTheodore Ts'o * changes so we cache the previous value until it does. 3630952fc18eSTheodore Ts'o */ 3631952fc18eSTheodore Ts'o 3632952fc18eSTheodore Ts'o /* 3633952fc18eSTheodore Ts'o * All of the blocks before first_data_block are overhead 3634952fc18eSTheodore Ts'o */ 3635952fc18eSTheodore Ts'o overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3636952fc18eSTheodore Ts'o 3637952fc18eSTheodore Ts'o /* 3638952fc18eSTheodore Ts'o * Add the overhead found in each block group 3639952fc18eSTheodore Ts'o */ 3640952fc18eSTheodore Ts'o for (i = 0; i < ngroups; i++) { 3641952fc18eSTheodore Ts'o int blks; 3642952fc18eSTheodore Ts'o 3643952fc18eSTheodore Ts'o blks = count_overhead(sb, i, buf); 3644952fc18eSTheodore Ts'o overhead += blks; 3645952fc18eSTheodore Ts'o if (blks) 3646952fc18eSTheodore Ts'o memset(buf, 0, PAGE_SIZE); 3647952fc18eSTheodore Ts'o cond_resched(); 3648952fc18eSTheodore Ts'o } 36493c816dedSEric Whitney 36503c816dedSEric Whitney /* 36513c816dedSEric Whitney * Add the internal journal blocks whether the journal has been 36523c816dedSEric Whitney * loaded or not 36533c816dedSEric Whitney */ 3654b003b524SEric Sandeen if (sbi->s_journal && !sbi->journal_bdev) 3655810da240SLukas Czerner overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen); 3656f1eec3b0SRitesh Harjani else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) { 3657f1eec3b0SRitesh Harjani /* j_inum for internal journal is non-zero */ 36583c816dedSEric Whitney j_inode = ext4_get_journal_inode(sb, j_inum); 36593c816dedSEric Whitney if (j_inode) { 36603c816dedSEric Whitney j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 36613c816dedSEric Whitney overhead += EXT4_NUM_B2C(sbi, j_blocks); 36623c816dedSEric Whitney iput(j_inode); 36633c816dedSEric Whitney } else { 36643c816dedSEric Whitney ext4_msg(sb, KERN_ERR, "can't get journal size"); 36653c816dedSEric Whitney } 36663c816dedSEric Whitney } 3667952fc18eSTheodore Ts'o sbi->s_overhead = overhead; 3668952fc18eSTheodore Ts'o smp_wmb(); 3669952fc18eSTheodore Ts'o free_page((unsigned long) buf); 3670952fc18eSTheodore Ts'o return 0; 3671952fc18eSTheodore Ts'o } 3672952fc18eSTheodore Ts'o 3673b5799018STheodore Ts'o static void ext4_set_resv_clusters(struct super_block *sb) 367427dd4385SLukas Czerner { 367527dd4385SLukas Czerner ext4_fsblk_t resv_clusters; 3676b5799018STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 367727dd4385SLukas Czerner 367827dd4385SLukas Czerner /* 367930fac0f7SJan Kara * There's no need to reserve anything when we aren't using extents. 368030fac0f7SJan Kara * The space estimates are exact, there are no unwritten extents, 368130fac0f7SJan Kara * hole punching doesn't need new metadata... This is needed especially 368230fac0f7SJan Kara * to keep ext2/3 backward compatibility. 368330fac0f7SJan Kara */ 3684e2b911c5SDarrick J. Wong if (!ext4_has_feature_extents(sb)) 3685b5799018STheodore Ts'o return; 368630fac0f7SJan Kara /* 368727dd4385SLukas Czerner * By default we reserve 2% or 4096 clusters, whichever is smaller. 368827dd4385SLukas Czerner * This should cover the situations where we can not afford to run 368927dd4385SLukas Czerner * out of space like for example punch hole, or converting 3690556615dcSLukas Czerner * unwritten extents in delalloc path. In most cases such 369127dd4385SLukas Czerner * allocation would require 1, or 2 blocks, higher numbers are 369227dd4385SLukas Czerner * very rare. 369327dd4385SLukas Czerner */ 3694b5799018STheodore Ts'o resv_clusters = (ext4_blocks_count(sbi->s_es) >> 3695b5799018STheodore Ts'o sbi->s_cluster_bits); 369627dd4385SLukas Czerner 369727dd4385SLukas Czerner do_div(resv_clusters, 50); 369827dd4385SLukas Czerner resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 369927dd4385SLukas Czerner 3700b5799018STheodore Ts'o atomic64_set(&sbi->s_resv_clusters, resv_clusters); 370127dd4385SLukas Czerner } 370227dd4385SLukas Czerner 3703617ba13bSMingming Cao static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3704ac27a0ecSDave Kleikamp { 37055e405595SDan Williams struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev); 3706d4c402d9SCurt Wohlgemuth char *orig_data = kstrdup(data, GFP_KERNEL); 37071d0c3924STheodore Ts'o struct buffer_head *bh, **group_desc; 3708617ba13bSMingming Cao struct ext4_super_block *es = NULL; 37095aee0f8aSTheodore Ts'o struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 37107c990728SSuraj Jitindar Singh struct flex_groups **flex_groups; 3711617ba13bSMingming Cao ext4_fsblk_t block; 3712617ba13bSMingming Cao ext4_fsblk_t sb_block = get_sb_block(&data); 371370bbb3e0SAndrew Morton ext4_fsblk_t logical_sb_block; 3714ac27a0ecSDave Kleikamp unsigned long offset = 0; 3715ac27a0ecSDave Kleikamp unsigned long journal_devnum = 0; 3716ac27a0ecSDave Kleikamp unsigned long def_mount_opts; 3717ac27a0ecSDave Kleikamp struct inode *root; 37180390131bSFrank Mayhar const char *descr; 3719dcc7dae3SCyrill Gorcunov int ret = -ENOMEM; 3720281b5995STheodore Ts'o int blocksize, clustersize; 37214ec11028STheodore Ts'o unsigned int db_count; 37224ec11028STheodore Ts'o unsigned int i; 3723281b5995STheodore Ts'o int needs_recovery, has_huge_files, has_bigalloc; 3724bd81d8eeSLaurent Vivier __u64 blocks_count; 372507aa2ea1SLukas Czerner int err = 0; 3726b3881f74STheodore Ts'o unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3727bfff6873SLukas Czerner ext4_group_t first_not_zeroed; 3728ac27a0ecSDave Kleikamp 37295aee0f8aSTheodore Ts'o if ((data && !orig_data) || !sbi) 37305aee0f8aSTheodore Ts'o goto out_free_base; 3731705895b6SPekka Enberg 3732aed9eb1bSColin Ian King sbi->s_daxdev = dax_dev; 3733705895b6SPekka Enberg sbi->s_blockgroup_lock = 3734705895b6SPekka Enberg kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 37355aee0f8aSTheodore Ts'o if (!sbi->s_blockgroup_lock) 37365aee0f8aSTheodore Ts'o goto out_free_base; 37375aee0f8aSTheodore Ts'o 3738ac27a0ecSDave Kleikamp sb->s_fs_info = sbi; 37392c0544b2STheodore Ts'o sbi->s_sb = sb; 3740240799cdSTheodore Ts'o sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3741d9c9bef1SMiklos Szeredi sbi->s_sb_block = sb_block; 3742f613dfcbSTheodore Ts'o if (sb->s_bdev->bd_part) 3743f613dfcbSTheodore Ts'o sbi->s_sectors_written_start = 3744dbae2c55SMichael Callahan part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]); 3745ac27a0ecSDave Kleikamp 37469f6200bbSTheodore Ts'o /* Cleanup superblock name */ 3747ec3904dcSRasmus Villemoes strreplace(sb->s_id, '/', '!'); 37489f6200bbSTheodore Ts'o 374907aa2ea1SLukas Czerner /* -EINVAL is default */ 3750dcc7dae3SCyrill Gorcunov ret = -EINVAL; 3751617ba13bSMingming Cao blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3752ac27a0ecSDave Kleikamp if (!blocksize) { 3753b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3754ac27a0ecSDave Kleikamp goto out_fail; 3755ac27a0ecSDave Kleikamp } 3756ac27a0ecSDave Kleikamp 3757ac27a0ecSDave Kleikamp /* 3758617ba13bSMingming Cao * The ext4 superblock will not be buffer aligned for other than 1kB 3759ac27a0ecSDave Kleikamp * block sizes. We need to calculate the offset from buffer start. 3760ac27a0ecSDave Kleikamp */ 3761617ba13bSMingming Cao if (blocksize != EXT4_MIN_BLOCK_SIZE) { 376270bbb3e0SAndrew Morton logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 376370bbb3e0SAndrew Morton offset = do_div(logical_sb_block, blocksize); 3764ac27a0ecSDave Kleikamp } else { 376570bbb3e0SAndrew Morton logical_sb_block = sb_block; 3766ac27a0ecSDave Kleikamp } 3767ac27a0ecSDave Kleikamp 3768a8ac900bSGioh Kim if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) { 3769b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3770ac27a0ecSDave Kleikamp goto out_fail; 3771ac27a0ecSDave Kleikamp } 3772ac27a0ecSDave Kleikamp /* 3773ac27a0ecSDave Kleikamp * Note: s_es must be initialized as soon as possible because 3774617ba13bSMingming Cao * some ext4 macro-instructions depend on its value 3775ac27a0ecSDave Kleikamp */ 37762716b802STheodore Ts'o es = (struct ext4_super_block *) (bh->b_data + offset); 3777ac27a0ecSDave Kleikamp sbi->s_es = es; 3778ac27a0ecSDave Kleikamp sb->s_magic = le16_to_cpu(es->s_magic); 3779617ba13bSMingming Cao if (sb->s_magic != EXT4_SUPER_MAGIC) 3780617ba13bSMingming Cao goto cantfind_ext4; 3781afc32f7eSTheodore Ts'o sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3782ac27a0ecSDave Kleikamp 3783feb0ab32SDarrick J. Wong /* Warn if metadata_csum and gdt_csum are both set. */ 3784e2b911c5SDarrick J. Wong if (ext4_has_feature_metadata_csum(sb) && 3785e2b911c5SDarrick J. Wong ext4_has_feature_gdt_csum(sb)) 3786363307e6SJakub Wilk ext4_warning(sb, "metadata_csum and uninit_bg are " 3787feb0ab32SDarrick J. Wong "redundant flags; please run fsck."); 3788feb0ab32SDarrick J. Wong 3789d25425f8SDarrick J. Wong /* Check for a known checksum algorithm */ 3790d25425f8SDarrick J. Wong if (!ext4_verify_csum_type(sb, es)) { 3791d25425f8SDarrick J. Wong ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3792d25425f8SDarrick J. Wong "unknown checksum algorithm."); 3793d25425f8SDarrick J. Wong silent = 1; 3794d25425f8SDarrick J. Wong goto cantfind_ext4; 3795d25425f8SDarrick J. Wong } 3796d25425f8SDarrick J. Wong 37970441984aSDarrick J. Wong /* Load the checksum driver */ 37980441984aSDarrick J. Wong sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 37990441984aSDarrick J. Wong if (IS_ERR(sbi->s_chksum_driver)) { 38000441984aSDarrick J. Wong ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 38010441984aSDarrick J. Wong ret = PTR_ERR(sbi->s_chksum_driver); 38020441984aSDarrick J. Wong sbi->s_chksum_driver = NULL; 38030441984aSDarrick J. Wong goto failed_mount; 38040441984aSDarrick J. Wong } 38050441984aSDarrick J. Wong 3806a9c47317SDarrick J. Wong /* Check superblock checksum */ 3807a9c47317SDarrick J. Wong if (!ext4_superblock_csum_verify(sb, es)) { 3808a9c47317SDarrick J. Wong ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3809a9c47317SDarrick J. Wong "invalid superblock checksum. Run e2fsck?"); 3810a9c47317SDarrick J. Wong silent = 1; 38116a797d27SDarrick J. Wong ret = -EFSBADCRC; 3812a9c47317SDarrick J. Wong goto cantfind_ext4; 3813a9c47317SDarrick J. Wong } 3814a9c47317SDarrick J. Wong 3815a9c47317SDarrick J. Wong /* Precompute checksum seed for all metadata */ 3816e2b911c5SDarrick J. Wong if (ext4_has_feature_csum_seed(sb)) 38178c81bd8fSDarrick J. Wong sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 3818dec214d0STahsin Erdogan else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) 3819a9c47317SDarrick J. Wong sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 3820a9c47317SDarrick J. Wong sizeof(es->s_uuid)); 3821a9c47317SDarrick J. Wong 3822ac27a0ecSDave Kleikamp /* Set defaults before we parse the mount options */ 3823ac27a0ecSDave Kleikamp def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3824fd8c37ecSTheodore Ts'o set_opt(sb, INIT_INODE_TABLE); 3825617ba13bSMingming Cao if (def_mount_opts & EXT4_DEFM_DEBUG) 3826fd8c37ecSTheodore Ts'o set_opt(sb, DEBUG); 382787f26807STheodore Ts'o if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 3828fd8c37ecSTheodore Ts'o set_opt(sb, GRPID); 3829617ba13bSMingming Cao if (def_mount_opts & EXT4_DEFM_UID16) 3830fd8c37ecSTheodore Ts'o set_opt(sb, NO_UID32); 3831ea663336SEric Sandeen /* xattr user namespace & acls are now defaulted on */ 3832fd8c37ecSTheodore Ts'o set_opt(sb, XATTR_USER); 383303010a33STheodore Ts'o #ifdef CONFIG_EXT4_FS_POSIX_ACL 3834fd8c37ecSTheodore Ts'o set_opt(sb, POSIX_ACL); 38352e7842b8SHugh Dickins #endif 383698c1a759SDarrick J. Wong /* don't forget to enable journal_csum when metadata_csum is enabled. */ 383798c1a759SDarrick J. Wong if (ext4_has_metadata_csum(sb)) 383898c1a759SDarrick J. Wong set_opt(sb, JOURNAL_CHECKSUM); 383998c1a759SDarrick J. Wong 3840617ba13bSMingming Cao if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 3841fd8c37ecSTheodore Ts'o set_opt(sb, JOURNAL_DATA); 3842617ba13bSMingming Cao else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 3843fd8c37ecSTheodore Ts'o set_opt(sb, ORDERED_DATA); 3844617ba13bSMingming Cao else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 3845fd8c37ecSTheodore Ts'o set_opt(sb, WRITEBACK_DATA); 3846ac27a0ecSDave Kleikamp 3847617ba13bSMingming Cao if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 3848fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_PANIC); 3849bb4f397aSAneesh Kumar K.V else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 3850fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_CONT); 3851bb4f397aSAneesh Kumar K.V else 3852fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_RO); 385345f1a9c3SDarrick J. Wong /* block_validity enabled by default; disable with noblock_validity */ 3854fd8c37ecSTheodore Ts'o set_opt(sb, BLOCK_VALIDITY); 38558b67f04aSTheodore Ts'o if (def_mount_opts & EXT4_DEFM_DISCARD) 3856fd8c37ecSTheodore Ts'o set_opt(sb, DISCARD); 3857ac27a0ecSDave Kleikamp 385808cefc7aSEric W. Biederman sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 385908cefc7aSEric W. Biederman sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 386030773840STheodore Ts'o sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 386130773840STheodore Ts'o sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 386230773840STheodore Ts'o sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3863ac27a0ecSDave Kleikamp 38648b67f04aSTheodore Ts'o if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 3865fd8c37ecSTheodore Ts'o set_opt(sb, BARRIER); 3866ac27a0ecSDave Kleikamp 38671e2462f9SMingming Cao /* 3868dd919b98SAneesh Kumar K.V * enable delayed allocation by default 3869dd919b98SAneesh Kumar K.V * Use -o nodelalloc to turn it off 3870dd919b98SAneesh Kumar K.V */ 3871bc0b75f7STheodore Ts'o if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 38728b67f04aSTheodore Ts'o ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 3873fd8c37ecSTheodore Ts'o set_opt(sb, DELALLOC); 3874dd919b98SAneesh Kumar K.V 387551ce6511SLukas Czerner /* 387651ce6511SLukas Czerner * set default s_li_wait_mult for lazyinit, for the case there is 387751ce6511SLukas Czerner * no mount option specified. 387851ce6511SLukas Czerner */ 387951ce6511SLukas Czerner sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 388051ce6511SLukas Czerner 38814f97a681STheodore Ts'o blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 3882626b035bSRitesh Harjani 3883626b035bSRitesh Harjani if (blocksize == PAGE_SIZE) 3884626b035bSRitesh Harjani set_opt(sb, DIOREAD_NOLOCK); 3885626b035bSRitesh Harjani 38864f97a681STheodore Ts'o if (blocksize < EXT4_MIN_BLOCK_SIZE || 38874f97a681STheodore Ts'o blocksize > EXT4_MAX_BLOCK_SIZE) { 38884f97a681STheodore Ts'o ext4_msg(sb, KERN_ERR, 38894f97a681STheodore Ts'o "Unsupported filesystem blocksize %d (%d log_block_size)", 38904f97a681STheodore Ts'o blocksize, le32_to_cpu(es->s_log_block_size)); 38914f97a681STheodore Ts'o goto failed_mount; 38924f97a681STheodore Ts'o } 38934f97a681STheodore Ts'o 38949803387cSTheodore Ts'o if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 38959803387cSTheodore Ts'o sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 38969803387cSTheodore Ts'o sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 38979803387cSTheodore Ts'o } else { 38989803387cSTheodore Ts'o sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 38999803387cSTheodore Ts'o sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 39009803387cSTheodore Ts'o if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 39019803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 39029803387cSTheodore Ts'o sbi->s_first_ino); 39039803387cSTheodore Ts'o goto failed_mount; 39049803387cSTheodore Ts'o } 39059803387cSTheodore Ts'o if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 39069803387cSTheodore Ts'o (!is_power_of_2(sbi->s_inode_size)) || 39079803387cSTheodore Ts'o (sbi->s_inode_size > blocksize)) { 39089803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 39099803387cSTheodore Ts'o "unsupported inode size: %d", 39109803387cSTheodore Ts'o sbi->s_inode_size); 39114f97a681STheodore Ts'o ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize); 39129803387cSTheodore Ts'o goto failed_mount; 39139803387cSTheodore Ts'o } 39149803387cSTheodore Ts'o /* 39159803387cSTheodore Ts'o * i_atime_extra is the last extra field available for 39169803387cSTheodore Ts'o * [acm]times in struct ext4_inode. Checking for that 39179803387cSTheodore Ts'o * field should suffice to ensure we have extra space 39189803387cSTheodore Ts'o * for all three. 