1ac27a0ecSDave Kleikamp /* 2617ba13bSMingming Cao * linux/fs/ext4/super.c 3ac27a0ecSDave Kleikamp * 4ac27a0ecSDave Kleikamp * Copyright (C) 1992, 1993, 1994, 1995 5ac27a0ecSDave Kleikamp * Remy Card (card@masi.ibp.fr) 6ac27a0ecSDave Kleikamp * Laboratoire MASI - Institut Blaise Pascal 7ac27a0ecSDave Kleikamp * Universite Pierre et Marie Curie (Paris VI) 8ac27a0ecSDave Kleikamp * 9ac27a0ecSDave Kleikamp * from 10ac27a0ecSDave Kleikamp * 11ac27a0ecSDave Kleikamp * linux/fs/minix/inode.c 12ac27a0ecSDave Kleikamp * 13ac27a0ecSDave Kleikamp * Copyright (C) 1991, 1992 Linus Torvalds 14ac27a0ecSDave Kleikamp * 15ac27a0ecSDave Kleikamp * Big-endian to little-endian byte-swapping/bitmaps by 16ac27a0ecSDave Kleikamp * David S. Miller (davem@caip.rutgers.edu), 1995 17ac27a0ecSDave Kleikamp */ 18ac27a0ecSDave Kleikamp 19ac27a0ecSDave Kleikamp #include <linux/module.h> 20ac27a0ecSDave Kleikamp #include <linux/string.h> 21ac27a0ecSDave Kleikamp #include <linux/fs.h> 22ac27a0ecSDave Kleikamp #include <linux/time.h> 23c5ca7c76STheodore Ts'o #include <linux/vmalloc.h> 24ac27a0ecSDave Kleikamp #include <linux/slab.h> 25ac27a0ecSDave Kleikamp #include <linux/init.h> 26ac27a0ecSDave Kleikamp #include <linux/blkdev.h> 2766114cadSTejun Heo #include <linux/backing-dev.h> 28ac27a0ecSDave Kleikamp #include <linux/parser.h> 29ac27a0ecSDave Kleikamp #include <linux/buffer_head.h> 30a5694255SChristoph Hellwig #include <linux/exportfs.h> 31ac27a0ecSDave Kleikamp #include <linux/vfs.h> 32ac27a0ecSDave Kleikamp #include <linux/random.h> 33ac27a0ecSDave Kleikamp #include <linux/mount.h> 34ac27a0ecSDave Kleikamp #include <linux/namei.h> 35ac27a0ecSDave Kleikamp #include <linux/quotaops.h> 36ac27a0ecSDave Kleikamp #include <linux/seq_file.h> 373197ebdbSTheodore Ts'o #include <linux/ctype.h> 381330593eSVignesh Babu #include <linux/log2.h> 39717d50e4SAndreas Dilger #include <linux/crc16.h> 407abc52c2SDan Magenheimer #include <linux/cleancache.h> 41ac27a0ecSDave Kleikamp #include <asm/uaccess.h> 42ac27a0ecSDave Kleikamp 43bfff6873SLukas Czerner #include <linux/kthread.h> 44bfff6873SLukas Czerner #include <linux/freezer.h> 45bfff6873SLukas Czerner 463dcf5451SChristoph Hellwig #include "ext4.h" 474a092d73STheodore Ts'o #include "ext4_extents.h" /* Needed for trace points definition */ 483dcf5451SChristoph Hellwig #include "ext4_jbd2.h" 49ac27a0ecSDave Kleikamp #include "xattr.h" 50ac27a0ecSDave Kleikamp #include "acl.h" 513661d286STheodore Ts'o #include "mballoc.h" 52ac27a0ecSDave Kleikamp 539bffad1eSTheodore Ts'o #define CREATE_TRACE_POINTS 549bffad1eSTheodore Ts'o #include <trace/events/ext4.h> 559bffad1eSTheodore Ts'o 560b75a840SLukas Czerner static struct ext4_lazy_init *ext4_li_info; 570b75a840SLukas Czerner static struct mutex ext4_li_mtx; 58e294a537STheodore Ts'o static struct ratelimit_state ext4_mount_msg_ratelimit; 599f6200bbSTheodore Ts'o 60617ba13bSMingming Cao static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 61ac27a0ecSDave Kleikamp unsigned long journal_devnum); 622adf6da8STheodore Ts'o static int ext4_show_options(struct seq_file *seq, struct dentry *root); 63e2d67052STheodore Ts'o static int ext4_commit_super(struct super_block *sb, int sync); 64617ba13bSMingming Cao static void ext4_mark_recovery_complete(struct super_block *sb, 65617ba13bSMingming Cao struct ext4_super_block *es); 66617ba13bSMingming Cao static void ext4_clear_journal_err(struct super_block *sb, 67617ba13bSMingming Cao struct ext4_super_block *es); 68617ba13bSMingming Cao static int ext4_sync_fs(struct super_block *sb, int wait); 69617ba13bSMingming Cao static int ext4_remount(struct super_block *sb, int *flags, char *data); 70617ba13bSMingming Cao static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 71c4be0c1dSTakashi Sato static int ext4_unfreeze(struct super_block *sb); 72c4be0c1dSTakashi Sato static int ext4_freeze(struct super_block *sb); 73152a0836SAl Viro static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 74152a0836SAl Viro const char *dev_name, void *data); 752035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb); 762035e776STheodore Ts'o static inline int ext3_feature_set_ok(struct super_block *sb); 77d39195c3SAmir Goldstein static int ext4_feature_set_ok(struct super_block *sb, int readonly); 78bfff6873SLukas Czerner static void ext4_destroy_lazyinit_thread(void); 79bfff6873SLukas Czerner static void ext4_unregister_li_request(struct super_block *sb); 808f1f7453SEric Sandeen static void ext4_clear_request_list(void); 81ac27a0ecSDave Kleikamp 82e74031fdSJan Kara /* 83e74031fdSJan Kara * Lock ordering 84e74031fdSJan Kara * 85e74031fdSJan Kara * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and 86e74031fdSJan Kara * i_mmap_rwsem (inode->i_mmap_rwsem)! 87e74031fdSJan Kara * 88e74031fdSJan Kara * page fault path: 89e74031fdSJan Kara * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start -> 90e74031fdSJan Kara * page lock -> i_data_sem (rw) 91e74031fdSJan Kara * 92e74031fdSJan Kara * buffered write path: 93e74031fdSJan Kara * sb_start_write -> i_mutex -> mmap_sem 94e74031fdSJan Kara * sb_start_write -> i_mutex -> transaction start -> page lock -> 95e74031fdSJan Kara * i_data_sem (rw) 96e74031fdSJan Kara * 97e74031fdSJan Kara * truncate: 98e74031fdSJan Kara * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) -> 99e74031fdSJan Kara * i_mmap_rwsem (w) -> page lock 100e74031fdSJan Kara * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) -> 101e74031fdSJan Kara * transaction start -> i_data_sem (rw) 102e74031fdSJan Kara * 103e74031fdSJan Kara * direct IO: 104e74031fdSJan Kara * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem 105e74031fdSJan Kara * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> 106e74031fdSJan Kara * transaction start -> i_data_sem (rw) 107e74031fdSJan Kara * 108e74031fdSJan Kara * writepages: 109e74031fdSJan Kara * transaction start -> page lock(s) -> i_data_sem (rw) 110e74031fdSJan Kara */ 111e74031fdSJan Kara 112c290ea01SJan Kara #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 1132035e776STheodore Ts'o static struct file_system_type ext2_fs_type = { 1142035e776STheodore Ts'o .owner = THIS_MODULE, 1152035e776STheodore Ts'o .name = "ext2", 1162035e776STheodore Ts'o .mount = ext4_mount, 1172035e776STheodore Ts'o .kill_sb = kill_block_super, 1182035e776STheodore Ts'o .fs_flags = FS_REQUIRES_DEV, 1192035e776STheodore Ts'o }; 1207f78e035SEric W. Biederman MODULE_ALIAS_FS("ext2"); 121fa7614ddSEric W. Biederman MODULE_ALIAS("ext2"); 1222035e776STheodore Ts'o #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 1232035e776STheodore Ts'o #else 1242035e776STheodore Ts'o #define IS_EXT2_SB(sb) (0) 1252035e776STheodore Ts'o #endif 1262035e776STheodore Ts'o 1272035e776STheodore Ts'o 128ba69f9abSJan Kara static struct file_system_type ext3_fs_type = { 129ba69f9abSJan Kara .owner = THIS_MODULE, 130ba69f9abSJan Kara .name = "ext3", 131152a0836SAl Viro .mount = ext4_mount, 132ba69f9abSJan Kara .kill_sb = kill_block_super, 133ba69f9abSJan Kara .fs_flags = FS_REQUIRES_DEV, 134ba69f9abSJan Kara }; 1357f78e035SEric W. Biederman MODULE_ALIAS_FS("ext3"); 136fa7614ddSEric W. Biederman MODULE_ALIAS("ext3"); 137ba69f9abSJan Kara #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 138bd81d8eeSLaurent Vivier 139d25425f8SDarrick J. Wong static int ext4_verify_csum_type(struct super_block *sb, 140d25425f8SDarrick J. Wong struct ext4_super_block *es) 141d25425f8SDarrick J. Wong { 142e2b911c5SDarrick J. Wong if (!ext4_has_feature_metadata_csum(sb)) 143d25425f8SDarrick J. Wong return 1; 144d25425f8SDarrick J. Wong 145d25425f8SDarrick J. Wong return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 146d25425f8SDarrick J. Wong } 147d25425f8SDarrick J. Wong 148a9c47317SDarrick J. Wong static __le32 ext4_superblock_csum(struct super_block *sb, 149a9c47317SDarrick J. Wong struct ext4_super_block *es) 150a9c47317SDarrick J. Wong { 151a9c47317SDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 152a9c47317SDarrick J. Wong int offset = offsetof(struct ext4_super_block, s_checksum); 153a9c47317SDarrick J. Wong __u32 csum; 154a9c47317SDarrick J. Wong 155a9c47317SDarrick J. Wong csum = ext4_chksum(sbi, ~0, (char *)es, offset); 156a9c47317SDarrick J. Wong 157a9c47317SDarrick J. Wong return cpu_to_le32(csum); 158a9c47317SDarrick J. Wong } 159a9c47317SDarrick J. Wong 160c197855eSStephen Hemminger static int ext4_superblock_csum_verify(struct super_block *sb, 161a9c47317SDarrick J. Wong struct ext4_super_block *es) 162a9c47317SDarrick J. Wong { 1639aa5d32bSDmitry Monakhov if (!ext4_has_metadata_csum(sb)) 164a9c47317SDarrick J. Wong return 1; 165a9c47317SDarrick J. Wong 166a9c47317SDarrick J. Wong return es->s_checksum == ext4_superblock_csum(sb, es); 167a9c47317SDarrick J. Wong } 168a9c47317SDarrick J. Wong 16906db49e6STheodore Ts'o void ext4_superblock_csum_set(struct super_block *sb) 170a9c47317SDarrick J. Wong { 17106db49e6STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 17206db49e6STheodore Ts'o 1739aa5d32bSDmitry Monakhov if (!ext4_has_metadata_csum(sb)) 174a9c47317SDarrick J. Wong return; 175a9c47317SDarrick J. Wong 176a9c47317SDarrick J. Wong es->s_checksum = ext4_superblock_csum(sb, es); 177a9c47317SDarrick J. Wong } 178a9c47317SDarrick J. Wong 1799933fc0aSTheodore Ts'o void *ext4_kvmalloc(size_t size, gfp_t flags) 1809933fc0aSTheodore Ts'o { 1819933fc0aSTheodore Ts'o void *ret; 1829933fc0aSTheodore Ts'o 1838be04b93SJoe Perches ret = kmalloc(size, flags | __GFP_NOWARN); 1849933fc0aSTheodore Ts'o if (!ret) 1859933fc0aSTheodore Ts'o ret = __vmalloc(size, flags, PAGE_KERNEL); 1869933fc0aSTheodore Ts'o return ret; 1879933fc0aSTheodore Ts'o } 1889933fc0aSTheodore Ts'o 1899933fc0aSTheodore Ts'o void *ext4_kvzalloc(size_t size, gfp_t flags) 1909933fc0aSTheodore Ts'o { 1919933fc0aSTheodore Ts'o void *ret; 1929933fc0aSTheodore Ts'o 1938be04b93SJoe Perches ret = kzalloc(size, flags | __GFP_NOWARN); 1949933fc0aSTheodore Ts'o if (!ret) 1959933fc0aSTheodore Ts'o ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL); 1969933fc0aSTheodore Ts'o return ret; 1979933fc0aSTheodore Ts'o } 1989933fc0aSTheodore Ts'o 1998fadc143SAlexandre Ratchov ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 2008fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 201bd81d8eeSLaurent Vivier { 2023a14589cSAneesh Kumar K.V return le32_to_cpu(bg->bg_block_bitmap_lo) | 2038fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 2048fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 205bd81d8eeSLaurent Vivier } 206bd81d8eeSLaurent Vivier 2078fadc143SAlexandre Ratchov ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 2088fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 209bd81d8eeSLaurent Vivier { 2105272f837SAneesh Kumar K.V return le32_to_cpu(bg->bg_inode_bitmap_lo) | 2118fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 2128fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 213bd81d8eeSLaurent Vivier } 214bd81d8eeSLaurent Vivier 2158fadc143SAlexandre Ratchov ext4_fsblk_t ext4_inode_table(struct super_block *sb, 2168fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 217bd81d8eeSLaurent Vivier { 2185272f837SAneesh Kumar K.V return le32_to_cpu(bg->bg_inode_table_lo) | 2198fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 2208fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 221bd81d8eeSLaurent Vivier } 222bd81d8eeSLaurent Vivier 223021b65bbSTheodore Ts'o __u32 ext4_free_group_clusters(struct super_block *sb, 224560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 225560671a0SAneesh Kumar K.V { 226560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_free_blocks_count_lo) | 227560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 228560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 229560671a0SAneesh Kumar K.V } 230560671a0SAneesh Kumar K.V 231560671a0SAneesh Kumar K.V __u32 ext4_free_inodes_count(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_inodes_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_inodes_count_hi) << 16 : 0); 237560671a0SAneesh Kumar K.V } 238560671a0SAneesh Kumar K.V 239560671a0SAneesh Kumar K.V __u32 ext4_used_dirs_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_used_dirs_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_used_dirs_count_hi) << 16 : 0); 245560671a0SAneesh Kumar K.V } 246560671a0SAneesh Kumar K.V 247560671a0SAneesh Kumar K.V __u32 ext4_itable_unused_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_itable_unused_lo) | 251560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 