39199803387cSTheodore Ts'o */ 39209803387cSTheodore Ts'o if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) + 39219803387cSTheodore Ts'o sizeof(((struct ext4_inode *)0)->i_atime_extra)) { 39229803387cSTheodore Ts'o sb->s_time_gran = 1; 39239803387cSTheodore Ts'o sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX; 39249803387cSTheodore Ts'o } else { 39259803387cSTheodore Ts'o sb->s_time_gran = NSEC_PER_SEC; 39269803387cSTheodore Ts'o sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX; 39279803387cSTheodore Ts'o } 39289803387cSTheodore Ts'o sb->s_time_min = EXT4_TIMESTAMP_MIN; 39299803387cSTheodore Ts'o } 39309803387cSTheodore Ts'o if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 39319803387cSTheodore Ts'o sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 39329803387cSTheodore Ts'o EXT4_GOOD_OLD_INODE_SIZE; 39339803387cSTheodore Ts'o if (ext4_has_feature_extra_isize(sb)) { 39349803387cSTheodore Ts'o unsigned v, max = (sbi->s_inode_size - 39359803387cSTheodore Ts'o EXT4_GOOD_OLD_INODE_SIZE); 39369803387cSTheodore Ts'o 39379803387cSTheodore Ts'o v = le16_to_cpu(es->s_want_extra_isize); 39389803387cSTheodore Ts'o if (v > max) { 39399803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 39409803387cSTheodore Ts'o "bad s_want_extra_isize: %d", v); 39419803387cSTheodore Ts'o goto failed_mount; 39429803387cSTheodore Ts'o } 39439803387cSTheodore Ts'o if (sbi->s_want_extra_isize < v) 39449803387cSTheodore Ts'o sbi->s_want_extra_isize = v; 39459803387cSTheodore Ts'o 39469803387cSTheodore Ts'o v = le16_to_cpu(es->s_min_extra_isize); 39479803387cSTheodore Ts'o if (v > max) { 39489803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 39499803387cSTheodore Ts'o "bad s_min_extra_isize: %d", v); 39509803387cSTheodore Ts'o goto failed_mount; 39519803387cSTheodore Ts'o } 39529803387cSTheodore Ts'o if (sbi->s_want_extra_isize < v) 39539803387cSTheodore Ts'o sbi->s_want_extra_isize = v; 39549803387cSTheodore Ts'o } 39559803387cSTheodore Ts'o } 39569803387cSTheodore Ts'o 39575aee0f8aSTheodore Ts'o if (sbi->s_es->s_mount_opts[0]) { 39585aee0f8aSTheodore Ts'o char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, 39595aee0f8aSTheodore Ts'o sizeof(sbi->s_es->s_mount_opts), 39605aee0f8aSTheodore Ts'o GFP_KERNEL); 39615aee0f8aSTheodore Ts'o if (!s_mount_opts) 39625aee0f8aSTheodore Ts'o goto failed_mount; 39635aee0f8aSTheodore Ts'o if (!parse_options(s_mount_opts, sb, &journal_devnum, 39645aee0f8aSTheodore Ts'o &journal_ioprio, 0)) { 39658b67f04aSTheodore Ts'o ext4_msg(sb, KERN_WARNING, 39668b67f04aSTheodore Ts'o "failed to parse options in superblock: %s", 39675aee0f8aSTheodore Ts'o s_mount_opts); 39685aee0f8aSTheodore Ts'o } 39695aee0f8aSTheodore Ts'o kfree(s_mount_opts); 39708b67f04aSTheodore Ts'o } 39715a916be1STheodore Ts'o sbi->s_def_mount_opt = sbi->s_mount_opt; 3972b3881f74STheodore Ts'o if (!parse_options((char *) data, sb, &journal_devnum, 3973661aa520SEric Sandeen &journal_ioprio, 0)) 3974ac27a0ecSDave Kleikamp goto failed_mount; 3975ac27a0ecSDave Kleikamp 3976c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 3977c83ad55eSGabriel Krisman Bertazi if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) { 3978c83ad55eSGabriel Krisman Bertazi const struct ext4_sb_encodings *encoding_info; 3979c83ad55eSGabriel Krisman Bertazi struct unicode_map *encoding; 3980c83ad55eSGabriel Krisman Bertazi __u16 encoding_flags; 3981c83ad55eSGabriel Krisman Bertazi 3982c83ad55eSGabriel Krisman Bertazi if (ext4_has_feature_encrypt(sb)) { 3983c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 3984c83ad55eSGabriel Krisman Bertazi "Can't mount with encoding and encryption"); 3985c83ad55eSGabriel Krisman Bertazi goto failed_mount; 3986c83ad55eSGabriel Krisman Bertazi } 3987c83ad55eSGabriel Krisman Bertazi 3988c83ad55eSGabriel Krisman Bertazi if (ext4_sb_read_encoding(es, &encoding_info, 3989c83ad55eSGabriel Krisman Bertazi &encoding_flags)) { 3990c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 3991c83ad55eSGabriel Krisman Bertazi "Encoding requested by superblock is unknown"); 3992c83ad55eSGabriel Krisman Bertazi goto failed_mount; 3993c83ad55eSGabriel Krisman Bertazi } 3994c83ad55eSGabriel Krisman Bertazi 3995c83ad55eSGabriel Krisman Bertazi encoding = utf8_load(encoding_info->version); 3996c83ad55eSGabriel Krisman Bertazi if (IS_ERR(encoding)) { 3997c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 3998c83ad55eSGabriel Krisman Bertazi "can't mount with superblock charset: %s-%s " 3999c83ad55eSGabriel Krisman Bertazi "not supported by the kernel. flags: 0x%x.", 4000c83ad55eSGabriel Krisman Bertazi encoding_info->name, encoding_info->version, 4001c83ad55eSGabriel Krisman Bertazi encoding_flags); 4002c83ad55eSGabriel Krisman Bertazi goto failed_mount; 4003c83ad55eSGabriel Krisman Bertazi } 4004c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " 4005c83ad55eSGabriel Krisman Bertazi "%s-%s with flags 0x%hx", encoding_info->name, 4006c83ad55eSGabriel Krisman Bertazi encoding_info->version?:"\b", encoding_flags); 4007c83ad55eSGabriel Krisman Bertazi 4008c83ad55eSGabriel Krisman Bertazi sbi->s_encoding = encoding; 4009c83ad55eSGabriel Krisman Bertazi sbi->s_encoding_flags = encoding_flags; 4010c83ad55eSGabriel Krisman Bertazi } 4011c83ad55eSGabriel Krisman Bertazi #endif 4012c83ad55eSGabriel Krisman Bertazi 401356889787STheodore Ts'o if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 4014244adf64STheodore Ts'o printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n"); 4015244adf64STheodore Ts'o clear_opt(sb, DIOREAD_NOLOCK); 401656889787STheodore Ts'o if (test_opt2(sb, EXPLICIT_DELALLOC)) { 401756889787STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't mount with " 401856889787STheodore Ts'o "both data=journal and delalloc"); 401956889787STheodore Ts'o goto failed_mount; 402056889787STheodore Ts'o } 402156889787STheodore Ts'o if (test_opt(sb, DIOREAD_NOLOCK)) { 402256889787STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't mount with " 40236ae6514bSPiotr Sarna "both data=journal and dioread_nolock"); 402456889787STheodore Ts'o goto failed_mount; 402556889787STheodore Ts'o } 4026fc626fe3SIra Weiny if (test_opt(sb, DAX_ALWAYS)) { 4027923ae0ffSRoss Zwisler ext4_msg(sb, KERN_ERR, "can't mount with " 4028923ae0ffSRoss Zwisler "both data=journal and dax"); 4029923ae0ffSRoss Zwisler goto failed_mount; 4030923ae0ffSRoss Zwisler } 403173b92a2aSSergey Karamov if (ext4_has_feature_encrypt(sb)) { 403273b92a2aSSergey Karamov ext4_msg(sb, KERN_WARNING, 403373b92a2aSSergey Karamov "encrypted files will use data=ordered " 403473b92a2aSSergey Karamov "instead of data journaling mode"); 403573b92a2aSSergey Karamov } 403656889787STheodore Ts'o if (test_opt(sb, DELALLOC)) 403756889787STheodore Ts'o clear_opt(sb, DELALLOC); 4038001e4a87STejun Heo } else { 4039001e4a87STejun Heo sb->s_iflags |= SB_I_CGROUPWB; 404056889787STheodore Ts'o } 404156889787STheodore Ts'o 40421751e8a6SLinus Torvalds sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 40431751e8a6SLinus Torvalds (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 4044ac27a0ecSDave Kleikamp 4045617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 4046e2b911c5SDarrick J. Wong (ext4_has_compat_features(sb) || 4047e2b911c5SDarrick J. Wong ext4_has_ro_compat_features(sb) || 4048e2b911c5SDarrick J. Wong ext4_has_incompat_features(sb))) 4049b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 4050b31e1552SEric Sandeen "feature flags set on rev 0 fs, " 4051b31e1552SEric Sandeen "running e2fsck is recommended"); 4052469108ffSTheodore Tso 4053ed3654ebSTheodore Ts'o if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 4054ed3654ebSTheodore Ts'o set_opt2(sb, HURD_COMPAT); 4055e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb)) { 4056ed3654ebSTheodore Ts'o ext4_msg(sb, KERN_ERR, 4057ed3654ebSTheodore Ts'o "The Hurd can't support 64-bit file systems"); 4058ed3654ebSTheodore Ts'o goto failed_mount; 4059ed3654ebSTheodore Ts'o } 4060dec214d0STahsin Erdogan 4061dec214d0STahsin Erdogan /* 4062dec214d0STahsin Erdogan * ea_inode feature uses l_i_version field which is not 4063dec214d0STahsin Erdogan * available in HURD_COMPAT mode. 4064dec214d0STahsin Erdogan */ 4065dec214d0STahsin Erdogan if (ext4_has_feature_ea_inode(sb)) { 4066dec214d0STahsin Erdogan ext4_msg(sb, KERN_ERR, 4067dec214d0STahsin Erdogan "ea_inode feature is not supported for Hurd"); 4068dec214d0STahsin Erdogan goto failed_mount; 4069dec214d0STahsin Erdogan } 4070ed3654ebSTheodore Ts'o } 4071ed3654ebSTheodore Ts'o 40722035e776STheodore Ts'o if (IS_EXT2_SB(sb)) { 40732035e776STheodore Ts'o if (ext2_feature_set_ok(sb)) 40742035e776STheodore Ts'o ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 40752035e776STheodore Ts'o "using the ext4 subsystem"); 40762035e776STheodore Ts'o else { 40770d9366d6SEric Sandeen /* 40780d9366d6SEric Sandeen * If we're probing be silent, if this looks like 40790d9366d6SEric Sandeen * it's actually an ext[34] filesystem. 40800d9366d6SEric Sandeen */ 40810d9366d6SEric Sandeen if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 40820d9366d6SEric Sandeen goto failed_mount; 40832035e776STheodore Ts'o ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 40842035e776STheodore Ts'o "to feature incompatibilities"); 40852035e776STheodore Ts'o goto failed_mount; 40862035e776STheodore Ts'o } 40872035e776STheodore Ts'o } 40882035e776STheodore Ts'o 40892035e776STheodore Ts'o if (IS_EXT3_SB(sb)) { 40902035e776STheodore Ts'o if (ext3_feature_set_ok(sb)) 40912035e776STheodore Ts'o ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 40922035e776STheodore Ts'o "using the ext4 subsystem"); 40932035e776STheodore Ts'o else { 40940d9366d6SEric Sandeen /* 40950d9366d6SEric Sandeen * If we're probing be silent, if this looks like 40960d9366d6SEric Sandeen * it's actually an ext4 filesystem. 40970d9366d6SEric Sandeen */ 40980d9366d6SEric Sandeen if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 40990d9366d6SEric Sandeen goto failed_mount; 41002035e776STheodore Ts'o ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 41012035e776STheodore Ts'o "to feature incompatibilities"); 41022035e776STheodore Ts'o goto failed_mount; 41032035e776STheodore Ts'o } 41042035e776STheodore Ts'o } 41052035e776STheodore Ts'o 4106469108ffSTheodore Tso /* 4107ac27a0ecSDave Kleikamp * Check feature flags regardless of the revision level, since we 4108ac27a0ecSDave Kleikamp * previously didn't change the revision level when setting the flags, 4109ac27a0ecSDave Kleikamp * so there is a chance incompat flags are set on a rev 0 filesystem. 4110ac27a0ecSDave Kleikamp */ 4111bc98a42cSDavid Howells if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 4112ac27a0ecSDave Kleikamp goto failed_mount; 4113a13fb1a4SEric Sandeen 41148cdf3372STheodore Ts'o if (le32_to_cpu(es->s_log_block_size) > 41158cdf3372STheodore Ts'o (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 41168cdf3372STheodore Ts'o ext4_msg(sb, KERN_ERR, 41178cdf3372STheodore Ts'o "Invalid log block size: %u", 41188cdf3372STheodore Ts'o le32_to_cpu(es->s_log_block_size)); 4119ac27a0ecSDave Kleikamp goto failed_mount; 4120ac27a0ecSDave Kleikamp } 4121bfe0a5f4STheodore Ts'o if (le32_to_cpu(es->s_log_cluster_size) > 4122bfe0a5f4STheodore Ts'o (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 4123bfe0a5f4STheodore Ts'o ext4_msg(sb, KERN_ERR, 4124bfe0a5f4STheodore Ts'o "Invalid log cluster size: %u", 4125bfe0a5f4STheodore Ts'o le32_to_cpu(es->s_log_cluster_size)); 4126bfe0a5f4STheodore Ts'o goto failed_mount; 4127bfe0a5f4STheodore Ts'o } 4128ac27a0ecSDave Kleikamp 41295b9554dcSTheodore Ts'o if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { 41305b9554dcSTheodore Ts'o ext4_msg(sb, KERN_ERR, 41315b9554dcSTheodore Ts'o "Number of reserved GDT blocks insanely large: %d", 41325b9554dcSTheodore Ts'o le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 41335b9554dcSTheodore Ts'o goto failed_mount; 41345b9554dcSTheodore Ts'o } 41355b9554dcSTheodore Ts'o 4136a8ab6d38SIra Weiny if (bdev_dax_supported(sb->s_bdev, blocksize)) 4137a8ab6d38SIra Weiny set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags); 4138a8ab6d38SIra Weiny 4139fc626fe3SIra Weiny if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) { 4140559db4c6SRoss Zwisler if (ext4_has_feature_inline_data(sb)) { 4141559db4c6SRoss Zwisler ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 4142559db4c6SRoss Zwisler " that may contain inline data"); 4143361d24d4SEric Sandeen goto failed_mount; 4144559db4c6SRoss Zwisler } 4145a8ab6d38SIra Weiny if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) { 414624f3478dSDan Williams ext4_msg(sb, KERN_ERR, 4147361d24d4SEric Sandeen "DAX unsupported by block device."); 4148361d24d4SEric Sandeen goto failed_mount; 414924f3478dSDan Williams } 4150923ae0ffSRoss Zwisler } 4151923ae0ffSRoss Zwisler 4152e2b911c5SDarrick J. Wong if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 41536ddb2447STheodore Ts'o ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 41546ddb2447STheodore Ts'o es->s_encryption_level); 41556ddb2447STheodore Ts'o goto failed_mount; 41566ddb2447STheodore Ts'o } 41576ddb2447STheodore Ts'o 4158ac27a0ecSDave Kleikamp if (sb->s_blocksize != blocksize) { 4159ce40733cSAneesh Kumar K.V /* Validate the filesystem blocksize */ 4160ce40733cSAneesh Kumar K.V if (!sb_set_blocksize(sb, blocksize)) { 4161b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "bad block size %d", 4162ce40733cSAneesh Kumar K.V blocksize); 4163ac27a0ecSDave Kleikamp goto failed_mount; 4164ac27a0ecSDave Kleikamp } 4165ac27a0ecSDave Kleikamp 4166ac27a0ecSDave Kleikamp brelse(bh); 416770bbb3e0SAndrew Morton logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 416870bbb3e0SAndrew Morton offset = do_div(logical_sb_block, blocksize); 4169a8ac900bSGioh Kim bh = sb_bread_unmovable(sb, logical_sb_block); 4170ac27a0ecSDave Kleikamp if (!bh) { 4171b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4172b31e1552SEric Sandeen "Can't read superblock on 2nd try"); 4173ac27a0ecSDave Kleikamp goto failed_mount; 4174ac27a0ecSDave Kleikamp } 41752716b802STheodore Ts'o es = (struct ext4_super_block *)(bh->b_data + offset); 4176ac27a0ecSDave Kleikamp sbi->s_es = es; 4177617ba13bSMingming Cao if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 4178b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4179b31e1552SEric Sandeen "Magic mismatch, very weird!"); 4180ac27a0ecSDave Kleikamp goto failed_mount; 4181ac27a0ecSDave Kleikamp } 4182ac27a0ecSDave Kleikamp } 4183ac27a0ecSDave Kleikamp 4184e2b911c5SDarrick J. Wong has_huge_files = ext4_has_feature_huge_file(sb); 4185f287a1a5STheodore Ts'o sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 4186f287a1a5STheodore Ts'o has_huge_files); 4187f287a1a5STheodore Ts'o sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 4188ac27a0ecSDave Kleikamp 41890d1ee42fSAlexandre Ratchov sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 4190e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb)) { 41918fadc143SAlexandre Ratchov if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 41920d1ee42fSAlexandre Ratchov sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 4193d8ea6cf8Svignesh babu !is_power_of_2(sbi->s_desc_size)) { 4194b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4195b31e1552SEric Sandeen "unsupported descriptor size %lu", 41960d1ee42fSAlexandre Ratchov sbi->s_desc_size); 41970d1ee42fSAlexandre Ratchov goto failed_mount; 41980d1ee42fSAlexandre Ratchov } 41990d1ee42fSAlexandre Ratchov } else 42000d1ee42fSAlexandre Ratchov sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 42010b8e58a1SAndreas Dilger 4202ac27a0ecSDave Kleikamp sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 4203ac27a0ecSDave Kleikamp sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 42040b8e58a1SAndreas Dilger 4205617ba13bSMingming Cao sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 4206ac27a0ecSDave Kleikamp if (sbi->s_inodes_per_block == 0) 4207617ba13bSMingming Cao goto cantfind_ext4; 4208cd6bb35bSTheodore Ts'o if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 4209cd6bb35bSTheodore Ts'o sbi->s_inodes_per_group > blocksize * 8) { 4210cd6bb35bSTheodore Ts'o ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 4211b9c538daSJosh Triplett sbi->s_inodes_per_group); 4212cd6bb35bSTheodore Ts'o goto failed_mount; 4213cd6bb35bSTheodore Ts'o } 4214ac27a0ecSDave Kleikamp sbi->s_itb_per_group = sbi->s_inodes_per_group / 4215ac27a0ecSDave Kleikamp sbi->s_inodes_per_block; 42160d1ee42fSAlexandre Ratchov sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 4217ac27a0ecSDave Kleikamp sbi->s_sbh = bh; 4218ac27a0ecSDave Kleikamp sbi->s_mount_state = le16_to_cpu(es->s_state); 4219e57aa839SFengguang Wu sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 4220e57aa839SFengguang Wu sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 42210b8e58a1SAndreas Dilger 4222ac27a0ecSDave Kleikamp for (i = 0; i < 4; i++) 4223ac27a0ecSDave Kleikamp sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 4224ac27a0ecSDave Kleikamp sbi->s_def_hash_version = es->s_def_hash_version; 4225e2b911c5SDarrick J. Wong if (ext4_has_feature_dir_index(sb)) { 4226f99b2589STheodore Ts'o i = le32_to_cpu(es->s_flags); 4227f99b2589STheodore Ts'o if (i & EXT2_FLAGS_UNSIGNED_HASH) 4228f99b2589STheodore Ts'o sbi->s_hash_unsigned = 3; 4229f99b2589STheodore Ts'o else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 4230f99b2589STheodore Ts'o #ifdef __CHAR_UNSIGNED__ 4231bc98a42cSDavid Howells if (!sb_rdonly(sb)) 423223301410STheodore Ts'o es->s_flags |= 423323301410STheodore Ts'o cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 4234f99b2589STheodore Ts'o sbi->s_hash_unsigned = 3; 4235f99b2589STheodore Ts'o #else 4236bc98a42cSDavid Howells if (!sb_rdonly(sb)) 423723301410STheodore Ts'o es->s_flags |= 423823301410STheodore Ts'o cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 4239f99b2589STheodore Ts'o #endif 4240f99b2589STheodore Ts'o } 424123301410STheodore Ts'o } 4242ac27a0ecSDave Kleikamp 4243281b5995STheodore Ts'o /* Handle clustersize */ 4244281b5995STheodore Ts'o clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 4245e2b911c5SDarrick J. Wong has_bigalloc = ext4_has_feature_bigalloc(sb); 4246281b5995STheodore Ts'o if (has_bigalloc) { 4247281b5995STheodore Ts'o if (clustersize < blocksize) { 4248281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, 4249281b5995STheodore Ts'o "cluster size (%d) smaller than " 4250281b5995STheodore Ts'o "block size (%d)", clustersize, blocksize); 4251281b5995STheodore Ts'o goto failed_mount; 4252281b5995STheodore Ts'o } 4253281b5995STheodore Ts'o sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 4254281b5995STheodore Ts'o le32_to_cpu(es->s_log_block_size); 4255281b5995STheodore Ts'o sbi->s_clusters_per_group = 4256281b5995STheodore Ts'o le32_to_cpu(es->s_clusters_per_group); 4257281b5995STheodore Ts'o if (sbi->s_clusters_per_group > blocksize * 8) { 4258281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, 4259281b5995STheodore Ts'o "#clusters per group too big: %lu", 4260281b5995STheodore Ts'o sbi->s_clusters_per_group); 4261281b5995STheodore Ts'o goto failed_mount; 4262281b5995STheodore Ts'o } 4263281b5995STheodore Ts'o if (sbi->s_blocks_per_group != 4264281b5995STheodore Ts'o (sbi->s_clusters_per_group * (clustersize / blocksize))) { 4265281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 4266281b5995STheodore Ts'o "clusters per group (%lu) inconsistent", 4267281b5995STheodore Ts'o sbi->s_blocks_per_group, 4268281b5995STheodore Ts'o sbi->s_clusters_per_group); 4269281b5995STheodore Ts'o goto failed_mount; 4270281b5995STheodore Ts'o } 4271281b5995STheodore Ts'o } else { 4272281b5995STheodore Ts'o if (clustersize != blocksize) { 4273bfe0a5f4STheodore Ts'o ext4_msg(sb, KERN_ERR, 4274bfe0a5f4STheodore Ts'o "fragment/cluster size (%d) != " 4275bfe0a5f4STheodore Ts'o "block size (%d)", clustersize, blocksize); 4276bfe0a5f4STheodore Ts'o goto failed_mount; 4277281b5995STheodore Ts'o } 4278ac27a0ecSDave Kleikamp if (sbi->s_blocks_per_group > blocksize * 8) { 4279b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4280b31e1552SEric Sandeen "#blocks per group too big: %lu", 4281ac27a0ecSDave Kleikamp sbi->s_blocks_per_group); 4282ac27a0ecSDave Kleikamp goto failed_mount; 4283ac27a0ecSDave Kleikamp } 4284281b5995STheodore Ts'o sbi->s_clusters_per_group = sbi->s_blocks_per_group; 4285281b5995STheodore Ts'o sbi->s_cluster_bits = 0; 4286281b5995STheodore Ts'o } 4287281b5995STheodore Ts'o sbi->s_cluster_ratio = clustersize / blocksize; 4288281b5995STheodore Ts'o 4289960fd856STheodore Ts'o /* Do we have standard group size of clustersize * 8 blocks ? */ 4290960fd856STheodore Ts'o if (sbi->s_blocks_per_group == clustersize << 3) 4291960fd856STheodore Ts'o set_opt2(sb, STD_GROUP_SIZE); 4292960fd856STheodore Ts'o 4293bf43d84bSEric Sandeen /* 4294bf43d84bSEric Sandeen * Test whether we have more sectors than will fit in sector_t, 4295bf43d84bSEric Sandeen * and whether the max offset is addressable by the page cache. 4296bf43d84bSEric Sandeen */ 42975a9ae68aSDarrick J. Wong err = generic_check_addressable(sb->s_blocksize_bits, 429830ca22c7SPatrick J. LoPresti ext4_blocks_count(es)); 42995a9ae68aSDarrick J. Wong if (err) { 4300b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "filesystem" 4301bf43d84bSEric Sandeen " too large to mount safely on this system"); 4302ac27a0ecSDave Kleikamp goto failed_mount; 4303ac27a0ecSDave Kleikamp } 4304ac27a0ecSDave Kleikamp 4305617ba13bSMingming Cao if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 4306617ba13bSMingming Cao goto cantfind_ext4; 4307e7c95593SEric Sandeen 43080f2ddca6SFrom: Thiemo Nagel /* check blocks count against device size */ 43090f2ddca6SFrom: Thiemo Nagel blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 43100f2ddca6SFrom: Thiemo Nagel if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4311b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4312b31e1552SEric Sandeen "exceeds size of device (%llu blocks)", 43130f2ddca6SFrom: Thiemo Nagel ext4_blocks_count(es), blocks_count); 43140f2ddca6SFrom: Thiemo Nagel goto failed_mount; 43150f2ddca6SFrom: Thiemo Nagel } 43160f2ddca6SFrom: Thiemo Nagel 43174ec11028STheodore Ts'o /* 43184ec11028STheodore Ts'o * It makes no sense for the first data block to be beyond the end 43194ec11028STheodore Ts'o * of the filesystem. 43204ec11028STheodore Ts'o */ 43214ec11028STheodore Ts'o if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4322b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4323b31e1552SEric Sandeen "block %u is beyond end of filesystem (%llu)", 4324e7c95593SEric Sandeen le32_to_cpu(es->s_first_data_block), 43254ec11028STheodore Ts'o ext4_blocks_count(es)); 4326e7c95593SEric Sandeen goto failed_mount; 4327e7c95593SEric Sandeen } 4328bfe0a5f4STheodore Ts'o if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4329bfe0a5f4STheodore Ts'o (sbi->s_cluster_ratio == 1)) { 4330bfe0a5f4STheodore Ts'o ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4331bfe0a5f4STheodore Ts'o "block is 0 with a 1k block and cluster size"); 4332bfe0a5f4STheodore Ts'o goto failed_mount; 4333bfe0a5f4STheodore Ts'o } 4334bfe0a5f4STheodore Ts'o 4335bd81d8eeSLaurent Vivier blocks_count = (ext4_blocks_count(es) - 4336bd81d8eeSLaurent Vivier le32_to_cpu(es->s_first_data_block) + 4337bd81d8eeSLaurent Vivier EXT4_BLOCKS_PER_GROUP(sb) - 1); 4338bd81d8eeSLaurent Vivier do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 43394ec11028STheodore Ts'o if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4340df41460aSJosh Triplett ext4_msg(sb, KERN_WARNING, "groups count too large: %llu " 43414ec11028STheodore Ts'o "(block count %llu, first data block %u, " 4342df41460aSJosh Triplett "blocks per group %lu)", blocks_count, 43434ec11028STheodore Ts'o ext4_blocks_count(es), 43444ec11028STheodore Ts'o le32_to_cpu(es->s_first_data_block), 43454ec11028STheodore Ts'o EXT4_BLOCKS_PER_GROUP(sb)); 43464ec11028STheodore Ts'o goto failed_mount; 43474ec11028STheodore Ts'o } 4348bd81d8eeSLaurent Vivier sbi->s_groups_count = blocks_count; 4349fb0a387dSEric Sandeen sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 4350fb0a387dSEric Sandeen (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 43519e463084STheodore Ts'o if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 43529e463084STheodore Ts'o le32_to_cpu(es->s_inodes_count)) { 43539e463084STheodore Ts'o ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 43549e463084STheodore Ts'o le32_to_cpu(es->s_inodes_count), 43559e463084STheodore Ts'o ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 43569e463084STheodore Ts'o ret = -EINVAL; 43579e463084STheodore Ts'o goto failed_mount; 43589e463084STheodore Ts'o } 4359617ba13bSMingming Cao db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4360617ba13bSMingming Cao EXT4_DESC_PER_BLOCK(sb); 43613a4b77cdSEryu Guan if (ext4_has_feature_meta_bg(sb)) { 43622ba3e6e8STheodore Ts'o if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 43633a4b77cdSEryu Guan ext4_msg(sb, KERN_WARNING, 43643a4b77cdSEryu Guan "first meta block group too large: %u " 43653a4b77cdSEryu Guan "(group descriptor block count %u)", 43663a4b77cdSEryu Guan le32_to_cpu(es->s_first_meta_bg), db_count); 43673a4b77cdSEryu Guan goto failed_mount; 43683a4b77cdSEryu Guan } 43693a4b77cdSEryu Guan } 43701d0c3924STheodore Ts'o rcu_assign_pointer(sbi->s_group_desc, 43711d0c3924STheodore Ts'o kvmalloc_array(db_count, 4372f18a5f21STheodore Ts'o sizeof(struct buffer_head *), 43731d0c3924STheodore Ts'o GFP_KERNEL)); 4374ac27a0ecSDave Kleikamp if (sbi->s_group_desc == NULL) { 4375b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "not enough memory"); 43762cde417dSTheodore Ts'o ret = -ENOMEM; 4377ac27a0ecSDave Kleikamp goto failed_mount; 4378ac27a0ecSDave Kleikamp } 4379ac27a0ecSDave Kleikamp 4380705895b6SPekka Enberg bgl_lock_init(sbi->s_blockgroup_lock); 4381ac27a0ecSDave Kleikamp 438285c8f176SAndrew Perepechko /* Pre-read the descriptors into the buffer cache */ 438385c8f176SAndrew Perepechko for (i = 0; i < db_count; i++) { 438485c8f176SAndrew Perepechko block = descriptor_loc(sb, logical_sb_block, i); 4385d87f6392SRoman Gushchin sb_breadahead_unmovable(sb, block); 438685c8f176SAndrew Perepechko } 438785c8f176SAndrew Perepechko 4388ac27a0ecSDave Kleikamp for (i = 0; i < db_count; i++) { 43891d0c3924STheodore Ts'o struct buffer_head *bh; 43901d0c3924STheodore Ts'o 439170bbb3e0SAndrew Morton block = descriptor_loc(sb, logical_sb_block, i); 43921d0c3924STheodore Ts'o bh = sb_bread_unmovable(sb, block); 43931d0c3924STheodore Ts'o if (!bh) { 4394b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4395b31e1552SEric Sandeen "can't read group descriptor %d", i); 4396ac27a0ecSDave Kleikamp db_count = i; 4397ac27a0ecSDave Kleikamp goto failed_mount2; 4398ac27a0ecSDave Kleikamp } 43991d0c3924STheodore Ts'o rcu_read_lock(); 44001d0c3924STheodore Ts'o rcu_dereference(sbi->s_group_desc)[i] = bh; 44011d0c3924STheodore Ts'o rcu_read_unlock(); 4402ac27a0ecSDave Kleikamp } 440344de022cSTheodore Ts'o sbi->s_gdb_count = db_count; 4404829fa70dSTheodore Ts'o if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { 4405b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 44066a797d27SDarrick J. Wong ret = -EFSCORRUPTED; 4407f9ae9cf5STheodore Ts'o goto failed_mount2; 4408ac27a0ecSDave Kleikamp } 4409772cb7c8SJose R. Santos 4410235699a8SKees Cook timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 441104496411STao Ma 4412a75ae78fSDmitry Monakhov /* Register extent status tree shrinker */ 4413eb68d0e2SZheng Liu if (ext4_es_register_shrinker(sbi)) 4414ce7e010aSTheodore Ts'o goto failed_mount3; 4415ce7e010aSTheodore Ts'o 4416c9de560dSAlex Tomas sbi->s_stripe = ext4_get_stripe_size(sbi); 441767a5da56SZheng Liu sbi->s_extent_max_zeroout_kb = 32; 4418c9de560dSAlex Tomas 4419f9ae9cf5STheodore Ts'o /* 4420f9ae9cf5STheodore Ts'o * set up enough so that it can read an inode 4421f9ae9cf5STheodore Ts'o */ 4422f9ae9cf5STheodore Ts'o sb->s_op = &ext4_sops; 4423617ba13bSMingming Cao sb->s_export_op = &ext4_export_ops; 4424617ba13bSMingming Cao sb->s_xattr = ext4_xattr_handlers; 4425643fa961SChandan Rajendra #ifdef CONFIG_FS_ENCRYPTION 4426a7550b30SJaegeuk Kim sb->s_cop = &ext4_cryptops; 4427ffcc4182SEric Biggers #endif 4428c93d8f88SEric Biggers #ifdef CONFIG_FS_VERITY 4429c93d8f88SEric Biggers sb->s_vop = &ext4_verityops; 4430c93d8f88SEric Biggers #endif 4431ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4432617ba13bSMingming Cao sb->dq_op = &ext4_quota_operations; 4433e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) 44341fa5efe3SJan Kara sb->s_qcop = &dquot_quotactl_sysfile_ops; 4435262b4662SJan Kara else 4436262b4662SJan Kara sb->s_qcop = &ext4_qctl_operations; 4437689c958cSLi Xi sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4438ac27a0ecSDave Kleikamp #endif 443985787090SChristoph Hellwig memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 4440f2fa2ffcSAneesh Kumar K.V 4441ac27a0ecSDave Kleikamp INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 44423b9d4ed2STheodore Ts'o mutex_init(&sbi->s_orphan_lock); 4443ac27a0ecSDave Kleikamp 4444ac27a0ecSDave Kleikamp sb->s_root = NULL; 4445ac27a0ecSDave Kleikamp 4446ac27a0ecSDave Kleikamp needs_recovery = (es->s_last_orphan != 0 || 4447e2b911c5SDarrick J. Wong ext4_has_feature_journal_needs_recovery(sb)); 4448ac27a0ecSDave Kleikamp 4449bc98a42cSDavid Howells if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) 4450c5e06d10SJohann Lombardi if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 445150460fe8SDarrick J. Wong goto failed_mount3a; 4452c5e06d10SJohann Lombardi 4453ac27a0ecSDave Kleikamp /* 4454ac27a0ecSDave Kleikamp * The first inode we look at is the journal inode. Don't try 4455ac27a0ecSDave Kleikamp * root first: it may be modified in the journal! 4456ac27a0ecSDave Kleikamp */ 4457e2b911c5SDarrick J. Wong if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 44584753d8a2STheodore Ts'o err = ext4_load_journal(sb, es, journal_devnum); 44594753d8a2STheodore Ts'o if (err) 446050460fe8SDarrick J. Wong goto failed_mount3a; 4461bc98a42cSDavid Howells } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 4462e2b911c5SDarrick J. Wong ext4_has_feature_journal_needs_recovery(sb)) { 4463b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "required journal recovery " 4464b31e1552SEric Sandeen "suppressed and not mounted read-only"); 4465744692dcSJiaying Zhang goto failed_mount_wq; 4466ac27a0ecSDave Kleikamp } else { 44671e381f60SDmitry Monakhov /* Nojournal mode, all journal mount options are illegal */ 44681e381f60SDmitry Monakhov if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 44691e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 44701e381f60SDmitry Monakhov "journal_checksum, fs mounted w/o journal"); 44711e381f60SDmitry Monakhov goto failed_mount_wq; 44721e381f60SDmitry Monakhov } 44731e381f60SDmitry Monakhov if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 44741e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 44751e381f60SDmitry Monakhov "journal_async_commit, fs mounted w/o journal"); 44761e381f60SDmitry Monakhov goto failed_mount_wq; 44771e381f60SDmitry Monakhov } 44781e381f60SDmitry Monakhov if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 44791e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 44801e381f60SDmitry Monakhov "commit=%lu, fs mounted w/o journal", 44811e381f60SDmitry Monakhov sbi->s_commit_interval / HZ); 44821e381f60SDmitry Monakhov goto failed_mount_wq; 44831e381f60SDmitry Monakhov } 44841e381f60SDmitry Monakhov if (EXT4_MOUNT_DATA_FLAGS & 44851e381f60SDmitry Monakhov (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 44861e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 44871e381f60SDmitry Monakhov "data=, fs mounted w/o journal"); 44881e381f60SDmitry Monakhov goto failed_mount_wq; 44891e381f60SDmitry Monakhov } 449050b29d8fSDebabrata Banerjee sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; 44911e381f60SDmitry Monakhov clear_opt(sb, JOURNAL_CHECKSUM); 4492fd8c37ecSTheodore Ts'o clear_opt(sb, DATA_FLAGS); 44930390131bSFrank Mayhar sbi->s_journal = NULL; 44940390131bSFrank Mayhar needs_recovery = 0; 44950390131bSFrank Mayhar goto no_journal; 4496ac27a0ecSDave Kleikamp } 4497ac27a0ecSDave Kleikamp 4498e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb) && 4499eb40a09cSJose R. Santos !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4500eb40a09cSJose R. Santos JBD2_FEATURE_INCOMPAT_64BIT)) { 4501b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 4502744692dcSJiaying Zhang goto failed_mount_wq; 4503eb40a09cSJose R. Santos } 4504eb40a09cSJose R. Santos 450525ed6e8aSDarrick J. Wong if (!set_journal_csum_feature_set(sb)) { 450625ed6e8aSDarrick J. Wong ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 450725ed6e8aSDarrick J. Wong "feature set"); 450825ed6e8aSDarrick J. Wong goto failed_mount_wq; 4509d4da6c9cSLinus Torvalds } 4510818d276cSGirish Shilamkar 4511ac27a0ecSDave Kleikamp /* We have now updated the journal if required, so we can 4512ac27a0ecSDave Kleikamp * validate the data journaling mode. */ 4513ac27a0ecSDave Kleikamp switch (test_opt(sb, DATA_FLAGS)) { 4514ac27a0ecSDave Kleikamp case 0: 4515ac27a0ecSDave Kleikamp /* No mode set, assume a default based on the journal 451663f57933SAndrew Morton * capabilities: ORDERED_DATA if the journal can 451763f57933SAndrew Morton * cope, else JOURNAL_DATA 451863f57933SAndrew Morton */ 4519dab291afSMingming Cao if (jbd2_journal_check_available_features 452027f394a7STyson Nottingham (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4521fd8c37ecSTheodore Ts'o set_opt(sb, ORDERED_DATA); 452227f394a7STyson Nottingham sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 452327f394a7STyson Nottingham } else { 4524fd8c37ecSTheodore Ts'o set_opt(sb, JOURNAL_DATA); 452527f394a7STyson Nottingham sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 452627f394a7STyson Nottingham } 4527ac27a0ecSDave Kleikamp break; 4528ac27a0ecSDave Kleikamp 4529617ba13bSMingming Cao case EXT4_MOUNT_ORDERED_DATA: 4530617ba13bSMingming Cao case EXT4_MOUNT_WRITEBACK_DATA: 4531dab291afSMingming Cao if (!jbd2_journal_check_available_features 4532dab291afSMingming Cao (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4533b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Journal does not support " 4534b31e1552SEric Sandeen "requested data journaling mode"); 4535744692dcSJiaying Zhang goto failed_mount_wq; 4536ac27a0ecSDave Kleikamp } 4537ac27a0ecSDave Kleikamp default: 4538ac27a0ecSDave Kleikamp break; 4539ac27a0ecSDave Kleikamp } 4540ab04df78SJan Kara 4541ab04df78SJan Kara if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 4542ab04df78SJan Kara test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4543ab04df78SJan Kara ext4_msg(sb, KERN_ERR, "can't mount with " 4544ab04df78SJan Kara "journal_async_commit in data=ordered mode"); 4545ab04df78SJan Kara goto failed_mount_wq; 4546ab04df78SJan Kara } 4547ab04df78SJan Kara 4548b3881f74STheodore Ts'o set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4549ac27a0ecSDave Kleikamp 455018aadd47SBobi Jam sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; 455118aadd47SBobi Jam 4552ce7e010aSTheodore Ts'o no_journal: 4553cdb7ee4cSTahsin Erdogan if (!test_opt(sb, NO_MBCACHE)) { 455447387409STahsin Erdogan sbi->s_ea_block_cache = ext4_xattr_create_cache(); 455547387409STahsin Erdogan if (!sbi->s_ea_block_cache) { 4556cdb7ee4cSTahsin Erdogan ext4_msg(sb, KERN_ERR, 4557cdb7ee4cSTahsin Erdogan "Failed to create ea_block_cache"); 45589c191f70ST Makphaibulchoke goto failed_mount_wq; 45599c191f70ST Makphaibulchoke } 45609c191f70ST Makphaibulchoke 4561dec214d0STahsin Erdogan if (ext4_has_feature_ea_inode(sb)) { 4562dec214d0STahsin Erdogan sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 4563dec214d0STahsin Erdogan if (!sbi->s_ea_inode_cache) { 4564dec214d0STahsin Erdogan ext4_msg(sb, KERN_ERR, 4565dec214d0STahsin Erdogan "Failed to create ea_inode_cache"); 4566dec214d0STahsin Erdogan goto failed_mount_wq; 4567dec214d0STahsin Erdogan } 4568dec214d0STahsin Erdogan } 4569cdb7ee4cSTahsin Erdogan } 4570dec214d0STahsin Erdogan 4571c93d8f88SEric Biggers if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) { 4572c93d8f88SEric Biggers ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity"); 4573c93d8f88SEric Biggers goto failed_mount_wq; 4574c93d8f88SEric Biggers } 4575c93d8f88SEric Biggers 4576bc98a42cSDavid Howells if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) && 4577e2b911c5SDarrick J. Wong !ext4_has_feature_encrypt(sb)) { 4578e2b911c5SDarrick J. Wong ext4_set_feature_encrypt(sb); 45796ddb2447STheodore Ts'o ext4_commit_super(sb, 1); 45806ddb2447STheodore Ts'o } 45816ddb2447STheodore Ts'o 4582fd89d5f2STejun Heo /* 4583952fc18eSTheodore Ts'o * Get the # of file system overhead blocks from the 4584952fc18eSTheodore Ts'o * superblock if present. 