252560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 253560671a0SAneesh Kumar K.V } 254560671a0SAneesh Kumar K.V 2558fadc143SAlexandre Ratchov void ext4_block_bitmap_set(struct super_block *sb, 2568fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 257bd81d8eeSLaurent Vivier { 2583a14589cSAneesh Kumar K.V bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 2598fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 2608fadc143SAlexandre Ratchov bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 261bd81d8eeSLaurent Vivier } 262bd81d8eeSLaurent Vivier 2638fadc143SAlexandre Ratchov void ext4_inode_bitmap_set(struct super_block *sb, 2648fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 265bd81d8eeSLaurent Vivier { 2665272f837SAneesh Kumar K.V bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 2678fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 2688fadc143SAlexandre Ratchov bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 269bd81d8eeSLaurent Vivier } 270bd81d8eeSLaurent Vivier 2718fadc143SAlexandre Ratchov void ext4_inode_table_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_table_lo = cpu_to_le32((u32)blk); 2758fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 2768fadc143SAlexandre Ratchov bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 277bd81d8eeSLaurent Vivier } 278bd81d8eeSLaurent Vivier 279021b65bbSTheodore Ts'o void ext4_free_group_clusters_set(struct super_block *sb, 280560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 281560671a0SAneesh Kumar K.V { 282560671a0SAneesh Kumar K.V bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 283560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 284560671a0SAneesh Kumar K.V bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 285560671a0SAneesh Kumar K.V } 286560671a0SAneesh Kumar K.V 287560671a0SAneesh Kumar K.V void ext4_free_inodes_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_inodes_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_inodes_count_hi = cpu_to_le16(count >> 16); 293560671a0SAneesh Kumar K.V } 294560671a0SAneesh Kumar K.V 295560671a0SAneesh Kumar K.V void ext4_used_dirs_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_used_dirs_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_used_dirs_count_hi = cpu_to_le16(count >> 16); 301560671a0SAneesh Kumar K.V } 302560671a0SAneesh Kumar K.V 303560671a0SAneesh Kumar K.V void ext4_itable_unused_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_itable_unused_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_itable_unused_hi = cpu_to_le16(count >> 16); 309560671a0SAneesh Kumar K.V } 310560671a0SAneesh Kumar K.V 311d3d1faf6SCurt Wohlgemuth 3121c13d5c0STheodore Ts'o static void __save_error_info(struct super_block *sb, const char *func, 3131c13d5c0STheodore Ts'o unsigned int line) 3141c13d5c0STheodore Ts'o { 3151c13d5c0STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 3161c13d5c0STheodore Ts'o 3171c13d5c0STheodore Ts'o EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 3181b46617bSTheodore Ts'o if (bdev_read_only(sb->s_bdev)) 3191b46617bSTheodore Ts'o return; 3201c13d5c0STheodore Ts'o es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 3211c13d5c0STheodore Ts'o es->s_last_error_time = cpu_to_le32(get_seconds()); 3221c13d5c0STheodore Ts'o strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); 3231c13d5c0STheodore Ts'o es->s_last_error_line = cpu_to_le32(line); 3241c13d5c0STheodore Ts'o if (!es->s_first_error_time) { 3251c13d5c0STheodore Ts'o es->s_first_error_time = es->s_last_error_time; 3261c13d5c0STheodore Ts'o strncpy(es->s_first_error_func, func, 3271c13d5c0STheodore Ts'o sizeof(es->s_first_error_func)); 3281c13d5c0STheodore Ts'o es->s_first_error_line = cpu_to_le32(line); 3291c13d5c0STheodore Ts'o es->s_first_error_ino = es->s_last_error_ino; 3301c13d5c0STheodore Ts'o es->s_first_error_block = es->s_last_error_block; 3311c13d5c0STheodore Ts'o } 33266e61a9eSTheodore Ts'o /* 33366e61a9eSTheodore Ts'o * Start the daily error reporting function if it hasn't been 33466e61a9eSTheodore Ts'o * started already 33566e61a9eSTheodore Ts'o */ 33666e61a9eSTheodore Ts'o if (!es->s_error_count) 33766e61a9eSTheodore Ts'o mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); 338ba39ebb6SWei Yongjun le32_add_cpu(&es->s_error_count, 1); 3391c13d5c0STheodore Ts'o } 3401c13d5c0STheodore Ts'o 3411c13d5c0STheodore Ts'o static void save_error_info(struct super_block *sb, const char *func, 3421c13d5c0STheodore Ts'o unsigned int line) 3431c13d5c0STheodore Ts'o { 3441c13d5c0STheodore Ts'o __save_error_info(sb, func, line); 3451c13d5c0STheodore Ts'o ext4_commit_super(sb, 1); 3461c13d5c0STheodore Ts'o } 3471c13d5c0STheodore Ts'o 348bdfe0cbdSTheodore Ts'o /* 349bdfe0cbdSTheodore Ts'o * The del_gendisk() function uninitializes the disk-specific data 350bdfe0cbdSTheodore Ts'o * structures, including the bdi structure, without telling anyone 351bdfe0cbdSTheodore Ts'o * else. Once this happens, any attempt to call mark_buffer_dirty() 352bdfe0cbdSTheodore Ts'o * (for example, by ext4_commit_super), will cause a kernel OOPS. 353bdfe0cbdSTheodore Ts'o * This is a kludge to prevent these oops until we can put in a proper 354bdfe0cbdSTheodore Ts'o * hook in del_gendisk() to inform the VFS and file system layers. 355bdfe0cbdSTheodore Ts'o */ 356bdfe0cbdSTheodore Ts'o static int block_device_ejected(struct super_block *sb) 357bdfe0cbdSTheodore Ts'o { 358bdfe0cbdSTheodore Ts'o struct inode *bd_inode = sb->s_bdev->bd_inode; 359bdfe0cbdSTheodore Ts'o struct backing_dev_info *bdi = inode_to_bdi(bd_inode); 360bdfe0cbdSTheodore Ts'o 361bdfe0cbdSTheodore Ts'o return bdi->dev == NULL; 362bdfe0cbdSTheodore Ts'o } 363bdfe0cbdSTheodore Ts'o 36418aadd47SBobi Jam static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 36518aadd47SBobi Jam { 36618aadd47SBobi Jam struct super_block *sb = journal->j_private; 36718aadd47SBobi Jam struct ext4_sb_info *sbi = EXT4_SB(sb); 36818aadd47SBobi Jam int error = is_journal_aborted(journal); 3695d3ee208SDmitry Monakhov struct ext4_journal_cb_entry *jce; 37018aadd47SBobi Jam 3715d3ee208SDmitry Monakhov BUG_ON(txn->t_state == T_FINISHED); 37218aadd47SBobi Jam spin_lock(&sbi->s_md_lock); 3735d3ee208SDmitry Monakhov while (!list_empty(&txn->t_private_list)) { 3745d3ee208SDmitry Monakhov jce = list_entry(txn->t_private_list.next, 3755d3ee208SDmitry Monakhov struct ext4_journal_cb_entry, jce_list); 37618aadd47SBobi Jam list_del_init(&jce->jce_list); 37718aadd47SBobi Jam spin_unlock(&sbi->s_md_lock); 37818aadd47SBobi Jam jce->jce_func(sb, jce, error); 37918aadd47SBobi Jam spin_lock(&sbi->s_md_lock); 38018aadd47SBobi Jam } 38118aadd47SBobi Jam spin_unlock(&sbi->s_md_lock); 38218aadd47SBobi Jam } 3831c13d5c0STheodore Ts'o 384ac27a0ecSDave Kleikamp /* Deal with the reporting of failure conditions on a filesystem such as 385ac27a0ecSDave Kleikamp * inconsistencies detected or read IO failures. 386ac27a0ecSDave Kleikamp * 387ac27a0ecSDave Kleikamp * On ext2, we can store the error state of the filesystem in the 388617ba13bSMingming Cao * superblock. That is not possible on ext4, because we may have other 389ac27a0ecSDave Kleikamp * write ordering constraints on the superblock which prevent us from 390ac27a0ecSDave Kleikamp * writing it out straight away; and given that the journal is about to 391ac27a0ecSDave Kleikamp * be aborted, we can't rely on the current, or future, transactions to 392ac27a0ecSDave Kleikamp * write out the superblock safely. 393ac27a0ecSDave Kleikamp * 394dab291afSMingming Cao * We'll just use the jbd2_journal_abort() error code to record an error in 395d6b198bcSThadeu Lima de Souza Cascardo * the journal instead. On recovery, the journal will complain about 396ac27a0ecSDave Kleikamp * that error until we've noted it down and cleared it. 397ac27a0ecSDave Kleikamp */ 398ac27a0ecSDave Kleikamp 399617ba13bSMingming Cao static void ext4_handle_error(struct super_block *sb) 400ac27a0ecSDave Kleikamp { 401ac27a0ecSDave Kleikamp if (sb->s_flags & MS_RDONLY) 402ac27a0ecSDave Kleikamp return; 403ac27a0ecSDave Kleikamp 404ac27a0ecSDave Kleikamp if (!test_opt(sb, ERRORS_CONT)) { 405617ba13bSMingming Cao journal_t *journal = EXT4_SB(sb)->s_journal; 406ac27a0ecSDave Kleikamp 4074ab2f15bSTheodore Ts'o EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 408ac27a0ecSDave Kleikamp if (journal) 409dab291afSMingming Cao jbd2_journal_abort(journal, -EIO); 410ac27a0ecSDave Kleikamp } 411ac27a0ecSDave Kleikamp if (test_opt(sb, ERRORS_RO)) { 412b31e1552SEric Sandeen ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 4134418e141SDmitry Monakhov /* 4144418e141SDmitry Monakhov * Make sure updated value of ->s_mount_flags will be visible 4154418e141SDmitry Monakhov * before ->s_flags update 4164418e141SDmitry Monakhov */ 4174418e141SDmitry Monakhov smp_wmb(); 418ac27a0ecSDave Kleikamp sb->s_flags |= MS_RDONLY; 419ac27a0ecSDave Kleikamp } 4204327ba52SDaeho Jeong if (test_opt(sb, ERRORS_PANIC)) { 4214327ba52SDaeho Jeong if (EXT4_SB(sb)->s_journal && 4224327ba52SDaeho Jeong !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) 4234327ba52SDaeho Jeong return; 424617ba13bSMingming Cao panic("EXT4-fs (device %s): panic forced after error\n", 425ac27a0ecSDave Kleikamp sb->s_id); 426ac27a0ecSDave Kleikamp } 4274327ba52SDaeho Jeong } 428ac27a0ecSDave Kleikamp 429efbed4dcSTheodore Ts'o #define ext4_error_ratelimit(sb) \ 430efbed4dcSTheodore Ts'o ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 431efbed4dcSTheodore Ts'o "EXT4-fs error") 432efbed4dcSTheodore Ts'o 43312062dddSEric Sandeen void __ext4_error(struct super_block *sb, const char *function, 434c398eda0STheodore Ts'o unsigned int line, const char *fmt, ...) 435ac27a0ecSDave Kleikamp { 4360ff2ea7dSJoe Perches struct va_format vaf; 437ac27a0ecSDave Kleikamp va_list args; 438ac27a0ecSDave Kleikamp 439efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 440ac27a0ecSDave Kleikamp va_start(args, fmt); 4410ff2ea7dSJoe Perches vaf.fmt = fmt; 4420ff2ea7dSJoe Perches vaf.va = &args; 443efbed4dcSTheodore Ts'o printk(KERN_CRIT 444efbed4dcSTheodore Ts'o "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 4450ff2ea7dSJoe Perches sb->s_id, function, line, current->comm, &vaf); 446ac27a0ecSDave Kleikamp va_end(args); 447efbed4dcSTheodore Ts'o } 448f3fc0210STheodore Ts'o save_error_info(sb, function, line); 449617ba13bSMingming Cao ext4_handle_error(sb); 450ac27a0ecSDave Kleikamp } 451ac27a0ecSDave Kleikamp 452e7c96e8eSJoe Perches void __ext4_error_inode(struct inode *inode, const char *function, 453c398eda0STheodore Ts'o unsigned int line, ext4_fsblk_t block, 454273df556SFrank Mayhar const char *fmt, ...) 455273df556SFrank Mayhar { 456273df556SFrank Mayhar va_list args; 457f7c21177STheodore Ts'o struct va_format vaf; 4581c13d5c0STheodore Ts'o struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; 459273df556SFrank Mayhar 4601c13d5c0STheodore Ts'o es->s_last_error_ino = cpu_to_le32(inode->i_ino); 4611c13d5c0STheodore Ts'o es->s_last_error_block = cpu_to_le64(block); 462efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(inode->i_sb)) { 463273df556SFrank Mayhar va_start(args, fmt); 464f7c21177STheodore Ts'o vaf.fmt = fmt; 465f7c21177STheodore Ts'o vaf.va = &args; 466c398eda0STheodore Ts'o if (block) 467d9ee81daSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 468d9ee81daSJoe Perches "inode #%lu: block %llu: comm %s: %pV\n", 469d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 470d9ee81daSJoe Perches block, current->comm, &vaf); 471d9ee81daSJoe Perches else 472d9ee81daSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 473d9ee81daSJoe Perches "inode #%lu: comm %s: %pV\n", 474d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 475d9ee81daSJoe Perches current->comm, &vaf); 476273df556SFrank Mayhar va_end(args); 477efbed4dcSTheodore Ts'o } 478efbed4dcSTheodore Ts'o save_error_info(inode->i_sb, function, line); 479273df556SFrank Mayhar ext4_handle_error(inode->i_sb); 480273df556SFrank Mayhar } 481273df556SFrank Mayhar 482e7c96e8eSJoe Perches void __ext4_error_file(struct file *file, const char *function, 483f7c21177STheodore Ts'o unsigned int line, ext4_fsblk_t block, 484f7c21177STheodore Ts'o const char *fmt, ...) 