4585952fc18eSTheodore Ts'o */ 4586952fc18eSTheodore Ts'o if (es->s_overhead_clusters) 4587952fc18eSTheodore Ts'o sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 4588952fc18eSTheodore Ts'o else { 458907aa2ea1SLukas Czerner err = ext4_calculate_overhead(sb); 459007aa2ea1SLukas Czerner if (err) 4591952fc18eSTheodore Ts'o goto failed_mount_wq; 4592952fc18eSTheodore Ts'o } 4593952fc18eSTheodore Ts'o 4594952fc18eSTheodore Ts'o /* 4595fd89d5f2STejun Heo * The maximum number of concurrent works can be high and 4596fd89d5f2STejun Heo * concurrency isn't really necessary. Limit it to 1. 4597fd89d5f2STejun Heo */ 45982e8fa54eSJan Kara EXT4_SB(sb)->rsv_conversion_wq = 45992e8fa54eSJan Kara alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 46002e8fa54eSJan Kara if (!EXT4_SB(sb)->rsv_conversion_wq) { 46012e8fa54eSJan Kara printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 460207aa2ea1SLukas Czerner ret = -ENOMEM; 46032e8fa54eSJan Kara goto failed_mount4; 46042e8fa54eSJan Kara } 46052e8fa54eSJan Kara 4606ac27a0ecSDave Kleikamp /* 4607dab291afSMingming Cao * The jbd2_journal_load will have done any necessary log recovery, 4608ac27a0ecSDave Kleikamp * so we can safely mount the rest of the filesystem now. 4609ac27a0ecSDave Kleikamp */ 4610ac27a0ecSDave Kleikamp 46118a363970STheodore Ts'o root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 46121d1fe1eeSDavid Howells if (IS_ERR(root)) { 4613b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "get root inode failed"); 46141d1fe1eeSDavid Howells ret = PTR_ERR(root); 461532a9bb57SManish Katiyar root = NULL; 4616ac27a0ecSDave Kleikamp goto failed_mount4; 4617ac27a0ecSDave Kleikamp } 4618ac27a0ecSDave Kleikamp if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 4619b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 462094bf608aSAl Viro iput(root); 4621ac27a0ecSDave Kleikamp goto failed_mount4; 4622ac27a0ecSDave Kleikamp } 4623b886ee3eSGabriel Krisman Bertazi 4624b886ee3eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 4625b886ee3eSGabriel Krisman Bertazi if (sbi->s_encoding) 4626b886ee3eSGabriel Krisman Bertazi sb->s_d_op = &ext4_dentry_ops; 4627b886ee3eSGabriel Krisman Bertazi #endif 4628b886ee3eSGabriel Krisman Bertazi 462948fde701SAl Viro sb->s_root = d_make_root(root); 46301d1fe1eeSDavid Howells if (!sb->s_root) { 4631b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "get root dentry failed"); 46321d1fe1eeSDavid Howells ret = -ENOMEM; 46331d1fe1eeSDavid Howells goto failed_mount4; 46341d1fe1eeSDavid Howells } 4635ac27a0ecSDave Kleikamp 4636c89128a0SJaegeuk Kim ret = ext4_setup_super(sb, es, sb_rdonly(sb)); 4637c89128a0SJaegeuk Kim if (ret == -EROFS) { 46381751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 4639c89128a0SJaegeuk Kim ret = 0; 4640c89128a0SJaegeuk Kim } else if (ret) 4641c89128a0SJaegeuk Kim goto failed_mount4a; 4642ef7f3835SKalpak Shah 4643b5799018STheodore Ts'o ext4_set_resv_clusters(sb); 464427dd4385SLukas Czerner 46456fd058f7STheodore Ts'o err = ext4_setup_system_zone(sb); 46466fd058f7STheodore Ts'o if (err) { 4647b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "failed to initialize system " 4648fbe845ddSCurt Wohlgemuth "zone (%d)", err); 4649f9ae9cf5STheodore Ts'o goto failed_mount4a; 4650f9ae9cf5STheodore Ts'o } 4651f9ae9cf5STheodore Ts'o 4652f9ae9cf5STheodore Ts'o ext4_ext_init(sb); 4653f9ae9cf5STheodore Ts'o err = ext4_mb_init(sb); 4654f9ae9cf5STheodore Ts'o if (err) { 4655f9ae9cf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 4656f9ae9cf5STheodore Ts'o err); 4657dcf2d804STao Ma goto failed_mount5; 4658c2774d84SAneesh Kumar K.V } 4659c2774d84SAneesh Kumar K.V 4660d5e03cbbSTheodore Ts'o block = ext4_count_free_clusters(sb); 4661d5e03cbbSTheodore Ts'o ext4_free_blocks_count_set(sbi->s_es, 4662d5e03cbbSTheodore Ts'o EXT4_C2B(sbi, block)); 46634274f516STheodore Ts'o ext4_superblock_csum_set(sb); 4664908c7f19STejun Heo err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 4665908c7f19STejun Heo GFP_KERNEL); 4666d5e03cbbSTheodore Ts'o if (!err) { 4667d5e03cbbSTheodore Ts'o unsigned long freei = ext4_count_free_inodes(sb); 4668d5e03cbbSTheodore Ts'o sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 46694274f516STheodore Ts'o ext4_superblock_csum_set(sb); 4670908c7f19STejun Heo err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 4671908c7f19STejun Heo GFP_KERNEL); 4672d5e03cbbSTheodore Ts'o } 4673d5e03cbbSTheodore Ts'o if (!err) 4674d5e03cbbSTheodore Ts'o err = percpu_counter_init(&sbi->s_dirs_counter, 4675908c7f19STejun Heo ext4_count_dirs(sb), GFP_KERNEL); 4676d5e03cbbSTheodore Ts'o if (!err) 4677908c7f19STejun Heo err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 4678908c7f19STejun Heo GFP_KERNEL); 4679c8585c6fSDaeho Jeong if (!err) 4680bbd55937SEric Biggers err = percpu_init_rwsem(&sbi->s_writepages_rwsem); 4681c8585c6fSDaeho Jeong 4682d5e03cbbSTheodore Ts'o if (err) { 4683d5e03cbbSTheodore Ts'o ext4_msg(sb, KERN_ERR, "insufficient memory"); 4684d5e03cbbSTheodore Ts'o goto failed_mount6; 4685d5e03cbbSTheodore Ts'o } 4686d5e03cbbSTheodore Ts'o 4687e2b911c5SDarrick J. Wong if (ext4_has_feature_flex_bg(sb)) 4688d5e03cbbSTheodore Ts'o if (!ext4_fill_flex_info(sb)) { 4689d5e03cbbSTheodore Ts'o ext4_msg(sb, KERN_ERR, 4690d5e03cbbSTheodore Ts'o "unable to initialize " 4691d5e03cbbSTheodore Ts'o "flex_bg meta info!"); 4692d5e03cbbSTheodore Ts'o goto failed_mount6; 4693d5e03cbbSTheodore Ts'o } 4694d5e03cbbSTheodore Ts'o 4695bfff6873SLukas Czerner err = ext4_register_li_request(sb, first_not_zeroed); 4696bfff6873SLukas Czerner if (err) 4697dcf2d804STao Ma goto failed_mount6; 4698bfff6873SLukas Czerner 4699b5799018STheodore Ts'o err = ext4_register_sysfs(sb); 4700dcf2d804STao Ma if (err) 4701dcf2d804STao Ma goto failed_mount7; 47023197ebdbSTheodore Ts'o 47039b2ff357SJan Kara #ifdef CONFIG_QUOTA 47049b2ff357SJan Kara /* Enable quota usage during mount. */ 4705bc98a42cSDavid Howells if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 47069b2ff357SJan Kara err = ext4_enable_quotas(sb); 47079b2ff357SJan Kara if (err) 47089b2ff357SJan Kara goto failed_mount8; 47099b2ff357SJan Kara } 47109b2ff357SJan Kara #endif /* CONFIG_QUOTA */ 47119b2ff357SJan Kara 4712617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 4713617ba13bSMingming Cao ext4_orphan_cleanup(sb, es); 4714617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 47150390131bSFrank Mayhar if (needs_recovery) { 4716b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "recovery complete"); 4717617ba13bSMingming Cao ext4_mark_recovery_complete(sb, es); 47180390131bSFrank Mayhar } 47190390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal) { 47200390131bSFrank Mayhar if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 47210390131bSFrank Mayhar descr = " journalled data mode"; 47220390131bSFrank Mayhar else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 47230390131bSFrank Mayhar descr = " ordered data mode"; 47240390131bSFrank Mayhar else 47250390131bSFrank Mayhar descr = " writeback data mode"; 47260390131bSFrank Mayhar } else 47270390131bSFrank Mayhar descr = "out journal"; 47280390131bSFrank Mayhar 472979add3a3SLukas Czerner if (test_opt(sb, DISCARD)) { 473079add3a3SLukas Czerner struct request_queue *q = bdev_get_queue(sb->s_bdev); 473179add3a3SLukas Czerner if (!blk_queue_discard(q)) 473279add3a3SLukas Czerner ext4_msg(sb, KERN_WARNING, 473379add3a3SLukas Czerner "mounting with \"discard\" option, but " 473479add3a3SLukas Czerner "the device does not support discard"); 473579add3a3SLukas Czerner } 473679add3a3SLukas Czerner 4737e294a537STheodore Ts'o if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 4738d4c402d9SCurt Wohlgemuth ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 47395aee0f8aSTheodore Ts'o "Opts: %.*s%s%s", descr, 47405aee0f8aSTheodore Ts'o (int) sizeof(sbi->s_es->s_mount_opts), 47415aee0f8aSTheodore Ts'o sbi->s_es->s_mount_opts, 47428b67f04aSTheodore Ts'o *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 4743ac27a0ecSDave Kleikamp 474466e61a9eSTheodore Ts'o if (es->s_error_count) 474566e61a9eSTheodore Ts'o mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 4746ac27a0ecSDave Kleikamp 4747efbed4dcSTheodore Ts'o /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 4748efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 4749efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 4750efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 4751efbed4dcSTheodore Ts'o 4752d4c402d9SCurt Wohlgemuth kfree(orig_data); 4753ac27a0ecSDave Kleikamp return 0; 4754ac27a0ecSDave Kleikamp 4755617ba13bSMingming Cao cantfind_ext4: 4756ac27a0ecSDave Kleikamp if (!silent) 4757b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 4758ac27a0ecSDave Kleikamp goto failed_mount; 4759ac27a0ecSDave Kleikamp 476072ba7450STheodore Ts'o #ifdef CONFIG_QUOTA 476172ba7450STheodore Ts'o failed_mount8: 4762ebd173beSTheodore Ts'o ext4_unregister_sysfs(sb); 476372ba7450STheodore Ts'o #endif 4764dcf2d804STao Ma failed_mount7: 4765dcf2d804STao Ma ext4_unregister_li_request(sb); 4766dcf2d804STao Ma failed_mount6: 4767f9ae9cf5STheodore Ts'o ext4_mb_release(sb); 47687c990728SSuraj Jitindar Singh rcu_read_lock(); 47697c990728SSuraj Jitindar Singh flex_groups = rcu_dereference(sbi->s_flex_groups); 47707c990728SSuraj Jitindar Singh if (flex_groups) { 47717c990728SSuraj Jitindar Singh for (i = 0; i < sbi->s_flex_groups_allocated; i++) 47727c990728SSuraj Jitindar Singh kvfree(flex_groups[i]); 47737c990728SSuraj Jitindar Singh kvfree(flex_groups); 47747c990728SSuraj Jitindar Singh } 47757c990728SSuraj Jitindar Singh rcu_read_unlock(); 4776d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_freeclusters_counter); 4777d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_freeinodes_counter); 4778d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_dirs_counter); 4779d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 4780bbd55937SEric Biggers percpu_free_rwsem(&sbi->s_writepages_rwsem); 478100764937SAzat Khuzhin failed_mount5: 4782f9ae9cf5STheodore Ts'o ext4_ext_release(sb); 4783f9ae9cf5STheodore Ts'o ext4_release_system_zone(sb); 4784f9ae9cf5STheodore Ts'o failed_mount4a: 478594bf608aSAl Viro dput(sb->s_root); 478632a9bb57SManish Katiyar sb->s_root = NULL; 478794bf608aSAl Viro failed_mount4: 4788b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "mount failed"); 47892e8fa54eSJan Kara if (EXT4_SB(sb)->rsv_conversion_wq) 47902e8fa54eSJan Kara destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 47914c0425ffSMingming Cao failed_mount_wq: 4792dec214d0STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 4793dec214d0STahsin Erdogan sbi->s_ea_inode_cache = NULL; 479450c15df6SChengguang Xu 479547387409STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 479647387409STahsin Erdogan sbi->s_ea_block_cache = NULL; 479750c15df6SChengguang Xu 47980390131bSFrank Mayhar if (sbi->s_journal) { 4799dab291afSMingming Cao jbd2_journal_destroy(sbi->s_journal); 480047b4a50bSJan Kara sbi->s_journal = NULL; 48010390131bSFrank Mayhar } 480250460fe8SDarrick J. Wong failed_mount3a: 4803d3922a77SZheng Liu ext4_es_unregister_shrinker(sbi); 4804eb68d0e2SZheng Liu failed_mount3: 48059105bb14SAl Viro del_timer_sync(&sbi->s_err_report); 4806c5e06d10SJohann Lombardi if (sbi->s_mmp_tsk) 4807c5e06d10SJohann Lombardi kthread_stop(sbi->s_mmp_tsk); 4808ac27a0ecSDave Kleikamp failed_mount2: 48091d0c3924STheodore Ts'o rcu_read_lock(); 48101d0c3924STheodore Ts'o group_desc = rcu_dereference(sbi->s_group_desc); 4811ac27a0ecSDave Kleikamp for (i = 0; i < db_count; i++) 48121d0c3924STheodore Ts'o brelse(group_desc[i]); 48131d0c3924STheodore Ts'o kvfree(group_desc); 48141d0c3924STheodore Ts'o rcu_read_unlock(); 4815ac27a0ecSDave Kleikamp failed_mount: 48160441984aSDarrick J. Wong if (sbi->s_chksum_driver) 48170441984aSDarrick J. Wong crypto_free_shash(sbi->s_chksum_driver); 4818c83ad55eSGabriel Krisman Bertazi 4819c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 4820c83ad55eSGabriel Krisman Bertazi utf8_unload(sbi->s_encoding); 4821c83ad55eSGabriel Krisman Bertazi #endif 4822c83ad55eSGabriel Krisman Bertazi 4823ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4824a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 48250ba33facSTheodore Ts'o kfree(get_qf_name(sb, sbi, i)); 4826ac27a0ecSDave Kleikamp #endif 4827617ba13bSMingming Cao ext4_blkdev_remove(sbi); 4828ac27a0ecSDave Kleikamp brelse(bh); 4829ac27a0ecSDave Kleikamp out_fail: 4830ac27a0ecSDave Kleikamp sb->s_fs_info = NULL; 4831f6830165SManish Katiyar kfree(sbi->s_blockgroup_lock); 48325aee0f8aSTheodore Ts'o out_free_base: 4833ac27a0ecSDave Kleikamp kfree(sbi); 4834d4c402d9SCurt Wohlgemuth kfree(orig_data); 48355e405595SDan Williams fs_put_dax(dax_dev); 483607aa2ea1SLukas Czerner return err ? err : ret; 4837ac27a0ecSDave Kleikamp } 4838ac27a0ecSDave Kleikamp 4839ac27a0ecSDave Kleikamp /* 4840ac27a0ecSDave Kleikamp * Setup any per-fs journal parameters now. We'll do this both on 4841ac27a0ecSDave Kleikamp * initial mount, once the journal has been initialised but before we've 4842ac27a0ecSDave Kleikamp * done any recovery; and again on any subsequent remount. 4843ac27a0ecSDave Kleikamp */ 4844617ba13bSMingming Cao static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 4845ac27a0ecSDave Kleikamp { 4846617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 4847ac27a0ecSDave Kleikamp 4848ac27a0ecSDave Kleikamp journal->j_commit_interval = sbi->s_commit_interval; 484930773840STheodore Ts'o journal->j_min_batch_time = sbi->s_min_batch_time; 485030773840STheodore Ts'o journal->j_max_batch_time = sbi->s_max_batch_time; 4851ac27a0ecSDave Kleikamp 4852a931da6aSTheodore Ts'o write_lock(&journal->j_state_lock); 4853ac27a0ecSDave Kleikamp if (test_opt(sb, BARRIER)) 4854dab291afSMingming Cao journal->j_flags |= JBD2_BARRIER; 4855ac27a0ecSDave Kleikamp else 4856dab291afSMingming Cao journal->j_flags &= ~JBD2_BARRIER; 48575bf5683aSHidehiro Kawai if (test_opt(sb, DATA_ERR_ABORT)) 48585bf5683aSHidehiro Kawai journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 48595bf5683aSHidehiro Kawai else 48605bf5683aSHidehiro Kawai journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 4861a931da6aSTheodore Ts'o write_unlock(&journal->j_state_lock); 4862ac27a0ecSDave Kleikamp } 4863ac27a0ecSDave Kleikamp 4864c6cb7e77SEric Whitney static struct inode *ext4_get_journal_inode(struct super_block *sb, 4865ac27a0ecSDave Kleikamp unsigned int journal_inum) 4866ac27a0ecSDave Kleikamp { 4867ac27a0ecSDave Kleikamp struct inode *journal_inode; 4868ac27a0ecSDave Kleikamp 4869c6cb7e77SEric Whitney /* 4870c6cb7e77SEric Whitney * Test for the existence of a valid inode on disk. Bad things 4871c6cb7e77SEric Whitney * happen if we iget() an unused inode, as the subsequent iput() 4872c6cb7e77SEric Whitney * will try to delete it. 4873c6cb7e77SEric Whitney */ 48748a363970STheodore Ts'o journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 48751d1fe1eeSDavid Howells if (IS_ERR(journal_inode)) { 4876b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "no journal found"); 4877ac27a0ecSDave Kleikamp return NULL; 4878ac27a0ecSDave Kleikamp } 4879ac27a0ecSDave Kleikamp if (!journal_inode->i_nlink) { 4880ac27a0ecSDave Kleikamp make_bad_inode(journal_inode); 4881ac27a0ecSDave Kleikamp iput(journal_inode); 4882b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 4883ac27a0ecSDave Kleikamp return NULL; 4884ac27a0ecSDave Kleikamp } 4885ac27a0ecSDave Kleikamp 4886e5f8eab8STheodore Ts'o jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 4887ac27a0ecSDave Kleikamp journal_inode, journal_inode->i_size); 48881d1fe1eeSDavid Howells if (!S_ISREG(journal_inode->i_mode)) { 4889b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "invalid journal inode"); 4890ac27a0ecSDave Kleikamp iput(journal_inode); 4891ac27a0ecSDave Kleikamp return NULL; 4892ac27a0ecSDave Kleikamp } 4893c6cb7e77SEric Whitney return journal_inode; 4894c6cb7e77SEric Whitney } 4895c6cb7e77SEric Whitney 4896c6cb7e77SEric Whitney static journal_t *ext4_get_journal(struct super_block *sb, 4897c6cb7e77SEric Whitney unsigned int journal_inum) 4898c6cb7e77SEric Whitney { 4899c6cb7e77SEric Whitney struct inode *journal_inode; 4900c6cb7e77SEric Whitney journal_t *journal; 4901c6cb7e77SEric Whitney 4902c6cb7e77SEric Whitney BUG_ON(!ext4_has_feature_journal(sb)); 4903c6cb7e77SEric Whitney 4904c6cb7e77SEric Whitney journal_inode = ext4_get_journal_inode(sb, journal_inum); 4905c6cb7e77SEric Whitney if (!journal_inode) 4906c6cb7e77SEric Whitney return NULL; 4907ac27a0ecSDave Kleikamp 4908dab291afSMingming Cao journal = jbd2_journal_init_inode(journal_inode); 4909ac27a0ecSDave Kleikamp if (!journal) { 4910b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 4911ac27a0ecSDave Kleikamp iput(journal_inode); 4912ac27a0ecSDave Kleikamp return NULL; 4913ac27a0ecSDave Kleikamp } 4914ac27a0ecSDave Kleikamp journal->j_private = sb; 4915617ba13bSMingming Cao ext4_init_journal_params(sb, journal); 4916ac27a0ecSDave Kleikamp return journal; 4917ac27a0ecSDave Kleikamp } 4918ac27a0ecSDave Kleikamp 4919617ba13bSMingming Cao static journal_t *ext4_get_dev_journal(struct super_block *sb, 4920ac27a0ecSDave Kleikamp dev_t j_dev) 4921ac27a0ecSDave Kleikamp { 4922ac27a0ecSDave Kleikamp struct buffer_head *bh; 4923ac27a0ecSDave Kleikamp journal_t *journal; 4924617ba13bSMingming Cao ext4_fsblk_t start; 4925617ba13bSMingming Cao ext4_fsblk_t len; 4926ac27a0ecSDave Kleikamp int hblock, blocksize; 4927617ba13bSMingming Cao ext4_fsblk_t sb_block; 4928ac27a0ecSDave Kleikamp unsigned long offset; 4929617ba13bSMingming Cao struct ext4_super_block *es; 4930ac27a0ecSDave Kleikamp struct block_device *bdev; 4931ac27a0ecSDave Kleikamp 4932e2b911c5SDarrick J. Wong BUG_ON(!ext4_has_feature_journal(sb)); 49330390131bSFrank Mayhar 4934b31e1552SEric Sandeen bdev = ext4_blkdev_get(j_dev, sb); 4935ac27a0ecSDave Kleikamp if (bdev == NULL) 4936ac27a0ecSDave Kleikamp return NULL; 4937ac27a0ecSDave Kleikamp 4938ac27a0ecSDave Kleikamp blocksize = sb->s_blocksize; 4939e1defc4fSMartin K. Petersen hblock = bdev_logical_block_size(bdev); 4940ac27a0ecSDave Kleikamp if (blocksize < hblock) { 4941b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4942b31e1552SEric Sandeen "blocksize too small for journal device"); 4943ac27a0ecSDave Kleikamp goto out_bdev; 4944ac27a0ecSDave Kleikamp } 4945ac27a0ecSDave Kleikamp 4946617ba13bSMingming Cao sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 4947617ba13bSMingming Cao offset = EXT4_MIN_BLOCK_SIZE % blocksize; 4948ac27a0ecSDave Kleikamp set_blocksize(bdev, blocksize); 4949ac27a0ecSDave Kleikamp if (!(bh = __bread(bdev, sb_block, blocksize))) { 4950b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 4951b31e1552SEric Sandeen "external journal"); 4952ac27a0ecSDave Kleikamp goto out_bdev; 4953ac27a0ecSDave Kleikamp } 4954ac27a0ecSDave Kleikamp 49552716b802STheodore Ts'o es = (struct ext4_super_block *) (bh->b_data + offset); 4956617ba13bSMingming Cao if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 4957ac27a0ecSDave Kleikamp !(le32_to_cpu(es->s_feature_incompat) & 4958617ba13bSMingming Cao EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 4959b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "external journal has " 4960b31e1552SEric Sandeen "bad superblock"); 4961ac27a0ecSDave Kleikamp brelse(bh); 4962ac27a0ecSDave Kleikamp goto out_bdev; 4963ac27a0ecSDave Kleikamp } 4964ac27a0ecSDave Kleikamp 4965df4763beSDarrick J. Wong if ((le32_to_cpu(es->s_feature_ro_compat) & 4966df4763beSDarrick J. Wong EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 4967df4763beSDarrick J. Wong es->s_checksum != ext4_superblock_csum(sb, es)) { 4968df4763beSDarrick J. Wong ext4_msg(sb, KERN_ERR, "external journal has " 4969df4763beSDarrick J. Wong "corrupt superblock"); 4970df4763beSDarrick J. Wong brelse(bh); 4971df4763beSDarrick J. Wong goto out_bdev; 4972df4763beSDarrick J. Wong } 4973df4763beSDarrick J. Wong 4974617ba13bSMingming Cao if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 4975b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 4976ac27a0ecSDave Kleikamp brelse(bh); 4977ac27a0ecSDave Kleikamp goto out_bdev; 4978ac27a0ecSDave Kleikamp } 4979ac27a0ecSDave Kleikamp 4980bd81d8eeSLaurent Vivier len = ext4_blocks_count(es); 4981ac27a0ecSDave Kleikamp start = sb_block + 1; 4982ac27a0ecSDave Kleikamp brelse(bh); /* we're done with the superblock */ 4983ac27a0ecSDave Kleikamp 4984dab291afSMingming Cao journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 4985ac27a0ecSDave Kleikamp start, len, blocksize); 4986ac27a0ecSDave Kleikamp if (!journal) { 4987b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "failed to create device journal"); 4988ac27a0ecSDave Kleikamp goto out_bdev; 4989ac27a0ecSDave Kleikamp } 4990ac27a0ecSDave Kleikamp journal->j_private = sb; 4991dfec8a14SMike Christie ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer); 4992ac27a0ecSDave Kleikamp wait_on_buffer(journal->j_sb_buffer); 4993ac27a0ecSDave Kleikamp if (!buffer_uptodate(journal->j_sb_buffer)) { 4994b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 4995ac27a0ecSDave Kleikamp goto out_journal; 4996ac27a0ecSDave Kleikamp } 4997ac27a0ecSDave Kleikamp if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 4998b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "External journal has more than one " 4999b31e1552SEric Sandeen "user (unsupported) - %d", 5000ac27a0ecSDave Kleikamp be32_to_cpu(journal->j_superblock->s_nr_users)); 5001ac27a0ecSDave Kleikamp goto out_journal; 5002ac27a0ecSDave Kleikamp } 5003617ba13bSMingming Cao EXT4_SB(sb)->journal_bdev = bdev; 5004617ba13bSMingming Cao ext4_init_journal_params(sb, journal); 5005ac27a0ecSDave Kleikamp return journal; 50060b8e58a1SAndreas Dilger 5007ac27a0ecSDave Kleikamp out_journal: 5008dab291afSMingming Cao jbd2_journal_destroy(journal); 5009ac27a0ecSDave Kleikamp out_bdev: 5010617ba13bSMingming Cao ext4_blkdev_put(bdev); 5011ac27a0ecSDave Kleikamp return NULL; 5012ac27a0ecSDave Kleikamp } 5013ac27a0ecSDave Kleikamp 5014617ba13bSMingming Cao static int ext4_load_journal(struct super_block *sb, 5015617ba13bSMingming Cao struct ext4_super_block *es, 5016ac27a0ecSDave Kleikamp unsigned long journal_devnum) 5017ac27a0ecSDave Kleikamp { 5018ac27a0ecSDave Kleikamp journal_t *journal; 5019ac27a0ecSDave Kleikamp unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 5020ac27a0ecSDave Kleikamp dev_t journal_dev; 5021ac27a0ecSDave Kleikamp int err = 0; 5022ac27a0ecSDave Kleikamp int really_read_only; 5023ac27a0ecSDave Kleikamp 5024e2b911c5SDarrick J. Wong BUG_ON(!ext4_has_feature_journal(sb)); 50250390131bSFrank Mayhar 5026ac27a0ecSDave Kleikamp if (journal_devnum && 5027ac27a0ecSDave Kleikamp journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5028b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "external journal device major/minor " 5029b31e1552SEric Sandeen "numbers have changed"); 5030ac27a0ecSDave Kleikamp journal_dev = new_decode_dev(journal_devnum); 5031ac27a0ecSDave Kleikamp } else 5032ac27a0ecSDave Kleikamp journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 5033ac27a0ecSDave Kleikamp 5034ac27a0ecSDave Kleikamp really_read_only = bdev_read_only(sb->s_bdev); 5035ac27a0ecSDave Kleikamp 5036ac27a0ecSDave Kleikamp /* 5037ac27a0ecSDave Kleikamp * Are we loading a blank journal or performing recovery after a 5038ac27a0ecSDave Kleikamp * crash? For recovery, we need to check in advance whether we 5039ac27a0ecSDave Kleikamp * can get read-write access to the device. 5040ac27a0ecSDave Kleikamp */ 5041e2b911c5SDarrick J. Wong if (ext4_has_feature_journal_needs_recovery(sb)) { 5042bc98a42cSDavid Howells if (sb_rdonly(sb)) { 5043b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "INFO: recovery " 5044b31e1552SEric Sandeen "required on readonly filesystem"); 5045ac27a0ecSDave Kleikamp if (really_read_only) { 5046b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "write access " 5047d98bf8cdSSimon Ruderich "unavailable, cannot proceed " 5048d98bf8cdSSimon Ruderich "(try mounting with noload)"); 5049ac27a0ecSDave Kleikamp return -EROFS; 5050ac27a0ecSDave Kleikamp } 5051b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "write access will " 5052b31e1552SEric Sandeen "be enabled during recovery"); 5053ac27a0ecSDave Kleikamp } 5054ac27a0ecSDave Kleikamp } 5055ac27a0ecSDave Kleikamp 5056ac27a0ecSDave Kleikamp if (journal_inum && journal_dev) { 5057b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "filesystem has both journal " 5058b31e1552SEric Sandeen "and inode journals!"); 5059ac27a0ecSDave Kleikamp return -EINVAL; 5060ac27a0ecSDave Kleikamp } 5061ac27a0ecSDave Kleikamp 5062ac27a0ecSDave Kleikamp if (journal_inum) { 5063617ba13bSMingming Cao if (!(journal = ext4_get_journal(sb, journal_inum))) 5064ac27a0ecSDave Kleikamp return -EINVAL; 5065ac27a0ecSDave Kleikamp } else { 5066617ba13bSMingming Cao if (!(journal = ext4_get_dev_journal(sb, journal_dev))) 5067ac27a0ecSDave Kleikamp return -EINVAL; 5068ac27a0ecSDave Kleikamp } 5069ac27a0ecSDave Kleikamp 507090576c0bSTheodore Ts'o if (!(journal->j_flags & JBD2_BARRIER)) 5071b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "barriers disabled"); 50724776004fSTheodore Ts'o 5073e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal_needs_recovery(sb)) 5074dab291afSMingming Cao err = jbd2_journal_wipe(journal, !really_read_only); 50751c13d5c0STheodore Ts'o if (!err) { 50761c13d5c0STheodore Ts'o char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 50771c13d5c0STheodore Ts'o if (save) 50781c13d5c0STheodore Ts'o memcpy(save, ((char *) es) + 50791c13d5c0STheodore Ts'o EXT4_S_ERR_START, EXT4_S_ERR_LEN); 5080dab291afSMingming Cao err = jbd2_journal_load(journal); 50811c13d5c0STheodore Ts'o if (save) 50821c13d5c0STheodore Ts'o memcpy(((char *) es) + EXT4_S_ERR_START, 50831c13d5c0STheodore Ts'o save, EXT4_S_ERR_LEN); 50841c13d5c0STheodore Ts'o kfree(save); 50851c13d5c0STheodore Ts'o } 5086ac27a0ecSDave Kleikamp 5087ac27a0ecSDave Kleikamp if (err) { 5088b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "error loading journal"); 5089dab291afSMingming Cao jbd2_journal_destroy(journal); 5090ac27a0ecSDave Kleikamp return err; 5091ac27a0ecSDave Kleikamp } 5092ac27a0ecSDave Kleikamp 5093617ba13bSMingming Cao EXT4_SB(sb)->s_journal = journal; 5094617ba13bSMingming Cao ext4_clear_journal_err(sb, es); 5095ac27a0ecSDave Kleikamp 5096c41303ceSMaciej Żenczykowski if (!really_read_only && journal_devnum && 5097ac27a0ecSDave Kleikamp journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5098ac27a0ecSDave Kleikamp es->s_journal_dev = cpu_to_le32(journal_devnum); 5099ac27a0ecSDave Kleikamp 5100ac27a0ecSDave Kleikamp /* Make sure we flush the recovery flag to disk. */ 5101e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 5102ac27a0ecSDave Kleikamp } 5103ac27a0ecSDave Kleikamp 5104ac27a0ecSDave Kleikamp return 0; 5105ac27a0ecSDave Kleikamp } 5106ac27a0ecSDave Kleikamp 5107e2d67052STheodore Ts'o static int ext4_commit_super(struct super_block *sb, int sync) 5108ac27a0ecSDave Kleikamp { 5109e2d67052STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 5110617ba13bSMingming Cao struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 5111c4be0c1dSTakashi Sato int error = 0; 5112ac27a0ecSDave Kleikamp 5113bdfe0cbdSTheodore Ts'o if (!sbh || block_device_ejected(sb)) 5114c4be0c1dSTakashi Sato return error; 5115a17712c8SJon Derrick 5116a17712c8SJon Derrick /* 5117a17712c8SJon Derrick * The superblock bh should be mapped, but it might not be if the 5118a17712c8SJon Derrick * device was hot-removed. Not much we can do but fail the I/O. 5119a17712c8SJon Derrick */ 5120a17712c8SJon Derrick if (!buffer_mapped(sbh)) 5121a17712c8SJon Derrick return error; 5122a17712c8SJon Derrick 512371290b36STheodore Ts'o /* 512471290b36STheodore Ts'o * If the file system is mounted read-only, don't update the 512571290b36STheodore Ts'o * superblock write time. This avoids updating the superblock 512671290b36STheodore Ts'o * write time when we are mounting the root file system 512771290b36STheodore Ts'o * read/only but we need to replay the journal; at that point, 512871290b36STheodore Ts'o * for people who are east of GMT and who make their clock 512971290b36STheodore Ts'o * tick in localtime for Windows bug-for-bug compatibility, 513071290b36STheodore Ts'o * the clock is set in the future, and this will cause e2fsck 513171290b36STheodore Ts'o * to complain and force a full file system check. 513271290b36STheodore Ts'o */ 51331751e8a6SLinus Torvalds if (!(sb->s_flags & SB_RDONLY)) 51346a0678a7SArnd Bergmann ext4_update_tstamp(es, s_wtime); 5135f613dfcbSTheodore Ts'o if (sb->s_bdev->bd_part) 5136afc32f7eSTheodore Ts'o es->s_kbytes_written = 5137afc32f7eSTheodore Ts'o cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 5138dbae2c55SMichael Callahan ((part_stat_read(sb->s_bdev->bd_part, 5139dbae2c55SMichael Callahan sectors[STAT_WRITE]) - 5140afc32f7eSTheodore Ts'o EXT4_SB(sb)->s_sectors_written_start) >> 1)); 5141f613dfcbSTheodore Ts'o else 5142f613dfcbSTheodore Ts'o es->s_kbytes_written = 5143f613dfcbSTheodore Ts'o cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 5144d5e03cbbSTheodore Ts'o if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter)) 514557042651STheodore Ts'o ext4_free_blocks_count_set(es, 514657042651STheodore Ts'o EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 514757042651STheodore Ts'o &EXT4_SB(sb)->s_freeclusters_counter))); 5148d5e03cbbSTheodore Ts'o if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter)) 51497f93cff9STheodore Ts'o es->s_free_inodes_count = 51507f93cff9STheodore Ts'o cpu_to_le32(percpu_counter_sum_positive( 51515d1b1b3fSAneesh Kumar K.V &EXT4_SB(sb)->s_freeinodes_counter)); 5152ac27a0ecSDave Kleikamp BUFFER_TRACE(sbh, "marking dirty"); 515306db49e6STheodore Ts'o ext4_superblock_csum_set(sb); 51541566a48aSTheodore Ts'o if (sync) 51554743f839SPranay Kr. Srivastava lock_buffer(sbh); 5156e8680786STheodore Ts'o if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 51574743f839SPranay Kr. Srivastava /* 51584743f839SPranay Kr. Srivastava * Oh, dear. A previous attempt to write the 51594743f839SPranay Kr. Srivastava * superblock failed. This could happen because the 51604743f839SPranay Kr. Srivastava * USB device was yanked out. Or it could happen to 51614743f839SPranay Kr. Srivastava * be a transient write error and maybe the block will 51624743f839SPranay Kr. Srivastava * be remapped. Nothing we can do but to retry the 51634743f839SPranay Kr. Srivastava * write and hope for the best. 51644743f839SPranay Kr. Srivastava */ 51654743f839SPranay Kr. Srivastava ext4_msg(sb, KERN_ERR, "previous I/O error to " 51664743f839SPranay Kr. Srivastava "superblock detected"); 51674743f839SPranay Kr. Srivastava clear_buffer_write_io_error(sbh); 51684743f839SPranay Kr. Srivastava set_buffer_uptodate(sbh); 51694743f839SPranay Kr. Srivastava } 5170ac27a0ecSDave Kleikamp mark_buffer_dirty(sbh); 5171914258bfSTheodore Ts'o if (sync) { 51721566a48aSTheodore Ts'o unlock_buffer(sbh); 5173564bc402SDaeho Jeong error = __sync_dirty_buffer(sbh, 517400473374SJan Kara REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0)); 5175c89128a0SJaegeuk Kim if (buffer_write_io_error(sbh)) { 5176b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "I/O error while writing " 5177b31e1552SEric Sandeen "superblock"); 5178914258bfSTheodore Ts'o clear_buffer_write_io_error(sbh); 5179914258bfSTheodore Ts'o set_buffer_uptodate(sbh); 5180914258bfSTheodore Ts'o } 5181914258bfSTheodore Ts'o } 5182c4be0c1dSTakashi Sato return error; 5183ac27a0ecSDave Kleikamp } 5184ac27a0ecSDave Kleikamp 5185ac27a0ecSDave Kleikamp /* 5186ac27a0ecSDave Kleikamp * Have we just finished recovery? If so, and if we are mounting (or 5187ac27a0ecSDave Kleikamp * remounting) the filesystem readonly, then we will end up with a 5188ac27a0ecSDave Kleikamp * consistent fs on disk. Record that fact. 5189ac27a0ecSDave Kleikamp */ 5190617ba13bSMingming Cao static void ext4_mark_recovery_complete(struct super_block *sb, 5191617ba13bSMingming Cao struct ext4_super_block *es) 5192ac27a0ecSDave Kleikamp { 5193617ba13bSMingming Cao journal_t *journal = EXT4_SB(sb)->s_journal; 5194ac27a0ecSDave Kleikamp 5195e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal(sb)) { 51960390131bSFrank Mayhar BUG_ON(journal != NULL); 51970390131bSFrank Mayhar return; 51980390131bSFrank Mayhar } 5199dab291afSMingming Cao jbd2_journal_lock_updates(journal); 52007ffe1ea8SHidehiro Kawai if (jbd2_journal_flush(journal) < 0) 52017ffe1ea8SHidehiro Kawai goto out; 52027ffe1ea8SHidehiro Kawai 5203bc98a42cSDavid Howells if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) { 5204e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 5205e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 5206ac27a0ecSDave Kleikamp } 52077ffe1ea8SHidehiro Kawai 52087ffe1ea8SHidehiro Kawai out: 5209dab291afSMingming Cao jbd2_journal_unlock_updates(journal); 5210ac27a0ecSDave Kleikamp } 5211ac27a0ecSDave Kleikamp 5212ac27a0ecSDave Kleikamp /* 5213ac27a0ecSDave Kleikamp * If we are mounting (or read-write remounting) a filesystem whose journal 5214ac27a0ecSDave Kleikamp * has recorded an error from a previous lifetime, move that error to the 5215ac27a0ecSDave Kleikamp * main filesystem now. 5216ac27a0ecSDave Kleikamp */ 5217617ba13bSMingming Cao static void ext4_clear_journal_err(struct super_block *sb, 5218617ba13bSMingming Cao struct ext4_super_block *es) 5219ac27a0ecSDave Kleikamp { 5220ac27a0ecSDave Kleikamp journal_t *journal; 5221ac27a0ecSDave Kleikamp int j_errno; 5222ac27a0ecSDave Kleikamp const char *errstr; 5223ac27a0ecSDave Kleikamp 5224e2b911c5SDarrick J. Wong BUG_ON(!ext4_has_feature_journal(sb)); 52250390131bSFrank Mayhar 5226617ba13bSMingming Cao journal = EXT4_SB(sb)->s_journal; 5227ac27a0ecSDave Kleikamp 5228ac27a0ecSDave Kleikamp /* 5229ac27a0ecSDave Kleikamp * Now check for any error status which may have been recorded in the 5230617ba13bSMingming Cao * journal by a prior ext4_error() or ext4_abort() 5231ac27a0ecSDave Kleikamp */ 5232ac27a0ecSDave Kleikamp 5233dab291afSMingming Cao j_errno = jbd2_journal_errno(journal); 5234ac27a0ecSDave Kleikamp if (j_errno) { 5235ac27a0ecSDave Kleikamp char nbuf[16]; 5236ac27a0ecSDave Kleikamp 5237617ba13bSMingming Cao errstr = ext4_decode_error(sb, j_errno, nbuf); 523812062dddSEric Sandeen ext4_warning(sb, "Filesystem error recorded " 5239ac27a0ecSDave Kleikamp "from previous mount: %s", errstr); 524012062dddSEric Sandeen ext4_warning(sb, "Marking fs in need of filesystem check."); 5241ac27a0ecSDave Kleikamp 5242617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 5243617ba13bSMingming Cao es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 5244e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 5245ac27a0ecSDave Kleikamp 5246dab291afSMingming Cao jbd2_journal_clear_err(journal); 5247d796c52eSTheodore Ts'o jbd2_journal_update_sb_errno(journal); 5248ac27a0ecSDave Kleikamp } 5249ac27a0ecSDave Kleikamp } 5250ac27a0ecSDave Kleikamp 5251ac27a0ecSDave Kleikamp /* 5252ac27a0ecSDave Kleikamp * Force the running and committing transactions to commit, 5253ac27a0ecSDave Kleikamp * and wait on the commit. 