485273df556SFrank Mayhar { 486273df556SFrank Mayhar va_list args; 487f7c21177STheodore Ts'o struct va_format vaf; 4881c13d5c0STheodore Ts'o struct ext4_super_block *es; 489496ad9aaSAl Viro struct inode *inode = file_inode(file); 490273df556SFrank Mayhar char pathname[80], *path; 491273df556SFrank Mayhar 4921c13d5c0STheodore Ts'o es = EXT4_SB(inode->i_sb)->s_es; 4931c13d5c0STheodore Ts'o es->s_last_error_ino = cpu_to_le32(inode->i_ino); 494efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(inode->i_sb)) { 4959bf39ab2SMiklos Szeredi path = file_path(file, pathname, sizeof(pathname)); 496f9a62d09SDan Carpenter if (IS_ERR(path)) 497273df556SFrank Mayhar path = "(unknown)"; 498f7c21177STheodore Ts'o va_start(args, fmt); 499f7c21177STheodore Ts'o vaf.fmt = fmt; 500f7c21177STheodore Ts'o vaf.va = &args; 501d9ee81daSJoe Perches if (block) 502d9ee81daSJoe Perches printk(KERN_CRIT 503d9ee81daSJoe Perches "EXT4-fs error (device %s): %s:%d: inode #%lu: " 504d9ee81daSJoe Perches "block %llu: comm %s: path %s: %pV\n", 505d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 506d9ee81daSJoe Perches block, current->comm, path, &vaf); 507d9ee81daSJoe Perches else 508d9ee81daSJoe Perches printk(KERN_CRIT 509d9ee81daSJoe Perches "EXT4-fs error (device %s): %s:%d: inode #%lu: " 510d9ee81daSJoe Perches "comm %s: path %s: %pV\n", 511d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 512d9ee81daSJoe Perches current->comm, path, &vaf); 513273df556SFrank Mayhar va_end(args); 514efbed4dcSTheodore Ts'o } 515efbed4dcSTheodore Ts'o save_error_info(inode->i_sb, function, line); 516273df556SFrank Mayhar ext4_handle_error(inode->i_sb); 517273df556SFrank Mayhar } 518273df556SFrank Mayhar 519722887ddSTheodore Ts'o const char *ext4_decode_error(struct super_block *sb, int errno, 520ac27a0ecSDave Kleikamp char nbuf[16]) 521ac27a0ecSDave Kleikamp { 522ac27a0ecSDave Kleikamp char *errstr = NULL; 523ac27a0ecSDave Kleikamp 524ac27a0ecSDave Kleikamp switch (errno) { 5256a797d27SDarrick J. Wong case -EFSCORRUPTED: 5266a797d27SDarrick J. Wong errstr = "Corrupt filesystem"; 5276a797d27SDarrick J. Wong break; 5286a797d27SDarrick J. Wong case -EFSBADCRC: 5296a797d27SDarrick J. Wong errstr = "Filesystem failed CRC"; 5306a797d27SDarrick J. Wong break; 531ac27a0ecSDave Kleikamp case -EIO: 532ac27a0ecSDave Kleikamp errstr = "IO failure"; 533ac27a0ecSDave Kleikamp break; 534ac27a0ecSDave Kleikamp case -ENOMEM: 535ac27a0ecSDave Kleikamp errstr = "Out of memory"; 536ac27a0ecSDave Kleikamp break; 537ac27a0ecSDave Kleikamp case -EROFS: 53878f1ddbbSTheodore Ts'o if (!sb || (EXT4_SB(sb)->s_journal && 53978f1ddbbSTheodore Ts'o EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 540ac27a0ecSDave Kleikamp errstr = "Journal has aborted"; 541ac27a0ecSDave Kleikamp else 542ac27a0ecSDave Kleikamp errstr = "Readonly filesystem"; 543ac27a0ecSDave Kleikamp break; 544ac27a0ecSDave Kleikamp default: 545ac27a0ecSDave Kleikamp /* If the caller passed in an extra buffer for unknown 546ac27a0ecSDave Kleikamp * errors, textualise them now. Else we just return 547ac27a0ecSDave Kleikamp * NULL. */ 548ac27a0ecSDave Kleikamp if (nbuf) { 549ac27a0ecSDave Kleikamp /* Check for truncated error codes... */ 550ac27a0ecSDave Kleikamp if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 551ac27a0ecSDave Kleikamp errstr = nbuf; 552ac27a0ecSDave Kleikamp } 553ac27a0ecSDave Kleikamp break; 554ac27a0ecSDave Kleikamp } 555ac27a0ecSDave Kleikamp 556ac27a0ecSDave Kleikamp return errstr; 557ac27a0ecSDave Kleikamp } 558ac27a0ecSDave Kleikamp 559617ba13bSMingming Cao /* __ext4_std_error decodes expected errors from journaling functions 560ac27a0ecSDave Kleikamp * automatically and invokes the appropriate error response. */ 561ac27a0ecSDave Kleikamp 562c398eda0STheodore Ts'o void __ext4_std_error(struct super_block *sb, const char *function, 563c398eda0STheodore Ts'o unsigned int line, int errno) 564ac27a0ecSDave Kleikamp { 565ac27a0ecSDave Kleikamp char nbuf[16]; 566ac27a0ecSDave Kleikamp const char *errstr; 567ac27a0ecSDave Kleikamp 568ac27a0ecSDave Kleikamp /* Special case: if the error is EROFS, and we're not already 569ac27a0ecSDave Kleikamp * inside a transaction, then there's really no point in logging 570ac27a0ecSDave Kleikamp * an error. */ 571ac27a0ecSDave Kleikamp if (errno == -EROFS && journal_current_handle() == NULL && 572ac27a0ecSDave Kleikamp (sb->s_flags & MS_RDONLY)) 573ac27a0ecSDave Kleikamp return; 574ac27a0ecSDave Kleikamp 575efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 576617ba13bSMingming Cao errstr = ext4_decode_error(sb, errno, nbuf); 577c398eda0STheodore Ts'o printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 578c398eda0STheodore Ts'o sb->s_id, function, line, errstr); 579efbed4dcSTheodore Ts'o } 580ac27a0ecSDave Kleikamp 581efbed4dcSTheodore Ts'o save_error_info(sb, function, line); 582617ba13bSMingming Cao ext4_handle_error(sb); 583ac27a0ecSDave Kleikamp } 584ac27a0ecSDave Kleikamp 585ac27a0ecSDave Kleikamp /* 586617ba13bSMingming Cao * ext4_abort is a much stronger failure handler than ext4_error. The 587ac27a0ecSDave Kleikamp * abort function may be used to deal with unrecoverable failures such 588ac27a0ecSDave Kleikamp * as journal IO errors or ENOMEM at a critical moment in log management. 589ac27a0ecSDave Kleikamp * 590ac27a0ecSDave Kleikamp * We unconditionally force the filesystem into an ABORT|READONLY state, 591ac27a0ecSDave Kleikamp * unless the error response on the fs has been set to panic in which 592ac27a0ecSDave Kleikamp * case we take the easy way out and panic immediately. 593ac27a0ecSDave Kleikamp */ 594ac27a0ecSDave Kleikamp 595c67d859eSTheodore Ts'o void __ext4_abort(struct super_block *sb, const char *function, 596c398eda0STheodore Ts'o unsigned int line, const char *fmt, ...) 597ac27a0ecSDave Kleikamp { 598ac27a0ecSDave Kleikamp va_list args; 599ac27a0ecSDave Kleikamp 6001c13d5c0STheodore Ts'o save_error_info(sb, function, line); 601ac27a0ecSDave Kleikamp va_start(args, fmt); 602c398eda0STheodore Ts'o printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id, 603c398eda0STheodore Ts'o function, line); 604ac27a0ecSDave Kleikamp vprintk(fmt, args); 605ac27a0ecSDave Kleikamp printk("\n"); 606ac27a0ecSDave Kleikamp va_end(args); 607ac27a0ecSDave Kleikamp 6081c13d5c0STheodore Ts'o if ((sb->s_flags & MS_RDONLY) == 0) { 609b31e1552SEric Sandeen ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 6104ab2f15bSTheodore Ts'o EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 6114418e141SDmitry Monakhov /* 6124418e141SDmitry Monakhov * Make sure updated value of ->s_mount_flags will be visible 6134418e141SDmitry Monakhov * before ->s_flags update 6144418e141SDmitry Monakhov */ 6154418e141SDmitry Monakhov smp_wmb(); 6164418e141SDmitry Monakhov sb->s_flags |= MS_RDONLY; 617ef2cabf7SHidehiro Kawai if (EXT4_SB(sb)->s_journal) 618dab291afSMingming Cao jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); 6191c13d5c0STheodore Ts'o save_error_info(sb, function, line); 6201c13d5c0STheodore Ts'o } 6214327ba52SDaeho Jeong if (test_opt(sb, ERRORS_PANIC)) { 6224327ba52SDaeho Jeong if (EXT4_SB(sb)->s_journal && 6234327ba52SDaeho Jeong !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR)) 6244327ba52SDaeho Jeong return; 6251c13d5c0STheodore Ts'o panic("EXT4-fs panic from previous error\n"); 626ac27a0ecSDave Kleikamp } 6274327ba52SDaeho Jeong } 628ac27a0ecSDave Kleikamp 629e7c96e8eSJoe Perches void __ext4_msg(struct super_block *sb, 630e7c96e8eSJoe Perches const char *prefix, const char *fmt, ...) 631b31e1552SEric Sandeen { 6320ff2ea7dSJoe Perches struct va_format vaf; 633b31e1552SEric Sandeen va_list args; 634b31e1552SEric Sandeen 635efbed4dcSTheodore Ts'o if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs")) 636efbed4dcSTheodore Ts'o return; 637efbed4dcSTheodore Ts'o 638b31e1552SEric Sandeen va_start(args, fmt); 6390ff2ea7dSJoe Perches vaf.fmt = fmt; 6400ff2ea7dSJoe Perches vaf.va = &args; 6410ff2ea7dSJoe Perches printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 642b31e1552SEric Sandeen va_end(args); 643b31e1552SEric Sandeen } 644b31e1552SEric Sandeen 645b03a2f7eSAndreas Dilger #define ext4_warning_ratelimit(sb) \ 646b03a2f7eSAndreas Dilger ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \ 647b03a2f7eSAndreas Dilger "EXT4-fs warning") 648b03a2f7eSAndreas Dilger 64912062dddSEric Sandeen void __ext4_warning(struct super_block *sb, const char *function, 650c398eda0STheodore Ts'o unsigned int line, const char *fmt, ...) 651ac27a0ecSDave Kleikamp { 6520ff2ea7dSJoe Perches struct va_format vaf; 653ac27a0ecSDave Kleikamp va_list args; 654ac27a0ecSDave Kleikamp 655b03a2f7eSAndreas Dilger if (!ext4_warning_ratelimit(sb)) 656efbed4dcSTheodore Ts'o return; 657efbed4dcSTheodore Ts'o 658ac27a0ecSDave Kleikamp va_start(args, fmt); 6590ff2ea7dSJoe Perches vaf.fmt = fmt; 6600ff2ea7dSJoe Perches vaf.va = &args; 6610ff2ea7dSJoe Perches printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 6620ff2ea7dSJoe Perches sb->s_id, function, line, &vaf); 663ac27a0ecSDave Kleikamp va_end(args); 664ac27a0ecSDave Kleikamp } 665ac27a0ecSDave Kleikamp 666b03a2f7eSAndreas Dilger void __ext4_warning_inode(const struct inode *inode, const char *function, 667b03a2f7eSAndreas Dilger unsigned int line, const char *fmt, ...) 668b03a2f7eSAndreas Dilger { 669b03a2f7eSAndreas Dilger struct va_format vaf; 670b03a2f7eSAndreas Dilger va_list args; 671b03a2f7eSAndreas Dilger 672b03a2f7eSAndreas Dilger if (!ext4_warning_ratelimit(inode->i_sb)) 673b03a2f7eSAndreas Dilger return; 674b03a2f7eSAndreas Dilger 675b03a2f7eSAndreas Dilger va_start(args, fmt); 676b03a2f7eSAndreas Dilger vaf.fmt = fmt; 677b03a2f7eSAndreas Dilger vaf.va = &args; 678b03a2f7eSAndreas Dilger printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " 679b03a2f7eSAndreas Dilger "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, 680b03a2f7eSAndreas Dilger function, line, inode->i_ino, current->comm, &vaf); 681b03a2f7eSAndreas Dilger va_end(args); 682b03a2f7eSAndreas Dilger } 683b03a2f7eSAndreas Dilger 684e29136f8STheodore Ts'o void __ext4_grp_locked_error(const char *function, unsigned int line, 685e29136f8STheodore Ts'o struct super_block *sb, ext4_group_t grp, 686e29136f8STheodore Ts'o unsigned long ino, ext4_fsblk_t block, 687e29136f8STheodore Ts'o const char *fmt, ...) 6885d1b1b3fSAneesh Kumar K.V __releases(bitlock) 6895d1b1b3fSAneesh Kumar K.V __acquires(bitlock) 6905d1b1b3fSAneesh Kumar K.V { 6910ff2ea7dSJoe Perches struct va_format vaf; 6925d1b1b3fSAneesh Kumar K.V va_list args; 6935d1b1b3fSAneesh Kumar K.V struct ext4_super_block *es = EXT4_SB(sb)->s_es; 6945d1b1b3fSAneesh Kumar K.V 6951c13d5c0STheodore Ts'o es->s_last_error_ino = cpu_to_le32(ino); 6961c13d5c0STheodore Ts'o es->s_last_error_block = cpu_to_le64(block); 6971c13d5c0STheodore Ts'o __save_error_info(sb, function, line); 6980ff2ea7dSJoe Perches 699efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 7005d1b1b3fSAneesh Kumar K.V va_start(args, fmt); 7010ff2ea7dSJoe Perches vaf.fmt = fmt; 7020ff2ea7dSJoe Perches vaf.va = &args; 70321149d61SRobin Dong printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 704e29136f8STheodore Ts'o sb->s_id, function, line, grp); 705e29136f8STheodore Ts'o if (ino) 7060ff2ea7dSJoe Perches printk(KERN_CONT "inode %lu: ", ino); 707e29136f8STheodore Ts'o if (block) 708efbed4dcSTheodore Ts'o printk(KERN_CONT "block %llu:", 709efbed4dcSTheodore Ts'o (unsigned long long) block); 7100ff2ea7dSJoe Perches printk(KERN_CONT "%pV\n", &vaf); 7115d1b1b3fSAneesh Kumar K.V va_end(args); 712efbed4dcSTheodore Ts'o } 7135d1b1b3fSAneesh Kumar K.V 7145d1b1b3fSAneesh Kumar K.V if (test_opt(sb, ERRORS_CONT)) { 715e2d67052STheodore Ts'o ext4_commit_super(sb, 0); 7165d1b1b3fSAneesh Kumar K.V return; 7175d1b1b3fSAneesh Kumar K.V } 7181c13d5c0STheodore Ts'o 7195d1b1b3fSAneesh Kumar K.V ext4_unlock_group(sb, grp); 7205d1b1b3fSAneesh Kumar K.V ext4_handle_error(sb); 7215d1b1b3fSAneesh Kumar K.V /* 7225d1b1b3fSAneesh Kumar K.V * We only get here in the ERRORS_RO case; relocking the group 7235d1b1b3fSAneesh Kumar K.V * may be dangerous, but nothing bad will happen since the 7245d1b1b3fSAneesh Kumar K.V * filesystem will have already been marked read/only and the 7255d1b1b3fSAneesh Kumar K.V * journal has been aborted. We return 1 as a hint to callers 7265d1b1b3fSAneesh Kumar K.V * who might what to use the return value from 72725985edcSLucas De Marchi * ext4_grp_locked_error() to distinguish between the 7285d1b1b3fSAneesh Kumar K.V * ERRORS_CONT and ERRORS_RO case, and perhaps return more 7295d1b1b3fSAneesh Kumar K.V * aggressively from the ext4 function in question, with a 7305d1b1b3fSAneesh Kumar K.V * more appropriate error code. 7315d1b1b3fSAneesh Kumar K.V */ 7325d1b1b3fSAneesh Kumar K.V ext4_lock_group(sb, grp); 7335d1b1b3fSAneesh Kumar K.V return; 7345d1b1b3fSAneesh Kumar K.V } 7355d1b1b3fSAneesh Kumar K.V 736617ba13bSMingming Cao void ext4_update_dynamic_rev(struct super_block *sb) 737ac27a0ecSDave Kleikamp { 738617ba13bSMingming Cao struct ext4_super_block *es = EXT4_SB(sb)->s_es; 739ac27a0ecSDave Kleikamp 740617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 741ac27a0ecSDave Kleikamp return; 742ac27a0ecSDave Kleikamp 74312062dddSEric Sandeen ext4_warning(sb, 744ac27a0ecSDave Kleikamp "updating to rev %d because of new feature flag, " 745ac27a0ecSDave Kleikamp "running e2fsck is recommended", 746617ba13bSMingming Cao EXT4_DYNAMIC_REV); 747ac27a0ecSDave Kleikamp 748617ba13bSMingming Cao es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 749617ba13bSMingming Cao es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 750617ba13bSMingming Cao es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 751ac27a0ecSDave Kleikamp /* leave es->s_feature_*compat flags alone */ 752ac27a0ecSDave Kleikamp /* es->s_uuid will be set by e2fsck if empty */ 753ac27a0ecSDave Kleikamp 754ac27a0ecSDave Kleikamp /* 755ac27a0ecSDave Kleikamp * The rest of the superblock fields should be zero, and if not it 756ac27a0ecSDave Kleikamp * means they are likely already in use, so leave them alone. We 757ac27a0ecSDave Kleikamp * can leave it up to e2fsck to clean up any inconsistencies there. 