5254ac27a0ecSDave Kleikamp */ 5255617ba13bSMingming Cao int ext4_force_commit(struct super_block *sb) 5256ac27a0ecSDave Kleikamp { 5257ac27a0ecSDave Kleikamp journal_t *journal; 5258ac27a0ecSDave Kleikamp 5259bc98a42cSDavid Howells if (sb_rdonly(sb)) 5260ac27a0ecSDave Kleikamp return 0; 5261ac27a0ecSDave Kleikamp 5262617ba13bSMingming Cao journal = EXT4_SB(sb)->s_journal; 5263b1deefc9SGuo Chao return ext4_journal_force_commit(journal); 5264ac27a0ecSDave Kleikamp } 5265ac27a0ecSDave Kleikamp 5266617ba13bSMingming Cao static int ext4_sync_fs(struct super_block *sb, int wait) 5267ac27a0ecSDave Kleikamp { 526814ce0cb4STheodore Ts'o int ret = 0; 52699eddacf9SJan Kara tid_t target; 527006a407f1SDmitry Monakhov bool needs_barrier = false; 52718d5d02e6SMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 5272ac27a0ecSDave Kleikamp 527349598e04SJun Piao if (unlikely(ext4_forced_shutdown(sbi))) 52740db1ff22STheodore Ts'o return 0; 52750db1ff22STheodore Ts'o 52769bffad1eSTheodore Ts'o trace_ext4_sync_fs(sb, wait); 52772e8fa54eSJan Kara flush_workqueue(sbi->rsv_conversion_wq); 5278a1177825SJan Kara /* 5279a1177825SJan Kara * Writeback quota in non-journalled quota case - journalled quota has 5280a1177825SJan Kara * no dirty dquots 5281a1177825SJan Kara */ 5282a1177825SJan Kara dquot_writeback_dquots(sb, -1); 528306a407f1SDmitry Monakhov /* 528406a407f1SDmitry Monakhov * Data writeback is possible w/o journal transaction, so barrier must 528506a407f1SDmitry Monakhov * being sent at the end of the function. But we can skip it if 528606a407f1SDmitry Monakhov * transaction_commit will do it for us. 528706a407f1SDmitry Monakhov */ 5288bda32530STheodore Ts'o if (sbi->s_journal) { 528906a407f1SDmitry Monakhov target = jbd2_get_latest_transaction(sbi->s_journal); 529006a407f1SDmitry Monakhov if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 529106a407f1SDmitry Monakhov !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 529206a407f1SDmitry Monakhov needs_barrier = true; 529306a407f1SDmitry Monakhov 52948d5d02e6SMingming Cao if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 5295ac27a0ecSDave Kleikamp if (wait) 5296bda32530STheodore Ts'o ret = jbd2_log_wait_commit(sbi->s_journal, 5297bda32530STheodore Ts'o target); 52980390131bSFrank Mayhar } 5299bda32530STheodore Ts'o } else if (wait && test_opt(sb, BARRIER)) 5300bda32530STheodore Ts'o needs_barrier = true; 530106a407f1SDmitry Monakhov if (needs_barrier) { 530206a407f1SDmitry Monakhov int err; 530306a407f1SDmitry Monakhov err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); 530406a407f1SDmitry Monakhov if (!ret) 530506a407f1SDmitry Monakhov ret = err; 530606a407f1SDmitry Monakhov } 530706a407f1SDmitry Monakhov 530806a407f1SDmitry Monakhov return ret; 530906a407f1SDmitry Monakhov } 531006a407f1SDmitry Monakhov 5311ac27a0ecSDave Kleikamp /* 5312ac27a0ecSDave Kleikamp * LVM calls this function before a (read-only) snapshot is created. This 5313ac27a0ecSDave Kleikamp * gives us a chance to flush the journal completely and mark the fs clean. 5314be4f27d3SYongqiang Yang * 5315be4f27d3SYongqiang Yang * Note that only this function cannot bring a filesystem to be in a clean 53168e8ad8a5SJan Kara * state independently. It relies on upper layer to stop all data & metadata 53178e8ad8a5SJan Kara * modifications. 5318ac27a0ecSDave Kleikamp */ 5319c4be0c1dSTakashi Sato static int ext4_freeze(struct super_block *sb) 5320ac27a0ecSDave Kleikamp { 5321c4be0c1dSTakashi Sato int error = 0; 5322c4be0c1dSTakashi Sato journal_t *journal; 5323ac27a0ecSDave Kleikamp 5324bc98a42cSDavid Howells if (sb_rdonly(sb)) 53259ca92389STheodore Ts'o return 0; 53269ca92389STheodore Ts'o 5327c4be0c1dSTakashi Sato journal = EXT4_SB(sb)->s_journal; 5328ac27a0ecSDave Kleikamp 5329bb044576STheodore Ts'o if (journal) { 5330ac27a0ecSDave Kleikamp /* Now we set up the journal barrier. */ 5331dab291afSMingming Cao jbd2_journal_lock_updates(journal); 53327ffe1ea8SHidehiro Kawai 53337ffe1ea8SHidehiro Kawai /* 5334bb044576STheodore Ts'o * Don't clear the needs_recovery flag if we failed to 5335bb044576STheodore Ts'o * flush the journal. 53367ffe1ea8SHidehiro Kawai */ 5337c4be0c1dSTakashi Sato error = jbd2_journal_flush(journal); 53386b0310fbSEric Sandeen if (error < 0) 53396b0310fbSEric Sandeen goto out; 5340ac27a0ecSDave Kleikamp 5341ac27a0ecSDave Kleikamp /* Journal blocked and flushed, clear needs_recovery flag. */ 5342e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 5343c642dc9eSEric Sandeen } 5344c642dc9eSEric Sandeen 5345e2d67052STheodore Ts'o error = ext4_commit_super(sb, 1); 53466b0310fbSEric Sandeen out: 5347bb044576STheodore Ts'o if (journal) 53488e8ad8a5SJan Kara /* we rely on upper layer to stop further updates */ 5349bb044576STheodore Ts'o jbd2_journal_unlock_updates(journal); 53506b0310fbSEric Sandeen return error; 5351ac27a0ecSDave Kleikamp } 5352ac27a0ecSDave Kleikamp 5353ac27a0ecSDave Kleikamp /* 5354ac27a0ecSDave Kleikamp * Called by LVM after the snapshot is done. We need to reset the RECOVER 5355ac27a0ecSDave Kleikamp * flag here, even though the filesystem is not technically dirty yet. 5356ac27a0ecSDave Kleikamp */ 5357c4be0c1dSTakashi Sato static int ext4_unfreeze(struct super_block *sb) 5358ac27a0ecSDave Kleikamp { 5359bc98a42cSDavid Howells if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) 53609ca92389STheodore Ts'o return 0; 53619ca92389STheodore Ts'o 5362c642dc9eSEric Sandeen if (EXT4_SB(sb)->s_journal) { 53639ca92389STheodore Ts'o /* Reset the needs_recovery flag before the fs is unlocked. */ 5364e2b911c5SDarrick J. Wong ext4_set_feature_journal_needs_recovery(sb); 5365c642dc9eSEric Sandeen } 5366c642dc9eSEric Sandeen 5367e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 5368c4be0c1dSTakashi Sato return 0; 5369ac27a0ecSDave Kleikamp } 5370ac27a0ecSDave Kleikamp 5371673c6100STheodore Ts'o /* 5372673c6100STheodore Ts'o * Structure to save mount options for ext4_remount's benefit 5373673c6100STheodore Ts'o */ 5374673c6100STheodore Ts'o struct ext4_mount_options { 5375673c6100STheodore Ts'o unsigned long s_mount_opt; 5376a2595b8aSTheodore Ts'o unsigned long s_mount_opt2; 537708cefc7aSEric W. Biederman kuid_t s_resuid; 537808cefc7aSEric W. Biederman kgid_t s_resgid; 5379673c6100STheodore Ts'o unsigned long s_commit_interval; 5380673c6100STheodore Ts'o u32 s_min_batch_time, s_max_batch_time; 5381673c6100STheodore Ts'o #ifdef CONFIG_QUOTA 5382673c6100STheodore Ts'o int s_jquota_fmt; 5383a2d4a646SJan Kara char *s_qf_names[EXT4_MAXQUOTAS]; 5384673c6100STheodore Ts'o #endif 5385673c6100STheodore Ts'o }; 5386673c6100STheodore Ts'o 5387617ba13bSMingming Cao static int ext4_remount(struct super_block *sb, int *flags, char *data) 5388ac27a0ecSDave Kleikamp { 5389617ba13bSMingming Cao struct ext4_super_block *es; 5390617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 5391ac27a0ecSDave Kleikamp unsigned long old_sb_flags; 5392617ba13bSMingming Cao struct ext4_mount_options old_opts; 5393c79d967dSChristoph Hellwig int enable_quota = 0; 53948a266467STheodore Ts'o ext4_group_t g; 5395b3881f74STheodore Ts'o unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 5396c5e06d10SJohann Lombardi int err = 0; 5397ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 539803dafb5fSChen Gang int i, j; 539933458eabSTheodore Ts'o char *to_free[EXT4_MAXQUOTAS]; 5400ac27a0ecSDave Kleikamp #endif 5401d4c402d9SCurt Wohlgemuth char *orig_data = kstrdup(data, GFP_KERNEL); 5402ac27a0ecSDave Kleikamp 540321ac738eSChengguang Xu if (data && !orig_data) 540421ac738eSChengguang Xu return -ENOMEM; 540521ac738eSChengguang Xu 5406ac27a0ecSDave Kleikamp /* Store the original options */ 5407ac27a0ecSDave Kleikamp old_sb_flags = sb->s_flags; 5408ac27a0ecSDave Kleikamp old_opts.s_mount_opt = sbi->s_mount_opt; 5409a2595b8aSTheodore Ts'o old_opts.s_mount_opt2 = sbi->s_mount_opt2; 5410ac27a0ecSDave Kleikamp old_opts.s_resuid = sbi->s_resuid; 5411ac27a0ecSDave Kleikamp old_opts.s_resgid = sbi->s_resgid; 5412ac27a0ecSDave Kleikamp old_opts.s_commit_interval = sbi->s_commit_interval; 541330773840STheodore Ts'o old_opts.s_min_batch_time = sbi->s_min_batch_time; 541430773840STheodore Ts'o old_opts.s_max_batch_time = sbi->s_max_batch_time; 5415ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 5416ac27a0ecSDave Kleikamp old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 5417a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 541803dafb5fSChen Gang if (sbi->s_qf_names[i]) { 541933458eabSTheodore Ts'o char *qf_name = get_qf_name(sb, sbi, i); 542033458eabSTheodore Ts'o 542133458eabSTheodore Ts'o old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 542203dafb5fSChen Gang if (!old_opts.s_qf_names[i]) { 542303dafb5fSChen Gang for (j = 0; j < i; j++) 542403dafb5fSChen Gang kfree(old_opts.s_qf_names[j]); 54253e36a163SWei Yongjun kfree(orig_data); 542603dafb5fSChen Gang return -ENOMEM; 542703dafb5fSChen Gang } 542803dafb5fSChen Gang } else 542903dafb5fSChen Gang old_opts.s_qf_names[i] = NULL; 5430ac27a0ecSDave Kleikamp #endif 5431b3881f74STheodore Ts'o if (sbi->s_journal && sbi->s_journal->j_task->io_context) 5432b3881f74STheodore Ts'o journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 5433ac27a0ecSDave Kleikamp 5434661aa520SEric Sandeen if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 5435ac27a0ecSDave Kleikamp err = -EINVAL; 5436ac27a0ecSDave Kleikamp goto restore_opts; 5437ac27a0ecSDave Kleikamp } 5438ac27a0ecSDave Kleikamp 54396b992ff2SDarrick J. Wong if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 54406b992ff2SDarrick J. Wong test_opt(sb, JOURNAL_CHECKSUM)) { 54416b992ff2SDarrick J. Wong ext4_msg(sb, KERN_ERR, "changing journal_checksum " 54422d5b86e0SEric Sandeen "during remount not supported; ignoring"); 54432d5b86e0SEric Sandeen sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 5444c6d3d56dSDarrick J. Wong } 5445c6d3d56dSDarrick J. Wong 54466ae6514bSPiotr Sarna if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 54476ae6514bSPiotr Sarna if (test_opt2(sb, EXPLICIT_DELALLOC)) { 54486ae6514bSPiotr Sarna ext4_msg(sb, KERN_ERR, "can't mount with " 54496ae6514bSPiotr Sarna "both data=journal and delalloc"); 54506ae6514bSPiotr Sarna err = -EINVAL; 54516ae6514bSPiotr Sarna goto restore_opts; 54526ae6514bSPiotr Sarna } 54536ae6514bSPiotr Sarna if (test_opt(sb, DIOREAD_NOLOCK)) { 54546ae6514bSPiotr Sarna ext4_msg(sb, KERN_ERR, "can't mount with " 54556ae6514bSPiotr Sarna "both data=journal and dioread_nolock"); 54566ae6514bSPiotr Sarna err = -EINVAL; 54576ae6514bSPiotr Sarna goto restore_opts; 54586ae6514bSPiotr Sarna } 5459fc626fe3SIra Weiny if (test_opt(sb, DAX_ALWAYS)) { 5460923ae0ffSRoss Zwisler ext4_msg(sb, KERN_ERR, "can't mount with " 5461923ae0ffSRoss Zwisler "both data=journal and dax"); 5462923ae0ffSRoss Zwisler err = -EINVAL; 5463923ae0ffSRoss Zwisler goto restore_opts; 5464923ae0ffSRoss Zwisler } 5465ab04df78SJan Kara } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 5466ab04df78SJan Kara if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5467ab04df78SJan Kara ext4_msg(sb, KERN_ERR, "can't mount with " 5468ab04df78SJan Kara "journal_async_commit in data=ordered mode"); 5469ab04df78SJan Kara err = -EINVAL; 5470ab04df78SJan Kara goto restore_opts; 5471ab04df78SJan Kara } 5472923ae0ffSRoss Zwisler } 5473923ae0ffSRoss Zwisler 5474cdb7ee4cSTahsin Erdogan if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 5475cdb7ee4cSTahsin Erdogan ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 5476cdb7ee4cSTahsin Erdogan err = -EINVAL; 5477cdb7ee4cSTahsin Erdogan goto restore_opts; 5478cdb7ee4cSTahsin Erdogan } 5479cdb7ee4cSTahsin Erdogan 5480*9cb20f94SIra Weiny if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX_ALWAYS || 5481*9cb20f94SIra Weiny (sbi->s_mount_opt2 ^ old_opts.s_mount_opt2) & EXT4_MOUNT2_DAX_NEVER || 5482*9cb20f94SIra Weiny (sbi->s_mount_opt2 ^ old_opts.s_mount_opt2) & EXT4_MOUNT2_DAX_INODE) { 5483923ae0ffSRoss Zwisler ext4_msg(sb, KERN_WARNING, "warning: refusing change of " 5484*9cb20f94SIra Weiny "dax mount option with busy inodes while remounting"); 5485*9cb20f94SIra Weiny sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 5486*9cb20f94SIra Weiny sbi->s_mount_opt |= old_opts.s_mount_opt & EXT4_MOUNT_DAX_ALWAYS; 5487*9cb20f94SIra Weiny sbi->s_mount_opt2 &= ~(EXT4_MOUNT2_DAX_NEVER | EXT4_MOUNT2_DAX_INODE); 5488*9cb20f94SIra Weiny sbi->s_mount_opt2 |= old_opts.s_mount_opt2 & 5489*9cb20f94SIra Weiny (EXT4_MOUNT2_DAX_NEVER | EXT4_MOUNT2_DAX_INODE); 54906ae6514bSPiotr Sarna } 54916ae6514bSPiotr Sarna 54924ab2f15bSTheodore Ts'o if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 549354d3adbcSTheodore Ts'o ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user"); 5494ac27a0ecSDave Kleikamp 54951751e8a6SLinus Torvalds sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 54961751e8a6SLinus Torvalds (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 5497ac27a0ecSDave Kleikamp 5498ac27a0ecSDave Kleikamp es = sbi->s_es; 5499ac27a0ecSDave Kleikamp 5500b3881f74STheodore Ts'o if (sbi->s_journal) { 5501617ba13bSMingming Cao ext4_init_journal_params(sb, sbi->s_journal); 5502b3881f74STheodore Ts'o set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 5503b3881f74STheodore Ts'o } 5504ac27a0ecSDave Kleikamp 55051751e8a6SLinus Torvalds if (*flags & SB_LAZYTIME) 55061751e8a6SLinus Torvalds sb->s_flags |= SB_LAZYTIME; 5507a2fd66d0STheodore Ts'o 55081751e8a6SLinus Torvalds if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) { 55094ab2f15bSTheodore Ts'o if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 5510ac27a0ecSDave Kleikamp err = -EROFS; 5511ac27a0ecSDave Kleikamp goto restore_opts; 5512ac27a0ecSDave Kleikamp } 5513ac27a0ecSDave Kleikamp 55141751e8a6SLinus Torvalds if (*flags & SB_RDONLY) { 551538c03b34STheodore Ts'o err = sync_filesystem(sb); 551638c03b34STheodore Ts'o if (err < 0) 551738c03b34STheodore Ts'o goto restore_opts; 55180f0dd62fSChristoph Hellwig err = dquot_suspend(sb, -1); 55190f0dd62fSChristoph Hellwig if (err < 0) 5520c79d967dSChristoph Hellwig goto restore_opts; 5521c79d967dSChristoph Hellwig 5522ac27a0ecSDave Kleikamp /* 5523ac27a0ecSDave Kleikamp * First of all, the unconditional stuff we have to do 5524ac27a0ecSDave Kleikamp * to disable replay of the journal when we next remount 5525ac27a0ecSDave Kleikamp */ 55261751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 5527ac27a0ecSDave Kleikamp 5528ac27a0ecSDave Kleikamp /* 5529ac27a0ecSDave Kleikamp * OK, test if we are remounting a valid rw partition 5530ac27a0ecSDave Kleikamp * readonly, and if so set the rdonly flag and then 5531ac27a0ecSDave Kleikamp * mark the partition as valid again. 5532ac27a0ecSDave Kleikamp */ 5533617ba13bSMingming Cao if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 5534617ba13bSMingming Cao (sbi->s_mount_state & EXT4_VALID_FS)) 5535ac27a0ecSDave Kleikamp es->s_state = cpu_to_le16(sbi->s_mount_state); 5536ac27a0ecSDave Kleikamp 5537a63c9eb2STheodore Ts'o if (sbi->s_journal) 5538617ba13bSMingming Cao ext4_mark_recovery_complete(sb, es); 55392dca60d9STheodore Ts'o if (sbi->s_mmp_tsk) 55402dca60d9STheodore Ts'o kthread_stop(sbi->s_mmp_tsk); 5541ac27a0ecSDave Kleikamp } else { 5542a13fb1a4SEric Sandeen /* Make sure we can mount this feature set readwrite */ 5543e2b911c5SDarrick J. Wong if (ext4_has_feature_readonly(sb) || 55442cb5cc8bSDarrick J. Wong !ext4_feature_set_ok(sb, 0)) { 5545ac27a0ecSDave Kleikamp err = -EROFS; 5546ac27a0ecSDave Kleikamp goto restore_opts; 5547ac27a0ecSDave Kleikamp } 5548ead6596bSEric Sandeen /* 55498a266467STheodore Ts'o * Make sure the group descriptor checksums 55500b8e58a1SAndreas Dilger * are sane. If they aren't, refuse to remount r/w. 55518a266467STheodore Ts'o */ 55528a266467STheodore Ts'o for (g = 0; g < sbi->s_groups_count; g++) { 55538a266467STheodore Ts'o struct ext4_group_desc *gdp = 55548a266467STheodore Ts'o ext4_get_group_desc(sb, g, NULL); 55558a266467STheodore Ts'o 5556feb0ab32SDarrick J. Wong if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 5557b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 5558b31e1552SEric Sandeen "ext4_remount: Checksum for group %u failed (%u!=%u)", 5559e2b911c5SDarrick J. Wong g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 55608a266467STheodore Ts'o le16_to_cpu(gdp->bg_checksum)); 55616a797d27SDarrick J. Wong err = -EFSBADCRC; 55628a266467STheodore Ts'o goto restore_opts; 55638a266467STheodore Ts'o } 55648a266467STheodore Ts'o } 55658a266467STheodore Ts'o 55668a266467STheodore Ts'o /* 5567ead6596bSEric Sandeen * If we have an unprocessed orphan list hanging 5568ead6596bSEric Sandeen * around from a previously readonly bdev mount, 5569ead6596bSEric Sandeen * require a full umount/remount for now. 5570ead6596bSEric Sandeen */ 5571ead6596bSEric Sandeen if (es->s_last_orphan) { 5572b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "Couldn't " 5573ead6596bSEric Sandeen "remount RDWR because of unprocessed " 5574ead6596bSEric Sandeen "orphan inode list. Please " 5575b31e1552SEric Sandeen "umount/remount instead"); 5576ead6596bSEric Sandeen err = -EINVAL; 5577ead6596bSEric Sandeen goto restore_opts; 5578ead6596bSEric Sandeen } 5579ead6596bSEric Sandeen 5580ac27a0ecSDave Kleikamp /* 5581ac27a0ecSDave Kleikamp * Mounting a RDONLY partition read-write, so reread 5582ac27a0ecSDave Kleikamp * and store the current valid flag. (It may have 5583ac27a0ecSDave Kleikamp * been changed by e2fsck since we originally mounted 5584ac27a0ecSDave Kleikamp * the partition.) 