758ac27a0ecSDave Kleikamp */ 759ac27a0ecSDave Kleikamp } 760ac27a0ecSDave Kleikamp 761ac27a0ecSDave Kleikamp /* 762ac27a0ecSDave Kleikamp * Open the external journal device 763ac27a0ecSDave Kleikamp */ 764b31e1552SEric Sandeen static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 765ac27a0ecSDave Kleikamp { 766ac27a0ecSDave Kleikamp struct block_device *bdev; 767ac27a0ecSDave Kleikamp char b[BDEVNAME_SIZE]; 768ac27a0ecSDave Kleikamp 769d4d77629STejun Heo bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 770ac27a0ecSDave Kleikamp if (IS_ERR(bdev)) 771ac27a0ecSDave Kleikamp goto fail; 772ac27a0ecSDave Kleikamp return bdev; 773ac27a0ecSDave Kleikamp 774ac27a0ecSDave Kleikamp fail: 775b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", 776ac27a0ecSDave Kleikamp __bdevname(dev, b), PTR_ERR(bdev)); 777ac27a0ecSDave Kleikamp return NULL; 778ac27a0ecSDave Kleikamp } 779ac27a0ecSDave Kleikamp 780ac27a0ecSDave Kleikamp /* 781ac27a0ecSDave Kleikamp * Release the journal device 782ac27a0ecSDave Kleikamp */ 7834385bab1SAl Viro static void ext4_blkdev_put(struct block_device *bdev) 784ac27a0ecSDave Kleikamp { 7854385bab1SAl Viro blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 786ac27a0ecSDave Kleikamp } 787ac27a0ecSDave Kleikamp 7884385bab1SAl Viro static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 789ac27a0ecSDave Kleikamp { 790ac27a0ecSDave Kleikamp struct block_device *bdev; 791ac27a0ecSDave Kleikamp bdev = sbi->journal_bdev; 792ac27a0ecSDave Kleikamp if (bdev) { 7934385bab1SAl Viro ext4_blkdev_put(bdev); 794ac27a0ecSDave Kleikamp sbi->journal_bdev = NULL; 795ac27a0ecSDave Kleikamp } 796ac27a0ecSDave Kleikamp } 797ac27a0ecSDave Kleikamp 798ac27a0ecSDave Kleikamp static inline struct inode *orphan_list_entry(struct list_head *l) 799ac27a0ecSDave Kleikamp { 800617ba13bSMingming Cao return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 801ac27a0ecSDave Kleikamp } 802ac27a0ecSDave Kleikamp 803617ba13bSMingming Cao static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 804ac27a0ecSDave Kleikamp { 805ac27a0ecSDave Kleikamp struct list_head *l; 806ac27a0ecSDave Kleikamp 807b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 808ac27a0ecSDave Kleikamp le32_to_cpu(sbi->s_es->s_last_orphan)); 809ac27a0ecSDave Kleikamp 810ac27a0ecSDave Kleikamp printk(KERN_ERR "sb_info orphan list:\n"); 811ac27a0ecSDave Kleikamp list_for_each(l, &sbi->s_orphan) { 812ac27a0ecSDave Kleikamp struct inode *inode = orphan_list_entry(l); 813ac27a0ecSDave Kleikamp printk(KERN_ERR " " 814ac27a0ecSDave Kleikamp "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 815ac27a0ecSDave Kleikamp inode->i_sb->s_id, inode->i_ino, inode, 816ac27a0ecSDave Kleikamp inode->i_mode, inode->i_nlink, 817ac27a0ecSDave Kleikamp NEXT_ORPHAN(inode)); 818ac27a0ecSDave Kleikamp } 819ac27a0ecSDave Kleikamp } 820ac27a0ecSDave Kleikamp 821617ba13bSMingming Cao static void ext4_put_super(struct super_block *sb) 822ac27a0ecSDave Kleikamp { 823617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 824617ba13bSMingming Cao struct ext4_super_block *es = sbi->s_es; 825ef2cabf7SHidehiro Kawai int i, err; 826ac27a0ecSDave Kleikamp 827857ac889SLukas Czerner ext4_unregister_li_request(sb); 828e0ccfd95SChristoph Hellwig dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED); 829e0ccfd95SChristoph Hellwig 8302e8fa54eSJan Kara flush_workqueue(sbi->rsv_conversion_wq); 8312e8fa54eSJan Kara destroy_workqueue(sbi->rsv_conversion_wq); 8324c0425ffSMingming Cao 8330390131bSFrank Mayhar if (sbi->s_journal) { 834ef2cabf7SHidehiro Kawai err = jbd2_journal_destroy(sbi->s_journal); 83547b4a50bSJan Kara sbi->s_journal = NULL; 836ef2cabf7SHidehiro Kawai if (err < 0) 837c67d859eSTheodore Ts'o ext4_abort(sb, "Couldn't clean up the journal"); 8380390131bSFrank Mayhar } 839d4edac31SJosef Bacik 840ebd173beSTheodore Ts'o ext4_unregister_sysfs(sb); 841d3922a77SZheng Liu ext4_es_unregister_shrinker(sbi); 8429105bb14SAl Viro del_timer_sync(&sbi->s_err_report); 843d4edac31SJosef Bacik ext4_release_system_zone(sb); 844d4edac31SJosef Bacik ext4_mb_release(sb); 845d4edac31SJosef Bacik ext4_ext_release(sb); 846d4edac31SJosef Bacik 847ac27a0ecSDave Kleikamp if (!(sb->s_flags & MS_RDONLY)) { 848e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 849ac27a0ecSDave Kleikamp es->s_state = cpu_to_le16(sbi->s_mount_state); 850ac27a0ecSDave Kleikamp } 85158c5873aSArtem Bityutskiy if (!(sb->s_flags & MS_RDONLY)) 852a8e25a83SArtem Bityutskiy ext4_commit_super(sb, 1); 853a8e25a83SArtem Bityutskiy 854ac27a0ecSDave Kleikamp for (i = 0; i < sbi->s_gdb_count; i++) 855ac27a0ecSDave Kleikamp brelse(sbi->s_group_desc[i]); 856b93b41d4SAl Viro kvfree(sbi->s_group_desc); 857b93b41d4SAl Viro kvfree(sbi->s_flex_groups); 85857042651STheodore Ts'o percpu_counter_destroy(&sbi->s_freeclusters_counter); 859ac27a0ecSDave Kleikamp percpu_counter_destroy(&sbi->s_freeinodes_counter); 860ac27a0ecSDave Kleikamp percpu_counter_destroy(&sbi->s_dirs_counter); 86157042651STheodore Ts'o percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 862c8585c6fSDaeho Jeong percpu_free_rwsem(&sbi->s_journal_flag_rwsem); 863ac27a0ecSDave Kleikamp brelse(sbi->s_sbh); 864ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 865a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 866ac27a0ecSDave Kleikamp kfree(sbi->s_qf_names[i]); 867ac27a0ecSDave Kleikamp #endif 868ac27a0ecSDave Kleikamp 869ac27a0ecSDave Kleikamp /* Debugging code just in case the in-memory inode orphan list 870ac27a0ecSDave Kleikamp * isn't empty. The on-disk one can be non-empty if we've 871ac27a0ecSDave Kleikamp * detected an error and taken the fs readonly, but the 872ac27a0ecSDave Kleikamp * in-memory list had better be clean by this point. */ 873ac27a0ecSDave Kleikamp if (!list_empty(&sbi->s_orphan)) 874ac27a0ecSDave Kleikamp dump_orphan_list(sb, sbi); 875ac27a0ecSDave Kleikamp J_ASSERT(list_empty(&sbi->s_orphan)); 876ac27a0ecSDave Kleikamp 87789d96a6fSTheodore Ts'o sync_blockdev(sb->s_bdev); 878f98393a6SPeter Zijlstra invalidate_bdev(sb->s_bdev); 879ac27a0ecSDave Kleikamp if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { 880ac27a0ecSDave Kleikamp /* 881ac27a0ecSDave Kleikamp * Invalidate the journal device's buffers. We don't want them 882ac27a0ecSDave Kleikamp * floating about in memory - the physical journal device may 883ac27a0ecSDave Kleikamp * hotswapped, and it breaks the `ro-after' testing code. 884ac27a0ecSDave Kleikamp */ 885ac27a0ecSDave Kleikamp sync_blockdev(sbi->journal_bdev); 886f98393a6SPeter Zijlstra invalidate_bdev(sbi->journal_bdev); 887617ba13bSMingming Cao ext4_blkdev_remove(sbi); 888ac27a0ecSDave Kleikamp } 8899c191f70ST Makphaibulchoke if (sbi->s_mb_cache) { 8909c191f70ST Makphaibulchoke ext4_xattr_destroy_cache(sbi->s_mb_cache); 8919c191f70ST Makphaibulchoke sbi->s_mb_cache = NULL; 8929c191f70ST Makphaibulchoke } 893c5e06d10SJohann Lombardi if (sbi->s_mmp_tsk) 894c5e06d10SJohann Lombardi kthread_stop(sbi->s_mmp_tsk); 895ac27a0ecSDave Kleikamp sb->s_fs_info = NULL; 8963197ebdbSTheodore Ts'o /* 8973197ebdbSTheodore Ts'o * Now that we are completely done shutting down the 8983197ebdbSTheodore Ts'o * superblock, we need to actually destroy the kobject. 8993197ebdbSTheodore Ts'o */ 9003197ebdbSTheodore Ts'o kobject_put(&sbi->s_kobj); 9013197ebdbSTheodore Ts'o wait_for_completion(&sbi->s_kobj_unregister); 9020441984aSDarrick J. Wong if (sbi->s_chksum_driver) 9030441984aSDarrick J. Wong crypto_free_shash(sbi->s_chksum_driver); 904705895b6SPekka Enberg kfree(sbi->s_blockgroup_lock); 905ac27a0ecSDave Kleikamp kfree(sbi); 906ac27a0ecSDave Kleikamp } 907ac27a0ecSDave Kleikamp 908e18b890bSChristoph Lameter static struct kmem_cache *ext4_inode_cachep; 909ac27a0ecSDave Kleikamp 910ac27a0ecSDave Kleikamp /* 911ac27a0ecSDave Kleikamp * Called inside transaction, so use GFP_NOFS 912ac27a0ecSDave Kleikamp */ 913617ba13bSMingming Cao static struct inode *ext4_alloc_inode(struct super_block *sb) 914ac27a0ecSDave Kleikamp { 915617ba13bSMingming Cao struct ext4_inode_info *ei; 916ac27a0ecSDave Kleikamp 917e6b4f8daSChristoph Lameter ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 918ac27a0ecSDave Kleikamp if (!ei) 919ac27a0ecSDave Kleikamp return NULL; 9200b8e58a1SAndreas Dilger 921ac27a0ecSDave Kleikamp ei->vfs_inode.i_version = 1; 922202ee5dfSTheodore Ts'o spin_lock_init(&ei->i_raw_lock); 923c9de560dSAlex Tomas INIT_LIST_HEAD(&ei->i_prealloc_list); 924c9de560dSAlex Tomas spin_lock_init(&ei->i_prealloc_lock); 9259a26b661SZheng Liu ext4_es_init_tree(&ei->i_es_tree); 9269a26b661SZheng Liu rwlock_init(&ei->i_es_lock); 927edaa53caSZheng Liu INIT_LIST_HEAD(&ei->i_es_list); 928eb68d0e2SZheng Liu ei->i_es_all_nr = 0; 929edaa53caSZheng Liu ei->i_es_shk_nr = 0; 930dd475925SJan Kara ei->i_es_shrink_lblk = 0; 931d2a17637SMingming Cao ei->i_reserved_data_blocks = 0; 932d2a17637SMingming Cao ei->i_reserved_meta_blocks = 0; 933d2a17637SMingming Cao ei->i_allocated_meta_blocks = 0; 9349d0be502STheodore Ts'o ei->i_da_metadata_calc_len = 0; 9357e731bc9STheodore Ts'o ei->i_da_metadata_calc_last_lblock = 0; 936d2a17637SMingming Cao spin_lock_init(&(ei->i_block_reservation_lock)); 937a9e7f447SDmitry Monakhov #ifdef CONFIG_QUOTA 938a9e7f447SDmitry Monakhov ei->i_reserved_quota = 0; 93996c7e0d9SJan Kara memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 940a9e7f447SDmitry Monakhov #endif 9418aefcd55STheodore Ts'o ei->jinode = NULL; 9422e8fa54eSJan Kara INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 943744692dcSJiaying Zhang spin_lock_init(&ei->i_completed_io_lock); 944b436b9beSJan Kara ei->i_sync_tid = 0; 945b436b9beSJan Kara ei->i_datasync_tid = 0; 946e27f41e1SDmitry Monakhov atomic_set(&ei->i_unwritten, 0); 9472e8fa54eSJan Kara INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 948b30ab0e0SMichael Halcrow #ifdef CONFIG_EXT4_FS_ENCRYPTION 949b7236e21STheodore Ts'o ei->i_crypt_info = NULL; 950b30ab0e0SMichael Halcrow #endif 951ac27a0ecSDave Kleikamp return &ei->vfs_inode; 952ac27a0ecSDave Kleikamp } 953ac27a0ecSDave Kleikamp 9547ff9c073STheodore Ts'o static int ext4_drop_inode(struct inode *inode) 9557ff9c073STheodore Ts'o { 9567ff9c073STheodore Ts'o int drop = generic_drop_inode(inode); 9577ff9c073STheodore Ts'o 9587ff9c073STheodore Ts'o trace_ext4_drop_inode(inode, drop); 9597ff9c073STheodore Ts'o return drop; 9607ff9c073STheodore Ts'o } 9617ff9c073STheodore Ts'o 962fa0d7e3dSNick Piggin static void ext4_i_callback(struct rcu_head *head) 963fa0d7e3dSNick Piggin { 964fa0d7e3dSNick Piggin struct inode *inode = container_of(head, struct inode, i_rcu); 965fa0d7e3dSNick Piggin kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 966fa0d7e3dSNick Piggin } 967fa0d7e3dSNick Piggin 968617ba13bSMingming Cao static void ext4_destroy_inode(struct inode *inode) 969ac27a0ecSDave Kleikamp { 9709f7dd93dSVasily Averin if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 971b31e1552SEric Sandeen ext4_msg(inode->i_sb, KERN_ERR, 972b31e1552SEric Sandeen "Inode %lu (%p): orphan list check failed!", 973b31e1552SEric Sandeen inode->i_ino, EXT4_I(inode)); 9749f7dd93dSVasily Averin print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 9759f7dd93dSVasily Averin EXT4_I(inode), sizeof(struct ext4_inode_info), 9769f7dd93dSVasily Averin true); 9779f7dd93dSVasily Averin dump_stack(); 9789f7dd93dSVasily Averin } 979fa0d7e3dSNick Piggin call_rcu(&inode->i_rcu, ext4_i_callback); 980ac27a0ecSDave Kleikamp } 981ac27a0ecSDave Kleikamp 98251cc5068SAlexey Dobriyan static void init_once(void *foo) 983ac27a0ecSDave Kleikamp { 984617ba13bSMingming Cao struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 985ac27a0ecSDave Kleikamp 986ac27a0ecSDave Kleikamp INIT_LIST_HEAD(&ei->i_orphan); 987ac27a0ecSDave Kleikamp init_rwsem(&ei->xattr_sem); 9880e855ac8SAneesh Kumar K.V init_rwsem(&ei->i_data_sem); 989ea3d7209SJan Kara init_rwsem(&ei->i_mmap_sem); 990ac27a0ecSDave Kleikamp inode_init_once(&ei->vfs_inode); 991ac27a0ecSDave Kleikamp } 992ac27a0ecSDave Kleikamp 993e67bc2b3SFabian Frederick static int __init init_inodecache(void) 994ac27a0ecSDave Kleikamp { 995617ba13bSMingming Cao ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", 996617ba13bSMingming Cao sizeof(struct ext4_inode_info), 997ac27a0ecSDave Kleikamp 0, (SLAB_RECLAIM_ACCOUNT| 9985d097056SVladimir Davydov SLAB_MEM_SPREAD|SLAB_ACCOUNT), 99920c2df83SPaul Mundt init_once); 1000617ba13bSMingming Cao if (ext4_inode_cachep == NULL) 1001ac27a0ecSDave Kleikamp return -ENOMEM; 1002ac27a0ecSDave Kleikamp return 0; 1003ac27a0ecSDave Kleikamp } 1004ac27a0ecSDave Kleikamp 1005ac27a0ecSDave Kleikamp static void destroy_inodecache(void) 1006ac27a0ecSDave Kleikamp { 10078c0a8537SKirill A. Shutemov /* 10088c0a8537SKirill A. Shutemov * Make sure all delayed rcu free inodes are flushed before we 10098c0a8537SKirill A. Shutemov * destroy cache. 