5585ac27a0ecSDave Kleikamp */ 55860390131bSFrank Mayhar if (sbi->s_journal) 5587617ba13bSMingming Cao ext4_clear_journal_err(sb, es); 5588ac27a0ecSDave Kleikamp sbi->s_mount_state = le16_to_cpu(es->s_state); 5589c89128a0SJaegeuk Kim 5590c89128a0SJaegeuk Kim err = ext4_setup_super(sb, es, 0); 5591c89128a0SJaegeuk Kim if (err) 5592c89128a0SJaegeuk Kim goto restore_opts; 5593c89128a0SJaegeuk Kim 55941751e8a6SLinus Torvalds sb->s_flags &= ~SB_RDONLY; 5595e2b911c5SDarrick J. Wong if (ext4_has_feature_mmp(sb)) 5596c5e06d10SJohann Lombardi if (ext4_multi_mount_protect(sb, 5597c5e06d10SJohann Lombardi le64_to_cpu(es->s_mmp_block))) { 5598c5e06d10SJohann Lombardi err = -EROFS; 5599c5e06d10SJohann Lombardi goto restore_opts; 5600c5e06d10SJohann Lombardi } 5601c79d967dSChristoph Hellwig enable_quota = 1; 5602ac27a0ecSDave Kleikamp } 5603ac27a0ecSDave Kleikamp } 5604bfff6873SLukas Czerner 5605bfff6873SLukas Czerner /* 5606bfff6873SLukas Czerner * Reinitialize lazy itable initialization thread based on 5607bfff6873SLukas Czerner * current settings 5608bfff6873SLukas Czerner */ 5609bc98a42cSDavid Howells if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 5610bfff6873SLukas Czerner ext4_unregister_li_request(sb); 5611bfff6873SLukas Czerner else { 5612bfff6873SLukas Czerner ext4_group_t first_not_zeroed; 5613bfff6873SLukas Czerner first_not_zeroed = ext4_has_uninit_itable(sb); 5614bfff6873SLukas Czerner ext4_register_li_request(sb, first_not_zeroed); 5615bfff6873SLukas Czerner } 5616bfff6873SLukas Czerner 56176fd058f7STheodore Ts'o ext4_setup_system_zone(sb); 5618c89128a0SJaegeuk Kim if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 5619c89128a0SJaegeuk Kim err = ext4_commit_super(sb, 1); 5620c89128a0SJaegeuk Kim if (err) 5621c89128a0SJaegeuk Kim goto restore_opts; 5622c89128a0SJaegeuk Kim } 56230390131bSFrank Mayhar 5624ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 5625ac27a0ecSDave Kleikamp /* Release old quota file names */ 5626a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 5627ac27a0ecSDave Kleikamp kfree(old_opts.s_qf_names[i]); 56287c319d32SAditya Kali if (enable_quota) { 56297c319d32SAditya Kali if (sb_any_quota_suspended(sb)) 56300f0dd62fSChristoph Hellwig dquot_resume(sb, -1); 5631e2b911c5SDarrick J. Wong else if (ext4_has_feature_quota(sb)) { 56327c319d32SAditya Kali err = ext4_enable_quotas(sb); 563307724f98STheodore Ts'o if (err) 56347c319d32SAditya Kali goto restore_opts; 56357c319d32SAditya Kali } 56367c319d32SAditya Kali } 56377c319d32SAditya Kali #endif 5638d4c402d9SCurt Wohlgemuth 56391751e8a6SLinus Torvalds *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 5640d4c402d9SCurt Wohlgemuth ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 5641d4c402d9SCurt Wohlgemuth kfree(orig_data); 5642ac27a0ecSDave Kleikamp return 0; 56430b8e58a1SAndreas Dilger 5644ac27a0ecSDave Kleikamp restore_opts: 5645ac27a0ecSDave Kleikamp sb->s_flags = old_sb_flags; 5646ac27a0ecSDave Kleikamp sbi->s_mount_opt = old_opts.s_mount_opt; 5647a2595b8aSTheodore Ts'o sbi->s_mount_opt2 = old_opts.s_mount_opt2; 5648ac27a0ecSDave Kleikamp sbi->s_resuid = old_opts.s_resuid; 5649ac27a0ecSDave Kleikamp sbi->s_resgid = old_opts.s_resgid; 5650ac27a0ecSDave Kleikamp sbi->s_commit_interval = old_opts.s_commit_interval; 565130773840STheodore Ts'o sbi->s_min_batch_time = old_opts.s_min_batch_time; 565230773840STheodore Ts'o sbi->s_max_batch_time = old_opts.s_max_batch_time; 5653ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 5654ac27a0ecSDave Kleikamp sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 5655a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 565633458eabSTheodore Ts'o to_free[i] = get_qf_name(sb, sbi, i); 565733458eabSTheodore Ts'o rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 5658ac27a0ecSDave Kleikamp } 565933458eabSTheodore Ts'o synchronize_rcu(); 566033458eabSTheodore Ts'o for (i = 0; i < EXT4_MAXQUOTAS; i++) 566133458eabSTheodore Ts'o kfree(to_free[i]); 5662ac27a0ecSDave Kleikamp #endif 5663d4c402d9SCurt Wohlgemuth kfree(orig_data); 5664ac27a0ecSDave Kleikamp return err; 5665ac27a0ecSDave Kleikamp } 5666ac27a0ecSDave Kleikamp 5667689c958cSLi Xi #ifdef CONFIG_QUOTA 5668689c958cSLi Xi static int ext4_statfs_project(struct super_block *sb, 5669689c958cSLi Xi kprojid_t projid, struct kstatfs *buf) 5670689c958cSLi Xi { 5671689c958cSLi Xi struct kqid qid; 5672689c958cSLi Xi struct dquot *dquot; 5673689c958cSLi Xi u64 limit; 5674689c958cSLi Xi u64 curblock; 5675689c958cSLi Xi 5676689c958cSLi Xi qid = make_kqid_projid(projid); 5677689c958cSLi Xi dquot = dqget(sb, qid); 5678689c958cSLi Xi if (IS_ERR(dquot)) 5679689c958cSLi Xi return PTR_ERR(dquot); 56807b9ca4c6SJan Kara spin_lock(&dquot->dq_dqb_lock); 5681689c958cSLi Xi 5682a08fe66eSChengguang Xu limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 5683a08fe66eSChengguang Xu dquot->dq_dqb.dqb_bhardlimit); 568457c32ea4SChengguang Xu limit >>= sb->s_blocksize_bits; 568557c32ea4SChengguang Xu 5686689c958cSLi Xi if (limit && buf->f_blocks > limit) { 5687f06925c7SKonstantin Khlebnikov curblock = (dquot->dq_dqb.dqb_curspace + 5688f06925c7SKonstantin Khlebnikov dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 5689689c958cSLi Xi buf->f_blocks = limit; 5690689c958cSLi Xi buf->f_bfree = buf->f_bavail = 5691689c958cSLi Xi (buf->f_blocks > curblock) ? 5692689c958cSLi Xi (buf->f_blocks - curblock) : 0; 5693689c958cSLi Xi } 5694689c958cSLi Xi 5695a08fe66eSChengguang Xu limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 5696a08fe66eSChengguang Xu dquot->dq_dqb.dqb_ihardlimit); 5697689c958cSLi Xi if (limit && buf->f_files > limit) { 5698689c958cSLi Xi buf->f_files = limit; 5699689c958cSLi Xi buf->f_ffree = 5700689c958cSLi Xi (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 5701689c958cSLi Xi (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 5702689c958cSLi Xi } 5703689c958cSLi Xi 57047b9ca4c6SJan Kara spin_unlock(&dquot->dq_dqb_lock); 5705689c958cSLi Xi dqput(dquot); 5706689c958cSLi Xi return 0; 5707689c958cSLi Xi } 5708689c958cSLi Xi #endif 5709689c958cSLi Xi 5710617ba13bSMingming Cao static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 5711ac27a0ecSDave Kleikamp { 5712ac27a0ecSDave Kleikamp struct super_block *sb = dentry->d_sb; 5713617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 5714617ba13bSMingming Cao struct ext4_super_block *es = sbi->s_es; 571527dd4385SLukas Czerner ext4_fsblk_t overhead = 0, resv_blocks; 5716960cc398SPekka Enberg u64 fsid; 5717d02a9391SKazuya Mio s64 bfree; 571827dd4385SLukas Czerner resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 5719ac27a0ecSDave Kleikamp 5720952fc18eSTheodore Ts'o if (!test_opt(sb, MINIX_DF)) 5721952fc18eSTheodore Ts'o overhead = sbi->s_overhead; 5722ac27a0ecSDave Kleikamp 5723617ba13bSMingming Cao buf->f_type = EXT4_SUPER_MAGIC; 5724ac27a0ecSDave Kleikamp buf->f_bsize = sb->s_blocksize; 5725b72f78cbSEric Sandeen buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 572657042651STheodore Ts'o bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 572757042651STheodore Ts'o percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 5728d02a9391SKazuya Mio /* prevent underflow in case that few free space is available */ 572957042651STheodore Ts'o buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 573027dd4385SLukas Czerner buf->f_bavail = buf->f_bfree - 573127dd4385SLukas Czerner (ext4_r_blocks_count(es) + resv_blocks); 573227dd4385SLukas Czerner if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 5733ac27a0ecSDave Kleikamp buf->f_bavail = 0; 5734ac27a0ecSDave Kleikamp buf->f_files = le32_to_cpu(es->s_inodes_count); 573552d9f3b4SPeter Zijlstra buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 5736617ba13bSMingming Cao buf->f_namelen = EXT4_NAME_LEN; 5737960cc398SPekka Enberg fsid = le64_to_cpup((void *)es->s_uuid) ^ 5738960cc398SPekka Enberg le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 5739960cc398SPekka Enberg buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 5740960cc398SPekka Enberg buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 57410b8e58a1SAndreas Dilger 5742689c958cSLi Xi #ifdef CONFIG_QUOTA 5743689c958cSLi Xi if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 5744689c958cSLi Xi sb_has_quota_limits_enabled(sb, PRJQUOTA)) 5745689c958cSLi Xi ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 5746689c958cSLi Xi #endif 5747ac27a0ecSDave Kleikamp return 0; 5748ac27a0ecSDave Kleikamp } 5749ac27a0ecSDave Kleikamp 5750ac27a0ecSDave Kleikamp 5751ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 5752ac27a0ecSDave Kleikamp 5753bc8230eeSJan Kara /* 5754bc8230eeSJan Kara * Helper functions so that transaction is started before we acquire dqio_sem 5755bc8230eeSJan Kara * to keep correct lock ordering of transaction > dqio_sem 5756bc8230eeSJan Kara */ 5757ac27a0ecSDave Kleikamp static inline struct inode *dquot_to_inode(struct dquot *dquot) 5758ac27a0ecSDave Kleikamp { 57594c376dcaSEric W. Biederman return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 5760ac27a0ecSDave Kleikamp } 5761ac27a0ecSDave Kleikamp 5762617ba13bSMingming Cao static int ext4_write_dquot(struct dquot *dquot) 5763ac27a0ecSDave Kleikamp { 5764ac27a0ecSDave Kleikamp int ret, err; 5765ac27a0ecSDave Kleikamp handle_t *handle; 5766ac27a0ecSDave Kleikamp struct inode *inode; 5767ac27a0ecSDave Kleikamp 5768ac27a0ecSDave Kleikamp inode = dquot_to_inode(dquot); 57699924a92aSTheodore Ts'o handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 5770617ba13bSMingming Cao EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 5771ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 5772ac27a0ecSDave Kleikamp return PTR_ERR(handle); 5773ac27a0ecSDave Kleikamp ret = dquot_commit(dquot); 5774617ba13bSMingming Cao err = ext4_journal_stop(handle); 5775ac27a0ecSDave Kleikamp if (!ret) 5776ac27a0ecSDave Kleikamp ret = err; 5777ac27a0ecSDave Kleikamp return ret; 5778ac27a0ecSDave Kleikamp } 5779ac27a0ecSDave Kleikamp 5780617ba13bSMingming Cao static int ext4_acquire_dquot(struct dquot *dquot) 5781ac27a0ecSDave Kleikamp { 5782ac27a0ecSDave Kleikamp int ret, err; 5783ac27a0ecSDave Kleikamp handle_t *handle; 5784ac27a0ecSDave Kleikamp 57859924a92aSTheodore Ts'o handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5786617ba13bSMingming Cao EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 5787ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 5788ac27a0ecSDave Kleikamp return PTR_ERR(handle); 5789ac27a0ecSDave Kleikamp ret = dquot_acquire(dquot); 5790617ba13bSMingming Cao err = ext4_journal_stop(handle); 5791ac27a0ecSDave Kleikamp if (!ret) 5792ac27a0ecSDave Kleikamp ret = err; 5793ac27a0ecSDave Kleikamp return ret; 5794ac27a0ecSDave Kleikamp } 5795ac27a0ecSDave Kleikamp 5796617ba13bSMingming Cao static int ext4_release_dquot(struct dquot *dquot) 5797ac27a0ecSDave Kleikamp { 5798ac27a0ecSDave Kleikamp int ret, err; 5799ac27a0ecSDave Kleikamp handle_t *handle; 5800ac27a0ecSDave Kleikamp 58019924a92aSTheodore Ts'o handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5802617ba13bSMingming Cao EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 58039c3013e9SJan Kara if (IS_ERR(handle)) { 58049c3013e9SJan Kara /* Release dquot anyway to avoid endless cycle in dqput() */ 58059c3013e9SJan Kara dquot_release(dquot); 5806ac27a0ecSDave Kleikamp return PTR_ERR(handle); 58079c3013e9SJan Kara } 5808ac27a0ecSDave Kleikamp ret = dquot_release(dquot); 5809617ba13bSMingming Cao err = ext4_journal_stop(handle); 5810ac27a0ecSDave Kleikamp if (!ret) 5811ac27a0ecSDave Kleikamp ret = err; 5812ac27a0ecSDave Kleikamp return ret; 5813ac27a0ecSDave Kleikamp } 5814ac27a0ecSDave Kleikamp 5815617ba13bSMingming Cao static int ext4_mark_dquot_dirty(struct dquot *dquot) 5816ac27a0ecSDave Kleikamp { 5817262b4662SJan Kara struct super_block *sb = dquot->dq_sb; 5818262b4662SJan Kara struct ext4_sb_info *sbi = EXT4_SB(sb); 5819262b4662SJan Kara 58202c8be6b2SJan Kara /* Are we journaling quotas? */ 5821e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb) || 5822262b4662SJan Kara sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 5823ac27a0ecSDave Kleikamp dquot_mark_dquot_dirty(dquot); 5824617ba13bSMingming Cao return ext4_write_dquot(dquot); 5825ac27a0ecSDave Kleikamp } else { 5826ac27a0ecSDave Kleikamp return dquot_mark_dquot_dirty(dquot); 5827ac27a0ecSDave Kleikamp } 5828ac27a0ecSDave Kleikamp } 5829ac27a0ecSDave Kleikamp 5830617ba13bSMingming Cao static int ext4_write_info(struct super_block *sb, int type) 5831ac27a0ecSDave Kleikamp { 5832ac27a0ecSDave Kleikamp int ret, err; 5833ac27a0ecSDave Kleikamp handle_t *handle; 5834ac27a0ecSDave Kleikamp 5835ac27a0ecSDave Kleikamp /* Data block + inode block */ 58362b0143b5SDavid Howells handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); 5837ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 5838ac27a0ecSDave Kleikamp return PTR_ERR(handle); 5839ac27a0ecSDave Kleikamp ret = dquot_commit_info(sb, type); 5840617ba13bSMingming Cao err = ext4_journal_stop(handle); 5841ac27a0ecSDave Kleikamp if (!ret) 5842ac27a0ecSDave Kleikamp ret = err; 5843ac27a0ecSDave Kleikamp return ret; 5844ac27a0ecSDave Kleikamp } 5845ac27a0ecSDave Kleikamp 5846ac27a0ecSDave Kleikamp /* 5847ac27a0ecSDave Kleikamp * Turn on quotas during mount time - we need to find 5848ac27a0ecSDave Kleikamp * the quota file and such... 5849ac27a0ecSDave Kleikamp */ 5850617ba13bSMingming Cao static int ext4_quota_on_mount(struct super_block *sb, int type) 5851ac27a0ecSDave Kleikamp { 585233458eabSTheodore Ts'o return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type), 5853617ba13bSMingming Cao EXT4_SB(sb)->s_jquota_fmt, type); 5854ac27a0ecSDave Kleikamp } 5855ac27a0ecSDave Kleikamp 5856daf647d2STheodore Ts'o static void lockdep_set_quota_inode(struct inode *inode, int subclass) 5857daf647d2STheodore Ts'o { 5858daf647d2STheodore Ts'o struct ext4_inode_info *ei = EXT4_I(inode); 5859daf647d2STheodore Ts'o 5860daf647d2STheodore Ts'o /* The first argument of lockdep_set_subclass has to be 5861daf647d2STheodore Ts'o * *exactly* the same as the argument to init_rwsem() --- in 5862daf647d2STheodore Ts'o * this case, in init_once() --- or lockdep gets unhappy 5863daf647d2STheodore Ts'o * because the name of the lock is set using the 5864daf647d2STheodore Ts'o * stringification of the argument to init_rwsem(). 5865daf647d2STheodore Ts'o */ 5866daf647d2STheodore Ts'o (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 5867daf647d2STheodore Ts'o lockdep_set_subclass(&ei->i_data_sem, subclass); 5868daf647d2STheodore Ts'o } 5869daf647d2STheodore Ts'o 5870ac27a0ecSDave Kleikamp /* 5871ac27a0ecSDave Kleikamp * Standard function to be called on quota_on 5872ac27a0ecSDave Kleikamp */ 5873617ba13bSMingming Cao static int ext4_quota_on(struct super_block *sb, int type, int format_id, 58748c54ca9cSAl Viro const struct path *path) 5875ac27a0ecSDave Kleikamp { 5876ac27a0ecSDave Kleikamp int err; 5877ac27a0ecSDave Kleikamp 5878ac27a0ecSDave Kleikamp if (!test_opt(sb, QUOTA)) 5879ac27a0ecSDave Kleikamp return -EINVAL; 58800623543bSJan Kara 5881ac27a0ecSDave Kleikamp /* Quotafile not on the same filesystem? */ 5882d8c9584eSAl Viro if (path->dentry->d_sb != sb) 5883ac27a0ecSDave Kleikamp return -EXDEV; 58840623543bSJan Kara /* Journaling quota? */ 58850623543bSJan Kara if (EXT4_SB(sb)->s_qf_names[type]) { 58862b2d6d01STheodore Ts'o /* Quotafile not in fs root? */ 5887f00c9e44SJan Kara if (path->dentry->d_parent != sb->s_root) 5888b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 5889b31e1552SEric Sandeen "Quota file not on filesystem root. " 5890b31e1552SEric Sandeen "Journaled quota will not work"); 589191389240SJan Kara sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 589291389240SJan Kara } else { 589391389240SJan Kara /* 589491389240SJan Kara * Clear the flag just in case mount options changed since 589591389240SJan Kara * last time. 589691389240SJan Kara */ 589791389240SJan Kara sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 58980623543bSJan Kara } 58990623543bSJan Kara 59000623543bSJan Kara /* 59010623543bSJan Kara * When we journal data on quota file, we have to flush journal to see 59020623543bSJan Kara * all updates to the file when we bypass pagecache... 59030623543bSJan Kara */ 59040390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal && 59052b0143b5SDavid Howells ext4_should_journal_data(d_inode(path->dentry))) { 59060623543bSJan Kara /* 59070623543bSJan Kara * We don't need to lock updates but journal_flush() could 59080623543bSJan Kara * otherwise be livelocked... 59090623543bSJan Kara */ 59100623543bSJan Kara jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 59117ffe1ea8SHidehiro Kawai err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 59120623543bSJan Kara jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 5913f00c9e44SJan Kara if (err) 59147ffe1ea8SHidehiro Kawai return err; 59157ffe1ea8SHidehiro Kawai } 5916957153fcSJan Kara 5917daf647d2STheodore Ts'o lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 5918daf647d2STheodore Ts'o err = dquot_quota_on(sb, type, format_id, path); 5919957153fcSJan Kara if (err) { 5920daf647d2STheodore Ts'o lockdep_set_quota_inode(path->dentry->d_inode, 5921daf647d2STheodore Ts'o I_DATA_SEM_NORMAL); 5922957153fcSJan Kara } else { 5923957153fcSJan Kara struct inode *inode = d_inode(path->dentry); 5924957153fcSJan Kara handle_t *handle; 5925957153fcSJan Kara 592661a92987SJan Kara /* 592761a92987SJan Kara * Set inode flags to prevent userspace from messing with quota 592861a92987SJan Kara * files. If this fails, we return success anyway since quotas 592961a92987SJan Kara * are already enabled and this is not a hard failure. 