10108c0a8537SKirill A. Shutemov */ 10118c0a8537SKirill A. Shutemov rcu_barrier(); 1012617ba13bSMingming Cao kmem_cache_destroy(ext4_inode_cachep); 1013ac27a0ecSDave Kleikamp } 1014ac27a0ecSDave Kleikamp 10150930fcc1SAl Viro void ext4_clear_inode(struct inode *inode) 1016ac27a0ecSDave Kleikamp { 10170930fcc1SAl Viro invalidate_inode_buffers(inode); 1018dbd5768fSJan Kara clear_inode(inode); 10199f754758SChristoph Hellwig dquot_drop(inode); 1020c2ea3fdeSTheodore Ts'o ext4_discard_preallocations(inode); 102151865fdaSZheng Liu ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 10228aefcd55STheodore Ts'o if (EXT4_I(inode)->jinode) { 10238aefcd55STheodore Ts'o jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 10248aefcd55STheodore Ts'o EXT4_I(inode)->jinode); 10258aefcd55STheodore Ts'o jbd2_free_inode(EXT4_I(inode)->jinode); 10268aefcd55STheodore Ts'o EXT4_I(inode)->jinode = NULL; 10278aefcd55STheodore Ts'o } 1028b7236e21STheodore Ts'o #ifdef CONFIG_EXT4_FS_ENCRYPTION 1029b7236e21STheodore Ts'o if (EXT4_I(inode)->i_crypt_info) 1030c936e1ecSTheodore Ts'o ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info); 1031b7236e21STheodore Ts'o #endif 1032ac27a0ecSDave Kleikamp } 1033ac27a0ecSDave Kleikamp 10341b961ac0SChristoph Hellwig static struct inode *ext4_nfs_get_inode(struct super_block *sb, 10351b961ac0SChristoph Hellwig u64 ino, u32 generation) 1036ac27a0ecSDave Kleikamp { 1037ac27a0ecSDave Kleikamp struct inode *inode; 1038ac27a0ecSDave Kleikamp 1039617ba13bSMingming Cao if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO) 1040ac27a0ecSDave Kleikamp return ERR_PTR(-ESTALE); 1041617ba13bSMingming Cao if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)) 1042ac27a0ecSDave Kleikamp return ERR_PTR(-ESTALE); 1043ac27a0ecSDave Kleikamp 1044ac27a0ecSDave Kleikamp /* iget isn't really right if the inode is currently unallocated!! 1045ac27a0ecSDave Kleikamp * 1046617ba13bSMingming Cao * ext4_read_inode will return a bad_inode if the inode had been 1047ac27a0ecSDave Kleikamp * deleted, so we should be safe. 1048ac27a0ecSDave Kleikamp * 1049ac27a0ecSDave Kleikamp * Currently we don't know the generation for parent directory, so 1050ac27a0ecSDave Kleikamp * a generation of 0 means "accept any" 1051ac27a0ecSDave Kleikamp */ 1052f4bb2981STheodore Ts'o inode = ext4_iget_normal(sb, ino); 10531d1fe1eeSDavid Howells if (IS_ERR(inode)) 10541d1fe1eeSDavid Howells return ERR_CAST(inode); 10551d1fe1eeSDavid Howells if (generation && inode->i_generation != generation) { 1056ac27a0ecSDave Kleikamp iput(inode); 1057ac27a0ecSDave Kleikamp return ERR_PTR(-ESTALE); 1058ac27a0ecSDave Kleikamp } 10591b961ac0SChristoph Hellwig 10601b961ac0SChristoph Hellwig return inode; 1061ac27a0ecSDave Kleikamp } 10621b961ac0SChristoph Hellwig 10631b961ac0SChristoph Hellwig static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 10641b961ac0SChristoph Hellwig int fh_len, int fh_type) 10651b961ac0SChristoph Hellwig { 10661b961ac0SChristoph Hellwig return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 10671b961ac0SChristoph Hellwig ext4_nfs_get_inode); 10681b961ac0SChristoph Hellwig } 10691b961ac0SChristoph Hellwig 10701b961ac0SChristoph Hellwig static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 10711b961ac0SChristoph Hellwig int fh_len, int fh_type) 10721b961ac0SChristoph Hellwig { 10731b961ac0SChristoph Hellwig return generic_fh_to_parent(sb, fid, fh_len, fh_type, 10741b961ac0SChristoph Hellwig ext4_nfs_get_inode); 1075ac27a0ecSDave Kleikamp } 1076ac27a0ecSDave Kleikamp 1077c39a7f84SToshiyuki Okajima /* 1078c39a7f84SToshiyuki Okajima * Try to release metadata pages (indirect blocks, directories) which are 1079c39a7f84SToshiyuki Okajima * mapped via the block device. Since these pages could have journal heads 1080c39a7f84SToshiyuki Okajima * which would prevent try_to_free_buffers() from freeing them, we must use 1081c39a7f84SToshiyuki Okajima * jbd2 layer's try_to_free_buffers() function to release them. 1082c39a7f84SToshiyuki Okajima */ 10830b8e58a1SAndreas Dilger static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 10840b8e58a1SAndreas Dilger gfp_t wait) 1085c39a7f84SToshiyuki Okajima { 1086c39a7f84SToshiyuki Okajima journal_t *journal = EXT4_SB(sb)->s_journal; 1087c39a7f84SToshiyuki Okajima 1088c39a7f84SToshiyuki Okajima WARN_ON(PageChecked(page)); 1089c39a7f84SToshiyuki Okajima if (!page_has_buffers(page)) 1090c39a7f84SToshiyuki Okajima return 0; 1091c39a7f84SToshiyuki Okajima if (journal) 1092c39a7f84SToshiyuki Okajima return jbd2_journal_try_to_free_buffers(journal, page, 1093d0164adcSMel Gorman wait & ~__GFP_DIRECT_RECLAIM); 1094c39a7f84SToshiyuki Okajima return try_to_free_buffers(page); 1095c39a7f84SToshiyuki Okajima } 1096c39a7f84SToshiyuki Okajima 1097ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1098689c958cSLi Xi static char *quotatypes[] = INITQFNAMES; 1099689c958cSLi Xi #define QTYPE2NAME(t) (quotatypes[t]) 1100ac27a0ecSDave Kleikamp 1101617ba13bSMingming Cao static int ext4_write_dquot(struct dquot *dquot); 1102617ba13bSMingming Cao static int ext4_acquire_dquot(struct dquot *dquot); 1103617ba13bSMingming Cao static int ext4_release_dquot(struct dquot *dquot); 1104617ba13bSMingming Cao static int ext4_mark_dquot_dirty(struct dquot *dquot); 1105617ba13bSMingming Cao static int ext4_write_info(struct super_block *sb, int type); 11066f28e087SJan Kara static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1107f00c9e44SJan Kara struct path *path); 1108ca0e05e4SDmitry Monakhov static int ext4_quota_off(struct super_block *sb, int type); 1109617ba13bSMingming Cao static int ext4_quota_on_mount(struct super_block *sb, int type); 1110617ba13bSMingming Cao static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1111ac27a0ecSDave Kleikamp size_t len, loff_t off); 1112617ba13bSMingming Cao static ssize_t ext4_quota_write(struct super_block *sb, int type, 1113ac27a0ecSDave Kleikamp const char *data, size_t len, loff_t off); 11147c319d32SAditya Kali static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 11157c319d32SAditya Kali unsigned int flags); 11167c319d32SAditya Kali static int ext4_enable_quotas(struct super_block *sb); 11178f0e8746STheodore Ts'o static int ext4_get_next_id(struct super_block *sb, struct kqid *qid); 1118ac27a0ecSDave Kleikamp 111996c7e0d9SJan Kara static struct dquot **ext4_get_dquots(struct inode *inode) 112096c7e0d9SJan Kara { 112196c7e0d9SJan Kara return EXT4_I(inode)->i_dquot; 112296c7e0d9SJan Kara } 112396c7e0d9SJan Kara 112461e225dcSAlexey Dobriyan static const struct dquot_operations ext4_quota_operations = { 112560e58e0fSMingming Cao .get_reserved_space = ext4_get_reserved_space, 1126617ba13bSMingming Cao .write_dquot = ext4_write_dquot, 1127617ba13bSMingming Cao .acquire_dquot = ext4_acquire_dquot, 1128617ba13bSMingming Cao .release_dquot = ext4_release_dquot, 1129617ba13bSMingming Cao .mark_dirty = ext4_mark_dquot_dirty, 1130a5b5ee32SJan Kara .write_info = ext4_write_info, 1131a5b5ee32SJan Kara .alloc_dquot = dquot_alloc, 1132a5b5ee32SJan Kara .destroy_dquot = dquot_destroy, 1133040cb378SLi Xi .get_projid = ext4_get_projid, 11348f0e8746STheodore Ts'o .get_next_id = ext4_get_next_id, 1135ac27a0ecSDave Kleikamp }; 1136ac27a0ecSDave Kleikamp 11370d54b217SAlexey Dobriyan static const struct quotactl_ops ext4_qctl_operations = { 1138617ba13bSMingming Cao .quota_on = ext4_quota_on, 1139ca0e05e4SDmitry Monakhov .quota_off = ext4_quota_off, 1140287a8095SChristoph Hellwig .quota_sync = dquot_quota_sync, 11410a240339SJan Kara .get_state = dquot_get_state, 1142287a8095SChristoph Hellwig .set_info = dquot_set_dqinfo, 1143287a8095SChristoph Hellwig .get_dqblk = dquot_get_dqblk, 11446332b9b5SEric Sandeen .set_dqblk = dquot_set_dqblk, 11456332b9b5SEric Sandeen .get_nextdqblk = dquot_get_next_dqblk, 1146ac27a0ecSDave Kleikamp }; 1147ac27a0ecSDave Kleikamp #endif 1148ac27a0ecSDave Kleikamp 1149ee9b6d61SJosef 'Jeff' Sipek static const struct super_operations ext4_sops = { 1150617ba13bSMingming Cao .alloc_inode = ext4_alloc_inode, 1151617ba13bSMingming Cao .destroy_inode = ext4_destroy_inode, 1152617ba13bSMingming Cao .write_inode = ext4_write_inode, 1153617ba13bSMingming Cao .dirty_inode = ext4_dirty_inode, 11547ff9c073STheodore Ts'o .drop_inode = ext4_drop_inode, 11550930fcc1SAl Viro .evict_inode = ext4_evict_inode, 1156617ba13bSMingming Cao .put_super = ext4_put_super, 1157617ba13bSMingming Cao .sync_fs = ext4_sync_fs, 1158c4be0c1dSTakashi Sato .freeze_fs = ext4_freeze, 1159c4be0c1dSTakashi Sato .unfreeze_fs = ext4_unfreeze, 1160617ba13bSMingming Cao .statfs = ext4_statfs, 1161617ba13bSMingming Cao .remount_fs = ext4_remount, 1162617ba13bSMingming Cao .show_options = ext4_show_options, 1163ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1164617ba13bSMingming Cao .quota_read = ext4_quota_read, 1165617ba13bSMingming Cao .quota_write = ext4_quota_write, 116696c7e0d9SJan Kara .get_dquots = ext4_get_dquots, 1167ac27a0ecSDave Kleikamp #endif 1168c39a7f84SToshiyuki Okajima .bdev_try_to_free_page = bdev_try_to_free_page, 1169ac27a0ecSDave Kleikamp }; 1170ac27a0ecSDave Kleikamp 117139655164SChristoph Hellwig static const struct export_operations ext4_export_ops = { 11721b961ac0SChristoph Hellwig .fh_to_dentry = ext4_fh_to_dentry, 11731b961ac0SChristoph Hellwig .fh_to_parent = ext4_fh_to_parent, 1174617ba13bSMingming Cao .get_parent = ext4_get_parent, 1175ac27a0ecSDave Kleikamp }; 1176ac27a0ecSDave Kleikamp 1177ac27a0ecSDave Kleikamp enum { 1178ac27a0ecSDave Kleikamp Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1179ac27a0ecSDave Kleikamp Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 118072578c33STheodore Ts'o Opt_nouid32, Opt_debug, Opt_removed, 1181ac27a0ecSDave Kleikamp Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 118272578c33STheodore Ts'o Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1183ad4eec61SEric Sandeen Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1184ad4eec61SEric Sandeen Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1185ac27a0ecSDave Kleikamp Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 11866ddb2447STheodore Ts'o Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 1187ac27a0ecSDave Kleikamp Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 11885a20bdfcSJan Kara Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1189ee4a3fcdSTheodore Ts'o Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1190923ae0ffSRoss Zwisler Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax, 11911449032bSTheodore Ts'o Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit, 1192a26f4992STheodore Ts'o Opt_lazytime, Opt_nolazytime, 11931449032bSTheodore Ts'o Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 11945328e635SEric Sandeen Opt_inode_readahead_blks, Opt_journal_ioprio, 1195744692dcSJiaying Zhang Opt_dioread_nolock, Opt_dioread_lock, 1196fc6cb1cdSTheodore Ts'o Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1197c6d3d56dSDarrick J. Wong Opt_max_dir_size_kb, Opt_nojournal_checksum, 1198ac27a0ecSDave Kleikamp }; 1199ac27a0ecSDave Kleikamp 1200a447c093SSteven Whitehouse static const match_table_t tokens = { 1201ac27a0ecSDave Kleikamp {Opt_bsd_df, "bsddf"}, 1202ac27a0ecSDave Kleikamp {Opt_minix_df, "minixdf"}, 1203ac27a0ecSDave Kleikamp {Opt_grpid, "grpid"}, 1204ac27a0ecSDave Kleikamp {Opt_grpid, "bsdgroups"}, 1205ac27a0ecSDave Kleikamp {Opt_nogrpid, "nogrpid"}, 1206ac27a0ecSDave Kleikamp {Opt_nogrpid, "sysvgroups"}, 1207ac27a0ecSDave Kleikamp {Opt_resgid, "resgid=%u"}, 1208ac27a0ecSDave Kleikamp {Opt_resuid, "resuid=%u"}, 1209ac27a0ecSDave Kleikamp {Opt_sb, "sb=%u"}, 1210ac27a0ecSDave Kleikamp {Opt_err_cont, "errors=continue"}, 1211ac27a0ecSDave Kleikamp {Opt_err_panic, "errors=panic"}, 1212ac27a0ecSDave Kleikamp {Opt_err_ro, "errors=remount-ro"}, 1213ac27a0ecSDave Kleikamp {Opt_nouid32, "nouid32"}, 1214ac27a0ecSDave Kleikamp {Opt_debug, "debug"}, 121572578c33STheodore Ts'o {Opt_removed, "oldalloc"}, 121672578c33STheodore Ts'o {Opt_removed, "orlov"}, 1217ac27a0ecSDave Kleikamp {Opt_user_xattr, "user_xattr"}, 1218ac27a0ecSDave Kleikamp {Opt_nouser_xattr, "nouser_xattr"}, 1219ac27a0ecSDave Kleikamp {Opt_acl, "acl"}, 1220ac27a0ecSDave Kleikamp {Opt_noacl, "noacl"}, 1221e3bb52aeSEric Sandeen {Opt_noload, "norecovery"}, 12225a916be1STheodore Ts'o {Opt_noload, "noload"}, 122372578c33STheodore Ts'o {Opt_removed, "nobh"}, 122472578c33STheodore Ts'o {Opt_removed, "bh"}, 1225ac27a0ecSDave Kleikamp {Opt_commit, "commit=%u"}, 122630773840STheodore Ts'o {Opt_min_batch_time, "min_batch_time=%u"}, 122730773840STheodore Ts'o {Opt_max_batch_time, "max_batch_time=%u"}, 1228ac27a0ecSDave Kleikamp {Opt_journal_dev, "journal_dev=%u"}, 1229ad4eec61SEric Sandeen {Opt_journal_path, "journal_path=%s"}, 1230818d276cSGirish Shilamkar {Opt_journal_checksum, "journal_checksum"}, 1231c6d3d56dSDarrick J. Wong {Opt_nojournal_checksum, "nojournal_checksum"}, 1232818d276cSGirish Shilamkar {Opt_journal_async_commit, "journal_async_commit"}, 1233ac27a0ecSDave Kleikamp {Opt_abort, "abort"}, 1234ac27a0ecSDave Kleikamp {Opt_data_journal, "data=journal"}, 1235ac27a0ecSDave Kleikamp {Opt_data_ordered, "data=ordered"}, 1236ac27a0ecSDave Kleikamp {Opt_data_writeback, "data=writeback"}, 12375bf5683aSHidehiro Kawai {Opt_data_err_abort, "data_err=abort"}, 12385bf5683aSHidehiro Kawai {Opt_data_err_ignore, "data_err=ignore"}, 1239ac27a0ecSDave Kleikamp {Opt_offusrjquota, "usrjquota="}, 1240ac27a0ecSDave Kleikamp {Opt_usrjquota, "usrjquota=%s"}, 1241ac27a0ecSDave Kleikamp {Opt_offgrpjquota, "grpjquota="}, 1242ac27a0ecSDave Kleikamp {Opt_grpjquota, "grpjquota=%s"}, 1243ac27a0ecSDave Kleikamp {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1244ac27a0ecSDave Kleikamp {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 12455a20bdfcSJan Kara {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1246ac27a0ecSDave Kleikamp {Opt_grpquota, "grpquota"}, 1247ac27a0ecSDave Kleikamp {Opt_noquota, "noquota"}, 1248ac27a0ecSDave Kleikamp {Opt_quota, "quota"}, 1249ac27a0ecSDave Kleikamp {Opt_usrquota, "usrquota"}, 1250ac27a0ecSDave Kleikamp {Opt_barrier, "barrier=%u"}, 125106705bffSTheodore Ts'o {Opt_barrier, "barrier"}, 125206705bffSTheodore Ts'o {Opt_nobarrier, "nobarrier"}, 125325ec56b5SJean Noel Cordenner {Opt_i_version, "i_version"}, 1254923ae0ffSRoss Zwisler {Opt_dax, "dax"}, 1255c9de560dSAlex Tomas {Opt_stripe, "stripe=%u"}, 125664769240SAlex Tomas {Opt_delalloc, "delalloc"}, 1257a26f4992STheodore Ts'o {Opt_lazytime, "lazytime"}, 1258a26f4992STheodore Ts'o {Opt_nolazytime, "nolazytime"}, 1259dd919b98SAneesh Kumar K.V {Opt_nodelalloc, "nodelalloc"}, 126036ade451SJan Kara {Opt_removed, "mblk_io_submit"}, 126136ade451SJan Kara {Opt_removed, "nomblk_io_submit"}, 12626fd058f7STheodore Ts'o {Opt_block_validity, "block_validity"}, 12636fd058f7STheodore Ts'o {Opt_noblock_validity, "noblock_validity"}, 1264240799cdSTheodore Ts'o {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1265b3881f74STheodore Ts'o {Opt_journal_ioprio, "journal_ioprio=%u"}, 1266afd4672dSTheodore Ts'o {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 126706705bffSTheodore Ts'o {Opt_auto_da_alloc, "auto_da_alloc"}, 126806705bffSTheodore Ts'o {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1269744692dcSJiaying Zhang {Opt_dioread_nolock, "dioread_nolock"}, 1270744692dcSJiaying Zhang {Opt_dioread_lock, "dioread_lock"}, 12715328e635SEric Sandeen {Opt_discard, "discard"}, 12725328e635SEric Sandeen {Opt_nodiscard, "nodiscard"}, 1273fc6cb1cdSTheodore Ts'o {Opt_init_itable, "init_itable=%u"}, 1274fc6cb1cdSTheodore Ts'o {Opt_init_itable, "init_itable"}, 1275fc6cb1cdSTheodore Ts'o {Opt_noinit_itable, "noinit_itable"}, 1276df981d03STheodore Ts'o {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 12776ddb2447STheodore Ts'o {Opt_test_dummy_encryption, "test_dummy_encryption"}, 1278c7198b9cSTheodore Ts'o {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1279c7198b9cSTheodore Ts'o {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1280c7198b9cSTheodore Ts'o {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1281c7198b9cSTheodore Ts'o {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1282c7198b9cSTheodore Ts'o {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1283f3f12faaSJosef Bacik {Opt_err, NULL}, 1284ac27a0ecSDave Kleikamp }; 1285ac27a0ecSDave Kleikamp 1286617ba13bSMingming Cao static ext4_fsblk_t get_sb_block(void **data) 1287ac27a0ecSDave Kleikamp { 1288617ba13bSMingming Cao ext4_fsblk_t sb_block; 1289ac27a0ecSDave Kleikamp char *options = (char *) *data; 1290ac27a0ecSDave Kleikamp 1291ac27a0ecSDave Kleikamp if (!options || strncmp(options, "sb=", 3) != 0) 1292ac27a0ecSDave Kleikamp return 1; /* Default location */ 12930b8e58a1SAndreas Dilger 1294ac27a0ecSDave Kleikamp options += 3; 12950b8e58a1SAndreas Dilger /* TODO: use simple_strtoll with >32bit ext4 */ 1296ac27a0ecSDave Kleikamp sb_block = simple_strtoul(options, &options, 0); 1297ac27a0ecSDave Kleikamp if (*options && *options != ',') { 12984776004fSTheodore Ts'o printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1299ac27a0ecSDave Kleikamp (char *) *data); 1300ac27a0ecSDave Kleikamp return 1; 1301ac27a0ecSDave Kleikamp } 1302ac27a0ecSDave Kleikamp if (*options == ',') 1303ac27a0ecSDave Kleikamp options++; 1304ac27a0ecSDave Kleikamp *data = (void *) options; 13050b8e58a1SAndreas Dilger 1306ac27a0ecSDave Kleikamp return sb_block; 1307ac27a0ecSDave Kleikamp } 1308ac27a0ecSDave Kleikamp 1309b3881f74STheodore Ts'o #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1310437ca0fdSDmitry Monakhov static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n" 1311437ca0fdSDmitry Monakhov "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1312b3881f74STheodore Ts'o 131356c50f11SDmitry Monakhov #ifdef CONFIG_QUOTA 131456c50f11SDmitry Monakhov static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 131556c50f11SDmitry Monakhov { 131656c50f11SDmitry Monakhov struct ext4_sb_info *sbi = EXT4_SB(sb); 131756c50f11SDmitry Monakhov char *qname; 131803dafb5fSChen Gang int ret = -1; 131956c50f11SDmitry Monakhov 132056c50f11SDmitry Monakhov if (sb_any_quota_loaded(sb) && 132156c50f11SDmitry Monakhov !sbi->s_qf_names[qtype]) { 132256c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 132356c50f11SDmitry Monakhov "Cannot change journaled " 132456c50f11SDmitry Monakhov "quota options when quota turned on"); 132557f73c2cSTheodore Ts'o return -1; 132656c50f11SDmitry Monakhov } 1327e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) { 1328c325a67cSTheodore Ts'o ext4_msg(sb, KERN_INFO, "Journaled quota options " 1329c325a67cSTheodore Ts'o "ignored when QUOTA feature is enabled"); 1330c325a67cSTheodore Ts'o return 1; 1331262b4662SJan Kara } 133256c50f11SDmitry Monakhov qname = match_strdup(args); 133356c50f11SDmitry Monakhov if (!qname) { 133456c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 133556c50f11SDmitry Monakhov "Not enough memory for storing quotafile name"); 133657f73c2cSTheodore Ts'o return -1; 133756c50f11SDmitry Monakhov } 133803dafb5fSChen Gang if (sbi->s_qf_names[qtype]) { 133903dafb5fSChen Gang if (strcmp(sbi->s_qf_names[qtype], qname) == 0) 134003dafb5fSChen Gang ret = 1; 134103dafb5fSChen Gang else 134256c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 134303dafb5fSChen Gang "%s quota file already specified", 134403dafb5fSChen Gang QTYPE2NAME(qtype)); 134503dafb5fSChen Gang goto errout; 134656c50f11SDmitry Monakhov } 134703dafb5fSChen Gang if (strchr(qname, '/')) { 134856c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 134956c50f11SDmitry Monakhov "quotafile must be on filesystem root"); 135003dafb5fSChen Gang goto errout; 135156c50f11SDmitry Monakhov } 135203dafb5fSChen Gang sbi->s_qf_names[qtype] = qname; 1353fd8c37ecSTheodore Ts'o set_opt(sb, QUOTA); 135456c50f11SDmitry Monakhov return 1; 135503dafb5fSChen Gang errout: 135603dafb5fSChen Gang kfree(qname); 135703dafb5fSChen Gang return ret; 135856c50f11SDmitry Monakhov } 135956c50f11SDmitry Monakhov 136056c50f11SDmitry Monakhov static int clear_qf_name(struct super_block *sb, int qtype) 136156c50f11SDmitry Monakhov { 136256c50f11SDmitry Monakhov 136356c50f11SDmitry Monakhov struct ext4_sb_info *sbi = EXT4_SB(sb); 136456c50f11SDmitry Monakhov 136556c50f11SDmitry Monakhov if (sb_any_quota_loaded(sb) && 136656c50f11SDmitry Monakhov sbi->s_qf_names[qtype]) { 136756c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 136856c50f11SDmitry Monakhov " when quota turned on"); 136957f73c2cSTheodore Ts'o return -1; 137056c50f11SDmitry Monakhov } 137103dafb5fSChen Gang kfree(sbi->s_qf_names[qtype]); 137256c50f11SDmitry Monakhov sbi->s_qf_names[qtype] = NULL; 137356c50f11SDmitry Monakhov return 1; 137456c50f11SDmitry Monakhov } 137556c50f11SDmitry Monakhov #endif 137656c50f11SDmitry Monakhov 137726092bf5STheodore Ts'o #define MOPT_SET 0x0001 137826092bf5STheodore Ts'o #define MOPT_CLEAR 0x0002 137926092bf5STheodore Ts'o #define MOPT_NOSUPPORT 0x0004 138026092bf5STheodore Ts'o #define MOPT_EXPLICIT 0x0008 138126092bf5STheodore Ts'o #define MOPT_CLEAR_ERR 0x0010 138226092bf5STheodore Ts'o #define MOPT_GTE0 0x0020 138326092bf5STheodore Ts'o #ifdef CONFIG_QUOTA 138426092bf5STheodore Ts'o #define MOPT_Q 0 138526092bf5STheodore Ts'o #define MOPT_QFMT 0x0040 138626092bf5STheodore Ts'o #else 138726092bf5STheodore Ts'o #define MOPT_Q MOPT_NOSUPPORT 138826092bf5STheodore Ts'o #define MOPT_QFMT MOPT_NOSUPPORT 138926092bf5STheodore Ts'o #endif 139026092bf5STheodore Ts'o #define MOPT_DATAJ 0x0080 13918dc0aa8cSTheodore Ts'o #define MOPT_NO_EXT2 0x0100 13928dc0aa8cSTheodore Ts'o #define MOPT_NO_EXT3 0x0200 13938dc0aa8cSTheodore Ts'o #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1394ad4eec61SEric Sandeen #define MOPT_STRING 0x0400 139526092bf5STheodore Ts'o 139626092bf5STheodore Ts'o static const struct mount_opts { 139726092bf5STheodore Ts'o int token; 139826092bf5STheodore Ts'o int mount_opt; 139926092bf5STheodore Ts'o int flags; 140026092bf5STheodore Ts'o } ext4_mount_opts[] = { 140126092bf5STheodore Ts'o {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 140226092bf5STheodore Ts'o {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 140326092bf5STheodore Ts'o {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 140426092bf5STheodore Ts'o {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 140526092bf5STheodore Ts'o {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 140626092bf5STheodore Ts'o {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 14078dc0aa8cSTheodore Ts'o {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 14088dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_SET}, 14098dc0aa8cSTheodore Ts'o {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 14108dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_CLEAR}, 141126092bf5STheodore Ts'o {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 141226092bf5STheodore Ts'o {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 14138dc0aa8cSTheodore Ts'o {Opt_delalloc, EXT4_MOUNT_DELALLOC, 14148dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 14158dc0aa8cSTheodore Ts'o {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 141659d9fa5cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_CLEAR}, 1417c6d3d56dSDarrick J. Wong {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1418c6d3d56dSDarrick