593061a92987SJan Kara */ 5931957153fcSJan Kara inode_lock(inode); 5932957153fcSJan Kara handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 5933957153fcSJan Kara if (IS_ERR(handle)) 5934957153fcSJan Kara goto unlock_inode; 5935957153fcSJan Kara EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 5936957153fcSJan Kara inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 5937957153fcSJan Kara S_NOATIME | S_IMMUTABLE); 5938957153fcSJan Kara ext4_mark_inode_dirty(handle, inode); 5939957153fcSJan Kara ext4_journal_stop(handle); 5940957153fcSJan Kara unlock_inode: 5941957153fcSJan Kara inode_unlock(inode); 5942957153fcSJan Kara } 5943daf647d2STheodore Ts'o return err; 5944ac27a0ecSDave Kleikamp } 5945ac27a0ecSDave Kleikamp 59467c319d32SAditya Kali static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 59477c319d32SAditya Kali unsigned int flags) 59487c319d32SAditya Kali { 59497c319d32SAditya Kali int err; 59507c319d32SAditya Kali struct inode *qf_inode; 5951a2d4a646SJan Kara unsigned long qf_inums[EXT4_MAXQUOTAS] = { 59527c319d32SAditya Kali le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5953689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5954689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 59557c319d32SAditya Kali }; 59567c319d32SAditya Kali 5957e2b911c5SDarrick J. Wong BUG_ON(!ext4_has_feature_quota(sb)); 59587c319d32SAditya Kali 59597c319d32SAditya Kali if (!qf_inums[type]) 59607c319d32SAditya Kali return -EPERM; 59617c319d32SAditya Kali 59628a363970STheodore Ts'o qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 59637c319d32SAditya Kali if (IS_ERR(qf_inode)) { 59647c319d32SAditya Kali ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 59657c319d32SAditya Kali return PTR_ERR(qf_inode); 59667c319d32SAditya Kali } 59677c319d32SAditya Kali 5968bcb13850SJan Kara /* Don't account quota for quota files to avoid recursion */ 5969bcb13850SJan Kara qf_inode->i_flags |= S_NOQUOTA; 5970daf647d2STheodore Ts'o lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 59717212b95eSJan Kara err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 5972daf647d2STheodore Ts'o if (err) 5973daf647d2STheodore Ts'o lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 597461157b24SPan Bian iput(qf_inode); 59757c319d32SAditya Kali 59767c319d32SAditya Kali return err; 59777c319d32SAditya Kali } 59787c319d32SAditya Kali 59797c319d32SAditya Kali /* Enable usage tracking for all quota types. */ 59807c319d32SAditya Kali static int ext4_enable_quotas(struct super_block *sb) 59817c319d32SAditya Kali { 59827c319d32SAditya Kali int type, err = 0; 5983a2d4a646SJan Kara unsigned long qf_inums[EXT4_MAXQUOTAS] = { 59847c319d32SAditya Kali le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5985689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5986689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 59877c319d32SAditya Kali }; 598849da9392SJan Kara bool quota_mopt[EXT4_MAXQUOTAS] = { 598949da9392SJan Kara test_opt(sb, USRQUOTA), 599049da9392SJan Kara test_opt(sb, GRPQUOTA), 599149da9392SJan Kara test_opt(sb, PRJQUOTA), 599249da9392SJan Kara }; 59937c319d32SAditya Kali 599491389240SJan Kara sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 5995a2d4a646SJan Kara for (type = 0; type < EXT4_MAXQUOTAS; type++) { 59967c319d32SAditya Kali if (qf_inums[type]) { 59977c319d32SAditya Kali err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 599849da9392SJan Kara DQUOT_USAGE_ENABLED | 599949da9392SJan Kara (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 60007c319d32SAditya Kali if (err) { 60017c319d32SAditya Kali ext4_warning(sb, 600272ba7450STheodore Ts'o "Failed to enable quota tracking " 600372ba7450STheodore Ts'o "(type=%d, err=%d). Please run " 600472ba7450STheodore Ts'o "e2fsck to fix.", type, err); 60057f144fd0SJunichi Uekawa for (type--; type >= 0; type--) 60067f144fd0SJunichi Uekawa dquot_quota_off(sb, type); 60077f144fd0SJunichi Uekawa 60087c319d32SAditya Kali return err; 60097c319d32SAditya Kali } 60107c319d32SAditya Kali } 60117c319d32SAditya Kali } 60127c319d32SAditya Kali return 0; 60137c319d32SAditya Kali } 60147c319d32SAditya Kali 6015ca0e05e4SDmitry Monakhov static int ext4_quota_off(struct super_block *sb, int type) 6016ca0e05e4SDmitry Monakhov { 601721f97697SJan Kara struct inode *inode = sb_dqopt(sb)->files[type]; 601821f97697SJan Kara handle_t *handle; 6019957153fcSJan Kara int err; 602021f97697SJan Kara 602187009d86SDmitry Monakhov /* Force all delayed allocation blocks to be allocated. 602287009d86SDmitry Monakhov * Caller already holds s_umount sem */ 602387009d86SDmitry Monakhov if (test_opt(sb, DELALLOC)) 6024ca0e05e4SDmitry Monakhov sync_filesystem(sb); 6025ca0e05e4SDmitry Monakhov 6026957153fcSJan Kara if (!inode || !igrab(inode)) 60270b268590SAmir Goldstein goto out; 60280b268590SAmir Goldstein 6029957153fcSJan Kara err = dquot_quota_off(sb, type); 6030964edf66SJan Kara if (err || ext4_has_feature_quota(sb)) 6031957153fcSJan Kara goto out_put; 6032957153fcSJan Kara 6033957153fcSJan Kara inode_lock(inode); 603461a92987SJan Kara /* 603561a92987SJan Kara * Update modification times of quota files when userspace can 603661a92987SJan Kara * start looking at them. If we fail, we return success anyway since 603761a92987SJan Kara * this is not a hard failure and quotas are already disabled. 603861a92987SJan Kara */ 60399924a92aSTheodore Ts'o handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 604021f97697SJan Kara if (IS_ERR(handle)) 6041957153fcSJan Kara goto out_unlock; 6042957153fcSJan Kara EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 6043957153fcSJan Kara inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 6044eeca7ea1SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 604521f97697SJan Kara ext4_mark_inode_dirty(handle, inode); 604621f97697SJan Kara ext4_journal_stop(handle); 6047957153fcSJan Kara out_unlock: 6048957153fcSJan Kara inode_unlock(inode); 6049957153fcSJan Kara out_put: 6050964edf66SJan Kara lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 6051957153fcSJan Kara iput(inode); 6052957153fcSJan Kara return err; 605321f97697SJan Kara out: 6054ca0e05e4SDmitry Monakhov return dquot_quota_off(sb, type); 6055ca0e05e4SDmitry Monakhov } 6056ca0e05e4SDmitry Monakhov 6057ac27a0ecSDave Kleikamp /* Read data from quotafile - avoid pagecache and such because we cannot afford 6058ac27a0ecSDave Kleikamp * acquiring the locks... As quota files are never truncated and quota code 6059ac27a0ecSDave Kleikamp * itself serializes the operations (and no one else should touch the files) 6060ac27a0ecSDave Kleikamp * we don't have to be afraid of races */ 6061617ba13bSMingming Cao static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 6062ac27a0ecSDave Kleikamp size_t len, loff_t off) 6063ac27a0ecSDave Kleikamp { 6064ac27a0ecSDave Kleikamp struct inode *inode = sb_dqopt(sb)->files[type]; 6065725d26d3SAneesh Kumar K.V ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 6066ac27a0ecSDave Kleikamp int offset = off & (sb->s_blocksize - 1); 6067ac27a0ecSDave Kleikamp int tocopy; 6068ac27a0ecSDave Kleikamp size_t toread; 6069ac27a0ecSDave Kleikamp struct buffer_head *bh; 6070ac27a0ecSDave Kleikamp loff_t i_size = i_size_read(inode); 6071ac27a0ecSDave Kleikamp 6072ac27a0ecSDave Kleikamp if (off > i_size) 6073ac27a0ecSDave Kleikamp return 0; 6074ac27a0ecSDave Kleikamp if (off+len > i_size) 6075ac27a0ecSDave Kleikamp len = i_size-off; 6076ac27a0ecSDave Kleikamp toread = len; 6077ac27a0ecSDave Kleikamp while (toread > 0) { 6078ac27a0ecSDave Kleikamp tocopy = sb->s_blocksize - offset < toread ? 6079ac27a0ecSDave Kleikamp sb->s_blocksize - offset : toread; 60801c215028STheodore Ts'o bh = ext4_bread(NULL, inode, blk, 0); 60811c215028STheodore Ts'o if (IS_ERR(bh)) 60821c215028STheodore Ts'o return PTR_ERR(bh); 6083ac27a0ecSDave Kleikamp if (!bh) /* A hole? */ 6084ac27a0ecSDave Kleikamp memset(data, 0, tocopy); 6085ac27a0ecSDave Kleikamp else 6086ac27a0ecSDave Kleikamp memcpy(data, bh->b_data+offset, tocopy); 6087ac27a0ecSDave Kleikamp brelse(bh); 6088ac27a0ecSDave Kleikamp offset = 0; 6089ac27a0ecSDave Kleikamp toread -= tocopy; 6090ac27a0ecSDave Kleikamp data += tocopy; 6091ac27a0ecSDave Kleikamp blk++; 6092ac27a0ecSDave Kleikamp } 6093ac27a0ecSDave Kleikamp return len; 6094ac27a0ecSDave Kleikamp } 6095ac27a0ecSDave Kleikamp 6096ac27a0ecSDave Kleikamp /* Write to quotafile (we know the transaction is already started and has 6097ac27a0ecSDave Kleikamp * enough credits) */ 6098617ba13bSMingming Cao static ssize_t ext4_quota_write(struct super_block *sb, int type, 6099ac27a0ecSDave Kleikamp const char *data, size_t len, loff_t off) 6100ac27a0ecSDave Kleikamp { 6101ac27a0ecSDave Kleikamp struct inode *inode = sb_dqopt(sb)->files[type]; 6102725d26d3SAneesh Kumar K.V ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 61031c215028STheodore Ts'o int err, offset = off & (sb->s_blocksize - 1); 6104c5e298aeSTheodore Ts'o int retries = 0; 6105ac27a0ecSDave Kleikamp struct buffer_head *bh; 6106ac27a0ecSDave Kleikamp handle_t *handle = journal_current_handle(); 6107ac27a0ecSDave Kleikamp 61080390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal && !handle) { 6109b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 6110b31e1552SEric Sandeen " cancelled because transaction is not started", 61119c3013e9SJan Kara (unsigned long long)off, (unsigned long long)len); 61129c3013e9SJan Kara return -EIO; 61139c3013e9SJan Kara } 611467eeb568SDmitry Monakhov /* 611567eeb568SDmitry Monakhov * Since we account only one data block in transaction credits, 611667eeb568SDmitry Monakhov * then it is impossible to cross a block boundary. 611767eeb568SDmitry Monakhov */ 611867eeb568SDmitry Monakhov if (sb->s_blocksize - offset < len) { 611967eeb568SDmitry Monakhov ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 612067eeb568SDmitry Monakhov " cancelled because not block aligned", 612167eeb568SDmitry Monakhov (unsigned long long)off, (unsigned long long)len); 612267eeb568SDmitry Monakhov return -EIO; 612367eeb568SDmitry Monakhov } 612467eeb568SDmitry Monakhov 6125c5e298aeSTheodore Ts'o do { 6126c5e298aeSTheodore Ts'o bh = ext4_bread(handle, inode, blk, 6127c5e298aeSTheodore Ts'o EXT4_GET_BLOCKS_CREATE | 6128c5e298aeSTheodore Ts'o EXT4_GET_BLOCKS_METADATA_NOFAIL); 612945586c70SMasahiro Yamada } while (PTR_ERR(bh) == -ENOSPC && 6130c5e298aeSTheodore Ts'o ext4_should_retry_alloc(inode->i_sb, &retries)); 61311c215028STheodore Ts'o if (IS_ERR(bh)) 61321c215028STheodore Ts'o return PTR_ERR(bh); 6133ac27a0ecSDave Kleikamp if (!bh) 6134ac27a0ecSDave Kleikamp goto out; 61355d601255Sliang xie BUFFER_TRACE(bh, "get write access"); 6136617ba13bSMingming Cao err = ext4_journal_get_write_access(handle, bh); 6137ac27a0ecSDave Kleikamp if (err) { 6138ac27a0ecSDave Kleikamp brelse(bh); 61391c215028STheodore Ts'o return err; 6140ac27a0ecSDave Kleikamp } 6141ac27a0ecSDave Kleikamp lock_buffer(bh); 614267eeb568SDmitry Monakhov memcpy(bh->b_data+offset, data, len); 6143ac27a0ecSDave Kleikamp flush_dcache_page(bh->b_page); 6144ac27a0ecSDave Kleikamp unlock_buffer(bh); 61450390131bSFrank Mayhar err = ext4_handle_dirty_metadata(handle, NULL, bh); 6146ac27a0ecSDave Kleikamp brelse(bh); 6147ac27a0ecSDave Kleikamp out: 614867eeb568SDmitry Monakhov if (inode->i_size < off + len) { 614967eeb568SDmitry Monakhov i_size_write(inode, off + len); 6150617ba13bSMingming Cao EXT4_I(inode)->i_disksize = inode->i_size; 6151617ba13bSMingming Cao ext4_mark_inode_dirty(handle, inode); 615221f97697SJan Kara } 615367eeb568SDmitry Monakhov return len; 6154ac27a0ecSDave Kleikamp } 6155ac27a0ecSDave Kleikamp #endif 6156ac27a0ecSDave Kleikamp 6157152a0836SAl Viro static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 6158152a0836SAl Viro const char *dev_name, void *data) 6159ac27a0ecSDave Kleikamp { 6160152a0836SAl Viro return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 6161ac27a0ecSDave Kleikamp } 6162ac27a0ecSDave Kleikamp 6163c290ea01SJan Kara #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 616424b58424STheodore Ts'o static inline void register_as_ext2(void) 616524b58424STheodore Ts'o { 616624b58424STheodore Ts'o int err = register_filesystem(&ext2_fs_type); 616724b58424STheodore Ts'o if (err) 616824b58424STheodore Ts'o printk(KERN_WARNING 616924b58424STheodore Ts'o "EXT4-fs: Unable to register as ext2 (%d)\n", err); 617024b58424STheodore Ts'o } 617124b58424STheodore Ts'o 617224b58424STheodore Ts'o static inline void unregister_as_ext2(void) 617324b58424STheodore Ts'o { 617424b58424STheodore Ts'o unregister_filesystem(&ext2_fs_type); 617524b58424STheodore Ts'o } 61762035e776STheodore Ts'o 61772035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb) 61782035e776STheodore Ts'o { 6179e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext2_incompat_features(sb)) 61802035e776STheodore Ts'o return 0; 6181bc98a42cSDavid Howells if (sb_rdonly(sb)) 61822035e776STheodore Ts'o return 1; 6183e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext2_ro_compat_features(sb)) 61842035e776STheodore Ts'o return 0; 61852035e776STheodore Ts'o return 1; 61862035e776STheodore Ts'o } 618724b58424STheodore Ts'o #else 618824b58424STheodore Ts'o static inline void register_as_ext2(void) { } 618924b58424STheodore Ts'o static inline void unregister_as_ext2(void) { } 61902035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 619124b58424STheodore Ts'o #endif 619224b58424STheodore Ts'o 619324b58424STheodore Ts'o static inline void register_as_ext3(void) 619424b58424STheodore Ts'o { 619524b58424STheodore Ts'o int err = register_filesystem(&ext3_fs_type); 619624b58424STheodore Ts'o if (err) 619724b58424STheodore Ts'o printk(KERN_WARNING 619824b58424STheodore Ts'o "EXT4-fs: Unable to register as ext3 (%d)\n", err); 619924b58424STheodore Ts'o } 620024b58424STheodore Ts'o 620124b58424STheodore Ts'o static inline void unregister_as_ext3(void) 620224b58424STheodore Ts'o { 620324b58424STheodore Ts'o unregister_filesystem(&ext3_fs_type); 620424b58424STheodore Ts'o } 62052035e776STheodore Ts'o 62062035e776STheodore Ts'o static inline int ext3_feature_set_ok(struct super_block *sb) 62072035e776STheodore Ts'o { 6208e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext3_incompat_features(sb)) 62092035e776STheodore Ts'o return 0; 6210e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal(sb)) 62112035e776STheodore Ts'o return 0; 6212bc98a42cSDavid Howells if (sb_rdonly(sb)) 62132035e776STheodore Ts'o return 1; 6214e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext3_ro_compat_features(sb)) 62152035e776STheodore Ts'o return 0; 62162035e776STheodore Ts'o return 1; 62172035e776STheodore Ts'o } 621824b58424STheodore Ts'o 621903010a33STheodore Ts'o static struct file_system_type ext4_fs_type = { 6220ac27a0ecSDave Kleikamp .owner = THIS_MODULE, 622103010a33STheodore Ts'o .name = "ext4", 6222152a0836SAl Viro .mount = ext4_mount, 6223ac27a0ecSDave Kleikamp .kill_sb = kill_block_super, 6224ac27a0ecSDave Kleikamp .fs_flags = FS_REQUIRES_DEV, 6225ac27a0ecSDave Kleikamp }; 62267f78e035SEric W. Biederman MODULE_ALIAS_FS("ext4"); 6227ac27a0ecSDave Kleikamp 6228e9e3bcecSEric Sandeen /* Shared across all ext4 file systems */ 6229e9e3bcecSEric Sandeen wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 6230e9e3bcecSEric Sandeen 62315dabfc78STheodore Ts'o static int __init ext4_init_fs(void) 6232ac27a0ecSDave Kleikamp { 6233e9e3bcecSEric Sandeen int i, err; 6234c9de560dSAlex Tomas 6235e294a537STheodore Ts'o ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 623607c0c5d8SAl Viro ext4_li_info = NULL; 623707c0c5d8SAl Viro mutex_init(&ext4_li_mtx); 623807c0c5d8SAl Viro 62399a4c8019SCarlos Maiolino /* Build-time check for flags consistency */ 624012e9b892SDmitry Monakhov ext4_check_flag_values(); 6241e9e3bcecSEric Sandeen 6242e142d052SJan Kara for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 6243e9e3bcecSEric Sandeen init_waitqueue_head(&ext4__ioend_wq[i]); 6244e9e3bcecSEric Sandeen 624551865fdaSZheng Liu err = ext4_init_es(); 62466fd058f7STheodore Ts'o if (err) 62476fd058f7STheodore Ts'o return err; 624851865fdaSZheng Liu 62491dc0aa46SEric Whitney err = ext4_init_pending(); 62501dc0aa46SEric Whitney if (err) 625122cfe4b4SEric Biggers goto out7; 625222cfe4b4SEric Biggers 625322cfe4b4SEric Biggers err = ext4_init_post_read_processing(); 625422cfe4b4SEric Biggers if (err) 62551dc0aa46SEric Whitney goto out6; 62561dc0aa46SEric Whitney 625751865fdaSZheng Liu err = ext4_init_pageio(); 625851865fdaSZheng Liu if (err) 6259b5799018STheodore Ts'o goto out5; 626051865fdaSZheng Liu 62615dabfc78STheodore Ts'o err = ext4_init_system_zone(); 6262bd2d0210STheodore Ts'o if (err) 6263b5799018STheodore Ts'o goto out4; 6264857ac889SLukas Czerner 6265b5799018STheodore Ts'o err = ext4_init_sysfs(); 6266dd68314cSTheodore Ts'o if (err) 6267b5799018STheodore Ts'o goto out3; 6268857ac889SLukas Czerner 62695dabfc78STheodore Ts'o err = ext4_init_mballoc(); 6270ac27a0ecSDave Kleikamp if (err) 6271c9de560dSAlex Tomas goto out2; 6272ac27a0ecSDave Kleikamp err = init_inodecache(); 6273ac27a0ecSDave Kleikamp if (err) 6274ac27a0ecSDave Kleikamp goto out1; 627524b58424STheodore Ts'o register_as_ext3(); 62762035e776STheodore Ts'o register_as_ext2(); 627703010a33STheodore Ts'o err = register_filesystem(&ext4_fs_type); 6278ac27a0ecSDave Kleikamp if (err) 6279ac27a0ecSDave Kleikamp goto out; 6280bfff6873SLukas Czerner 6281ac27a0ecSDave Kleikamp return 0; 6282ac27a0ecSDave Kleikamp out: 628324b58424STheodore Ts'o unregister_as_ext2(); 628424b58424STheodore Ts'o unregister_as_ext3(); 6285ac27a0ecSDave Kleikamp destroy_inodecache(); 6286ac27a0ecSDave Kleikamp out1: 62875dabfc78STheodore Ts'o ext4_exit_mballoc(); 62889c191f70ST Makphaibulchoke out2: 6289b5799018STheodore Ts'o ext4_exit_sysfs(); 6290b5799018STheodore Ts'o out3: 6291dd68314cSTheodore Ts'o ext4_exit_system_zone(); 6292b5799018STheodore Ts'o out4: 62935dabfc78STheodore Ts'o ext4_exit_pageio(); 6294b5799018STheodore Ts'o out5: 629522cfe4b4SEric Biggers ext4_exit_post_read_processing(); 62961dc0aa46SEric Whitney out6: 629722cfe4b4SEric Biggers ext4_exit_pending(); 629822cfe4b4SEric Biggers out7: 629951865fdaSZheng Liu ext4_exit_es(); 630051865fdaSZheng Liu 6301ac27a0ecSDave Kleikamp return err; 6302ac27a0ecSDave Kleikamp } 6303ac27a0ecSDave Kleikamp 63045dabfc78STheodore Ts'o static void __exit ext4_exit_fs(void) 6305ac27a0ecSDave Kleikamp { 6306bfff6873SLukas Czerner ext4_destroy_lazyinit_thread(); 630724b58424STheodore Ts'o unregister_as_ext2(); 630824b58424STheodore Ts'o unregister_as_ext3(); 630903010a33STheodore Ts'o unregister_filesystem(&ext4_fs_type); 6310ac27a0ecSDave Kleikamp destroy_inodecache(); 63115dabfc78STheodore Ts'o ext4_exit_mballoc(); 6312b5799018STheodore Ts'o ext4_exit_sysfs(); 63135dabfc78STheodore Ts'o ext4_exit_system_zone(); 63145dabfc78STheodore Ts'o ext4_exit_pageio(); 631522cfe4b4SEric Biggers ext4_exit_post_read_processing(); 6316dd12ed14SEric Sandeen ext4_exit_es(); 63171dc0aa46SEric Whitney ext4_exit_pending(); 6318ac27a0ecSDave Kleikamp } 6319ac27a0ecSDave Kleikamp 6320ac27a0ecSDave Kleikamp MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 632183982b6fSTheodore Ts'o MODULE_DESCRIPTION("Fourth Extended Filesystem"); 6322ac27a0ecSDave Kleikamp MODULE_LICENSE("GPL"); 63237ef79ad5STheodore Ts'o MODULE_SOFTDEP("pre: crc32c"); 63245dabfc78STheodore Ts'o module_init(ext4_init_fs) 63255dabfc78STheodore Ts'o module_exit(ext4_exit_fs) 6326