J. Wong MOPT_EXT4_ONLY | MOPT_CLEAR}, 14198dc0aa8cSTheodore Ts'o {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 14201e381f60SDmitry Monakhov MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 142126092bf5STheodore Ts'o {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 14228dc0aa8cSTheodore Ts'o EXT4_MOUNT_JOURNAL_CHECKSUM), 14231e381f60SDmitry Monakhov MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 14248dc0aa8cSTheodore Ts'o {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 142526092bf5STheodore Ts'o {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 142626092bf5STheodore Ts'o {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 142726092bf5STheodore Ts'o {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 14288dc0aa8cSTheodore Ts'o {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 14297915a861SAles Novak MOPT_NO_EXT2}, 14308dc0aa8cSTheodore Ts'o {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 14317915a861SAles Novak MOPT_NO_EXT2}, 143226092bf5STheodore Ts'o {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 143326092bf5STheodore Ts'o {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 143426092bf5STheodore Ts'o {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 143526092bf5STheodore Ts'o {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 143626092bf5STheodore Ts'o {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 143726092bf5STheodore Ts'o {Opt_commit, 0, MOPT_GTE0}, 143826092bf5STheodore Ts'o {Opt_max_batch_time, 0, MOPT_GTE0}, 143926092bf5STheodore Ts'o {Opt_min_batch_time, 0, MOPT_GTE0}, 144026092bf5STheodore Ts'o {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 144126092bf5STheodore Ts'o {Opt_init_itable, 0, MOPT_GTE0}, 1442923ae0ffSRoss Zwisler {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET}, 144326092bf5STheodore Ts'o {Opt_stripe, 0, MOPT_GTE0}, 14440efb3b23SJan Kara {Opt_resuid, 0, MOPT_GTE0}, 14450efb3b23SJan Kara {Opt_resgid, 0, MOPT_GTE0}, 14465ba92bcfSCarlos Maiolino {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 14475ba92bcfSCarlos Maiolino {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING}, 14485ba92bcfSCarlos Maiolino {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 14498dc0aa8cSTheodore Ts'o {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 14508dc0aa8cSTheodore Ts'o {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 14518dc0aa8cSTheodore Ts'o {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 14528dc0aa8cSTheodore Ts'o MOPT_NO_EXT2 | MOPT_DATAJ}, 145326092bf5STheodore Ts'o {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 145426092bf5STheodore Ts'o {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 145526092bf5STheodore Ts'o #ifdef CONFIG_EXT4_FS_POSIX_ACL 145626092bf5STheodore Ts'o {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 145726092bf5STheodore Ts'o {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 145826092bf5STheodore Ts'o #else 145926092bf5STheodore Ts'o {Opt_acl, 0, MOPT_NOSUPPORT}, 146026092bf5STheodore Ts'o {Opt_noacl, 0, MOPT_NOSUPPORT}, 146126092bf5STheodore Ts'o #endif 146226092bf5STheodore Ts'o {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 146326092bf5STheodore Ts'o {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 146426092bf5STheodore Ts'o {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 146526092bf5STheodore Ts'o {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 146626092bf5STheodore Ts'o MOPT_SET | MOPT_Q}, 146726092bf5STheodore Ts'o {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 146826092bf5STheodore Ts'o MOPT_SET | MOPT_Q}, 146926092bf5STheodore Ts'o {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 147026092bf5STheodore Ts'o EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q}, 147126092bf5STheodore Ts'o {Opt_usrjquota, 0, MOPT_Q}, 147226092bf5STheodore Ts'o {Opt_grpjquota, 0, MOPT_Q}, 147326092bf5STheodore Ts'o {Opt_offusrjquota, 0, MOPT_Q}, 147426092bf5STheodore Ts'o {Opt_offgrpjquota, 0, MOPT_Q}, 147526092bf5STheodore Ts'o {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 147626092bf5STheodore Ts'o {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 147726092bf5STheodore Ts'o {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 1478df981d03STheodore Ts'o {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 14796ddb2447STheodore Ts'o {Opt_test_dummy_encryption, 0, MOPT_GTE0}, 148026092bf5STheodore Ts'o {Opt_err, 0, 0} 148126092bf5STheodore Ts'o }; 148226092bf5STheodore Ts'o 148326092bf5STheodore Ts'o static int handle_mount_opt(struct super_block *sb, char *opt, int token, 148426092bf5STheodore Ts'o substring_t *args, unsigned long *journal_devnum, 148526092bf5STheodore Ts'o unsigned int *journal_ioprio, int is_remount) 148626092bf5STheodore Ts'o { 148726092bf5STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 148826092bf5STheodore Ts'o const struct mount_opts *m; 148908cefc7aSEric W. Biederman kuid_t uid; 149008cefc7aSEric W. Biederman kgid_t gid; 149126092bf5STheodore Ts'o int arg = 0; 149226092bf5STheodore Ts'o 149357f73c2cSTheodore Ts'o #ifdef CONFIG_QUOTA 149457f73c2cSTheodore Ts'o if (token == Opt_usrjquota) 149557f73c2cSTheodore Ts'o return set_qf_name(sb, USRQUOTA, &args[0]); 149657f73c2cSTheodore Ts'o else if (token == Opt_grpjquota) 149757f73c2cSTheodore Ts'o return set_qf_name(sb, GRPQUOTA, &args[0]); 149857f73c2cSTheodore Ts'o else if (token == Opt_offusrjquota) 149957f73c2cSTheodore Ts'o return clear_qf_name(sb, USRQUOTA); 150057f73c2cSTheodore Ts'o else if (token == Opt_offgrpjquota) 150157f73c2cSTheodore Ts'o return clear_qf_name(sb, GRPQUOTA); 150257f73c2cSTheodore Ts'o #endif 150326092bf5STheodore Ts'o switch (token) { 1504f7048605STheodore Ts'o case Opt_noacl: 1505f7048605STheodore Ts'o case Opt_nouser_xattr: 1506f7048605STheodore Ts'o ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 1507f7048605STheodore Ts'o break; 150826092bf5STheodore Ts'o case Opt_sb: 150926092bf5STheodore Ts'o return 1; /* handled by get_sb_block() */ 151026092bf5STheodore Ts'o case Opt_removed: 15115f3633e3SJan Kara ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 151226092bf5STheodore Ts'o return 1; 151326092bf5STheodore Ts'o case Opt_abort: 151426092bf5STheodore Ts'o sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 151526092bf5STheodore Ts'o return 1; 151626092bf5STheodore Ts'o case Opt_i_version: 151726092bf5STheodore Ts'o sb->s_flags |= MS_I_VERSION; 151826092bf5STheodore Ts'o return 1; 1519a26f4992STheodore Ts'o case Opt_lazytime: 1520a26f4992STheodore Ts'o sb->s_flags |= MS_LAZYTIME; 1521a26f4992STheodore Ts'o return 1; 1522a26f4992STheodore Ts'o case Opt_nolazytime: 1523a26f4992STheodore Ts'o sb->s_flags &= ~MS_LAZYTIME; 1524a26f4992STheodore Ts'o return 1; 152526092bf5STheodore Ts'o } 152626092bf5STheodore Ts'o 15275f3633e3SJan Kara for (m = ext4_mount_opts; m->token != Opt_err; m++) 15285f3633e3SJan Kara if (token == m->token) 15295f3633e3SJan Kara break; 15305f3633e3SJan Kara 15315f3633e3SJan Kara if (m->token == Opt_err) { 15325f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 15335f3633e3SJan Kara "or missing value", opt); 15345f3633e3SJan Kara return -1; 15355f3633e3SJan Kara } 15365f3633e3SJan Kara 15378dc0aa8cSTheodore Ts'o if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 15388dc0aa8cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 15398dc0aa8cSTheodore Ts'o "Mount option \"%s\" incompatible with ext2", opt); 15408dc0aa8cSTheodore Ts'o return -1; 15418dc0aa8cSTheodore Ts'o } 15428dc0aa8cSTheodore Ts'o if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 15438dc0aa8cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 15448dc0aa8cSTheodore Ts'o "Mount option \"%s\" incompatible with ext3", opt); 15458dc0aa8cSTheodore Ts'o return -1; 15468dc0aa8cSTheodore Ts'o } 15478dc0aa8cSTheodore Ts'o 1548ad4eec61SEric Sandeen if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) 15490efb3b23SJan Kara return -1; 155026092bf5STheodore Ts'o if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 155126092bf5STheodore Ts'o return -1; 1552c93cf2d7SDmitry Monakhov if (m->flags & MOPT_EXPLICIT) { 1553c93cf2d7SDmitry Monakhov if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 155426092bf5STheodore Ts'o set_opt2(sb, EXPLICIT_DELALLOC); 15551e381f60SDmitry Monakhov } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 15561e381f60SDmitry Monakhov set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM); 1557c93cf2d7SDmitry Monakhov } else 1558c93cf2d7SDmitry Monakhov return -1; 1559c93cf2d7SDmitry Monakhov } 156026092bf5STheodore Ts'o if (m->flags & MOPT_CLEAR_ERR) 156126092bf5STheodore Ts'o clear_opt(sb, ERRORS_MASK); 156226092bf5STheodore Ts'o if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 156326092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "Cannot change quota " 156426092bf5STheodore Ts'o "options when quota turned on"); 156526092bf5STheodore Ts'o return -1; 156626092bf5STheodore Ts'o } 156726092bf5STheodore Ts'o 156826092bf5STheodore Ts'o if (m->flags & MOPT_NOSUPPORT) { 156926092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 157026092bf5STheodore Ts'o } else if (token == Opt_commit) { 157126092bf5STheodore Ts'o if (arg == 0) 157226092bf5STheodore Ts'o arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 157326092bf5STheodore Ts'o sbi->s_commit_interval = HZ * arg; 157426092bf5STheodore Ts'o } else if (token == Opt_max_batch_time) { 157526092bf5STheodore Ts'o sbi->s_max_batch_time = arg; 157626092bf5STheodore Ts'o } else if (token == Opt_min_batch_time) { 157726092bf5STheodore Ts'o sbi->s_min_batch_time = arg; 157826092bf5STheodore Ts'o } else if (token == Opt_inode_readahead_blks) { 1579e33e60eaSJan Kara if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 1580e33e60eaSJan Kara ext4_msg(sb, KERN_ERR, 1581e33e60eaSJan Kara "EXT4-fs: inode_readahead_blks must be " 1582e33e60eaSJan Kara "0 or a power of 2 smaller than 2^31"); 158326092bf5STheodore Ts'o return -1; 158426092bf5STheodore Ts'o } 158526092bf5STheodore Ts'o sbi->s_inode_readahead_blks = arg; 158626092bf5STheodore Ts'o } else if (token == Opt_init_itable) { 158726092bf5STheodore Ts'o set_opt(sb, INIT_INODE_TABLE); 158826092bf5STheodore Ts'o if (!args->from) 158926092bf5STheodore Ts'o arg = EXT4_DEF_LI_WAIT_MULT; 159026092bf5STheodore Ts'o sbi->s_li_wait_mult = arg; 1591df981d03STheodore Ts'o } else if (token == Opt_max_dir_size_kb) { 1592df981d03STheodore Ts'o sbi->s_max_dir_size_kb = arg; 159326092bf5STheodore Ts'o } else if (token == Opt_stripe) { 159426092bf5STheodore Ts'o sbi->s_stripe = arg; 15950efb3b23SJan Kara } else if (token == Opt_resuid) { 15960efb3b23SJan Kara uid = make_kuid(current_user_ns(), arg); 15970efb3b23SJan Kara if (!uid_valid(uid)) { 15985f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 15990efb3b23SJan Kara return -1; 16000efb3b23SJan Kara } 16010efb3b23SJan Kara sbi->s_resuid = uid; 16020efb3b23SJan Kara } else if (token == Opt_resgid) { 16030efb3b23SJan Kara gid = make_kgid(current_user_ns(), arg); 16040efb3b23SJan Kara if (!gid_valid(gid)) { 16055f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 16060efb3b23SJan Kara return -1; 16070efb3b23SJan Kara } 16080efb3b23SJan Kara sbi->s_resgid = gid; 16090efb3b23SJan Kara } else if (token == Opt_journal_dev) { 16100efb3b23SJan Kara if (is_remount) { 16110efb3b23SJan Kara ext4_msg(sb, KERN_ERR, 16120efb3b23SJan Kara "Cannot specify journal on remount"); 16130efb3b23SJan Kara return -1; 16140efb3b23SJan Kara } 16150efb3b23SJan Kara *journal_devnum = arg; 1616ad4eec61SEric Sandeen } else if (token == Opt_journal_path) { 1617ad4eec61SEric Sandeen char *journal_path; 1618ad4eec61SEric Sandeen struct inode *journal_inode; 1619ad4eec61SEric Sandeen struct path path; 1620ad4eec61SEric Sandeen int error; 1621ad4eec61SEric Sandeen 1622ad4eec61SEric Sandeen if (is_remount) { 1623ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, 1624ad4eec61SEric Sandeen "Cannot specify journal on remount"); 1625ad4eec61SEric Sandeen return -1; 1626ad4eec61SEric Sandeen } 1627ad4eec61SEric Sandeen journal_path = match_strdup(&args[0]); 1628ad4eec61SEric Sandeen if (!journal_path) { 1629ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: could not dup " 1630ad4eec61SEric Sandeen "journal device string"); 1631ad4eec61SEric Sandeen return -1; 1632ad4eec61SEric Sandeen } 1633ad4eec61SEric Sandeen 1634ad4eec61SEric Sandeen error = kern_path(journal_path, LOOKUP_FOLLOW, &path); 1635ad4eec61SEric Sandeen if (error) { 1636ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: could not find " 1637ad4eec61SEric Sandeen "journal device path: error %d", error); 1638ad4eec61SEric Sandeen kfree(journal_path); 1639ad4eec61SEric Sandeen return -1; 1640ad4eec61SEric Sandeen } 1641ad4eec61SEric Sandeen 16422b0143b5SDavid Howells journal_inode = d_inode(path.dentry); 1643ad4eec61SEric Sandeen if (!S_ISBLK(journal_inode->i_mode)) { 1644ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: journal path %s " 1645ad4eec61SEric Sandeen "is not a block device", journal_path); 1646ad4eec61SEric Sandeen path_put(&path); 1647ad4eec61SEric Sandeen kfree(journal_path); 1648ad4eec61SEric Sandeen return -1; 1649ad4eec61SEric Sandeen } 1650ad4eec61SEric Sandeen 1651ad4eec61SEric Sandeen *journal_devnum = new_encode_dev(journal_inode->i_rdev); 1652ad4eec61SEric Sandeen path_put(&path); 1653ad4eec61SEric Sandeen kfree(journal_path); 16540efb3b23SJan Kara } else if (token == Opt_journal_ioprio) { 16550efb3b23SJan Kara if (arg > 7) { 16565f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 16570efb3b23SJan Kara " (must be 0-7)"); 16580efb3b23SJan Kara return -1; 16590efb3b23SJan Kara } 16600efb3b23SJan Kara *journal_ioprio = 16610efb3b23SJan Kara IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 16626ddb2447STheodore Ts'o } else if (token == Opt_test_dummy_encryption) { 16636ddb2447STheodore Ts'o #ifdef CONFIG_EXT4_FS_ENCRYPTION 16646ddb2447STheodore Ts'o sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION; 16656ddb2447STheodore Ts'o ext4_msg(sb, KERN_WARNING, 16666ddb2447STheodore Ts'o "Test dummy encryption mode enabled"); 16676ddb2447STheodore Ts'o #else 16686ddb2447STheodore Ts'o ext4_msg(sb, KERN_WARNING, 16696ddb2447STheodore Ts'o "Test dummy encryption mount option ignored"); 16706ddb2447STheodore Ts'o #endif 167126092bf5STheodore Ts'o } else if (m->flags & MOPT_DATAJ) { 167226092bf5STheodore Ts'o if (is_remount) { 167326092bf5STheodore Ts'o if (!sbi->s_journal) 167426092bf5STheodore Ts'o ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 16755f3633e3SJan Kara else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 167626092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, 167726092bf5STheodore Ts'o "Cannot change data mode on remount"); 167826092bf5STheodore Ts'o return -1; 167926092bf5STheodore Ts'o } 168026092bf5STheodore Ts'o } else { 168126092bf5STheodore Ts'o clear_opt(sb, DATA_FLAGS); 168226092bf5STheodore Ts'o sbi->s_mount_opt |= m->mount_opt; 168326092bf5STheodore Ts'o } 168426092bf5STheodore Ts'o #ifdef CONFIG_QUOTA 168526092bf5STheodore Ts'o } else if (m->flags & MOPT_QFMT) { 168626092bf5STheodore Ts'o if (sb_any_quota_loaded(sb) && 168726092bf5STheodore Ts'o sbi->s_jquota_fmt != m->mount_opt) { 16885f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Cannot change journaled " 16895f3633e3SJan Kara "quota options when quota turned on"); 169026092bf5STheodore Ts'o return -1; 169126092bf5STheodore Ts'o } 1692e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) { 1693c325a67cSTheodore Ts'o ext4_msg(sb, KERN_INFO, 1694c325a67cSTheodore Ts'o "Quota format mount options ignored " 1695262b4662SJan Kara "when QUOTA feature is enabled"); 1696c325a67cSTheodore Ts'o return 1; 1697262b4662SJan Kara } 169826092bf5STheodore Ts'o sbi->s_jquota_fmt = m->mount_opt; 169926092bf5STheodore Ts'o #endif 1700923ae0ffSRoss Zwisler } else if (token == Opt_dax) { 1701ef83b6e8SDan Williams #ifdef CONFIG_FS_DAX 1702ef83b6e8SDan Williams ext4_msg(sb, KERN_WARNING, 1703ef83b6e8SDan Williams "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); 1704ef83b6e8SDan Williams sbi->s_mount_opt |= m->mount_opt; 1705ef83b6e8SDan Williams #else 1706923ae0ffSRoss Zwisler ext4_msg(sb, KERN_INFO, "dax option not supported"); 1707923ae0ffSRoss Zwisler return -1; 1708923ae0ffSRoss Zwisler #endif 17097915a861SAles Novak } else if (token == Opt_data_err_abort) { 17107915a861SAles Novak sbi->s_mount_opt |= m->mount_opt; 17117915a861SAles Novak } else if (token == Opt_data_err_ignore) { 17127915a861SAles Novak sbi->s_mount_opt &= ~m->mount_opt; 171326092bf5STheodore Ts'o } else { 171426092bf5STheodore Ts'o if (!args->from) 171526092bf5STheodore Ts'o arg = 1; 171626092bf5STheodore Ts'o if (m->flags & MOPT_CLEAR) 171726092bf5STheodore Ts'o arg = !arg; 171826092bf5STheodore Ts'o else if (unlikely(!(m->flags & MOPT_SET))) { 171926092bf5STheodore Ts'o ext4_msg(sb, KERN_WARNING, 172026092bf5STheodore Ts'o "buggy handling of option %s", opt); 172126092bf5STheodore Ts'o WARN_ON(1); 172226092bf5STheodore Ts'o return -1; 172326092bf5STheodore Ts'o } 172426092bf5STheodore Ts'o if (arg != 0) 172526092bf5STheodore Ts'o sbi->s_mount_opt |= m->mount_opt; 172626092bf5STheodore Ts'o else 172726092bf5STheodore Ts'o sbi->s_mount_opt &= ~m->mount_opt; 172826092bf5STheodore Ts'o } 172926092bf5STheodore Ts'o return 1; 173026092bf5STheodore Ts'o } 173126092bf5STheodore Ts'o 1732ac27a0ecSDave Kleikamp static int parse_options(char *options, struct super_block *sb, 1733c3191067STheodore Ts'o unsigned long *journal_devnum, 1734b3881f74STheodore Ts'o unsigned int *journal_ioprio, 1735661aa520SEric Sandeen int is_remount) 1736ac27a0ecSDave Kleikamp { 1737617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 1738ac27a0ecSDave Kleikamp char *p; 1739ac27a0ecSDave Kleikamp substring_t args[MAX_OPT_ARGS]; 174026092bf5STheodore Ts'o int token; 1741ac27a0ecSDave Kleikamp 1742ac27a0ecSDave Kleikamp if (!options) 1743ac27a0ecSDave Kleikamp return 1; 1744ac27a0ecSDave Kleikamp 1745ac27a0ecSDave Kleikamp while ((p = strsep(&options, ",")) != NULL) { 1746ac27a0ecSDave Kleikamp if (!*p) 1747ac27a0ecSDave Kleikamp continue; 174815121c18SEric Sandeen /* 174915121c18SEric Sandeen * Initialize args struct so we know whether arg was 175015121c18SEric Sandeen * found; some options take optional arguments. 175115121c18SEric Sandeen */ 1752caecd0afSSachin Kamat args[0].to = args[0].from = NULL; 1753ac27a0ecSDave Kleikamp token = match_token(p, tokens, args); 175426092bf5STheodore Ts'o if (handle_mount_opt(sb, p, token, args, journal_devnum, 175526092bf5STheodore Ts'o journal_ioprio, is_remount) < 0) 1756ac27a0ecSDave Kleikamp return 0; 1757ac27a0ecSDave Kleikamp } 1758ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1759e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb) && 1760262b4662SJan Kara (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) { 1761c325a67cSTheodore Ts'o ext4_msg(sb, KERN_INFO, "Quota feature enabled, usrquota and grpquota " 1762c325a67cSTheodore Ts'o "mount options ignored."); 1763c325a67cSTheodore Ts'o clear_opt(sb, USRQUOTA); 1764c325a67cSTheodore Ts'o clear_opt(sb, GRPQUOTA); 1765c325a67cSTheodore Ts'o } else if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 1766482a7425SDmitry Monakhov if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) 1767fd8c37ecSTheodore Ts'o clear_opt(sb, USRQUOTA); 1768ac27a0ecSDave Kleikamp 1769482a7425SDmitry Monakhov if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) 1770fd8c37ecSTheodore Ts'o clear_opt(sb, GRPQUOTA); 1771ac27a0ecSDave Kleikamp 177256c50f11SDmitry Monakhov if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 1773b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "old and new quota " 1774b31e1552SEric Sandeen "format mixing"); 1775ac27a0ecSDave Kleikamp return 0; 1776ac27a0ecSDave Kleikamp } 1777ac27a0ecSDave Kleikamp 1778ac27a0ecSDave Kleikamp if (!sbi->s_jquota_fmt) { 1779b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journaled quota format " 1780b31e1552SEric Sandeen "not specified"); 1781ac27a0ecSDave Kleikamp return 0; 1782ac27a0ecSDave Kleikamp } 1783ac27a0ecSDave Kleikamp } 1784ac27a0ecSDave Kleikamp #endif 1785261cb20cSJan Kara if (test_opt(sb, DIOREAD_NOLOCK)) { 1786261cb20cSJan Kara int blocksize = 1787261cb20cSJan Kara BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 1788261cb20cSJan Kara 178909cbfeafSKirill A. Shutemov if (blocksize < PAGE_SIZE) { 1790261cb20cSJan Kara ext4_msg(sb, KERN_ERR, "can't mount with " 1791261cb20cSJan Kara "dioread_nolock if block size != PAGE_SIZE"); 1792261cb20cSJan Kara return 0; 1793261cb20cSJan Kara } 1794261cb20cSJan Kara } 1795d4f76107SJan Kara if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 1796d4f76107SJan Kara test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 1797d4f76107SJan Kara ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit " 1798d4f76107SJan Kara "in data=ordered mode"); 1799d4f76107SJan Kara return 0; 1800d4f76107SJan Kara } 1801ac27a0ecSDave Kleikamp return 1; 1802ac27a0ecSDave Kleikamp } 1803ac27a0ecSDave Kleikamp 18042adf6da8STheodore Ts'o static inline void ext4_show_quota_options(struct seq_file *seq, 18052adf6da8STheodore Ts'o struct super_block *sb) 18062adf6da8STheodore Ts'o { 18072adf6da8STheodore Ts'o #if defined(CONFIG_QUOTA) 18082adf6da8STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 18092adf6da8STheodore Ts'o 18102adf6da8STheodore Ts'o if (sbi->s_jquota_fmt) { 18112adf6da8STheodore Ts'o char *fmtname = ""; 18122adf6da8STheodore Ts'o 18132adf6da8STheodore Ts'o switch (sbi->s_jquota_fmt) { 18142adf6da8STheodore Ts'o case QFMT_VFS_OLD: 18152adf6da8STheodore Ts'o fmtname = "vfsold"; 18162adf6da8STheodore Ts'o break; 18172adf6da8STheodore Ts'o case QFMT_VFS_V0: 18182adf6da8STheodore Ts'o fmtname = "vfsv0"; 18192adf6da8STheodore Ts'o break; 18202adf6da8STheodore Ts'o case QFMT_VFS_V1: 18212adf6da8STheodore Ts'o fmtname = "vfsv1"; 18222adf6da8STheodore Ts'o break; 18232adf6da8STheodore Ts'o } 18242adf6da8STheodore Ts'o seq_printf(seq, ",jqfmt=%s", fmtname); 18252adf6da8STheodore Ts'o } 18262adf6da8STheodore Ts'o 18272adf6da8STheodore Ts'o if (sbi->s_qf_names[USRQUOTA]) 1828a068acf2SKees Cook seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]); 18292adf6da8STheodore Ts'o 18302adf6da8STheodore Ts'o if (sbi->s_qf_names[GRPQUOTA]) 1831a068acf2SKees Cook seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]); 18322adf6da8STheodore Ts'o #endif 18332adf6da8STheodore Ts'o } 18342adf6da8STheodore Ts'o 18355a916be1STheodore Ts'o static const char *token2str(int token) 18365a916be1STheodore Ts'o { 183750df9fd5SHerton Ronaldo Krzesinski const struct match_token *t; 18385a916be1STheodore Ts'o 18395a916be1STheodore Ts'o for (t = tokens; t->token != Opt_err; t++) 18405a916be1STheodore Ts'o if (t->token == token && !strchr(t->pattern, '=')) 18415a916be1STheodore Ts'o break; 18425a916be1STheodore Ts'o return t->pattern; 18435a916be1STheodore Ts'o } 18445a916be1STheodore Ts'o 18452adf6da8STheodore Ts'o /* 18462adf6da8STheodore Ts'o * Show an option if 18472adf6da8STheodore Ts'o * - it's set to a non-default value OR 18482adf6da8STheodore Ts'o * - if the per-sb default is different from the global default 18492adf6da8STheodore Ts'o */ 185066acdcf4STheodore Ts'o static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 185166acdcf4STheodore Ts'o int nodefs) 18522adf6da8STheodore Ts'o { 18532adf6da8STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 18542adf6da8STheodore Ts'o struct ext4_super_block *es = sbi->s_es; 185566acdcf4STheodore Ts'o int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt; 18565a916be1STheodore Ts'o const struct mount_opts *m; 185766acdcf4STheodore Ts'o char sep = nodefs ? '\n' : ','; 18582adf6da8STheodore Ts'o 185966acdcf4STheodore Ts'o #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 186066acdcf4STheodore Ts'o #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 18612adf6da8STheodore Ts'o 18622adf6da8STheodore Ts'o if (sbi->s_sb_block != 1) 18635a916be1STheodore Ts'o SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 18645a916be1STheodore Ts'o 18655a916be1STheodore Ts'o for (m = ext4_mount_opts; m->token != Opt_err; m++) { 18665a916be1STheodore Ts'o int want_set = m->flags & MOPT_SET; 18675a916be1STheodore Ts'o if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 18685a916be1STheodore Ts'o (m->flags & MOPT_CLEAR_ERR)) 18695a916be1STheodore Ts'o continue; 187066acdcf4STheodore Ts'o if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 18715a916be1STheodore Ts'o continue; /* skip if same as the default */ 18725a916be1STheodore Ts'o if ((want_set && 18735a916be1STheodore Ts'o (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 18745a916be1STheodore Ts'o (!want_set && (sbi->s_mount_opt & m->mount_opt))) 18755a916be1STheodore Ts'o continue; /* select Opt_noFoo vs Opt_Foo */ 18765a916be1STheodore Ts'o SEQ_OPTS_PRINT("%s", token2str(m->token)); 18775a916be1STheodore Ts'o } 18785a916be1STheodore Ts'o 187908cefc7aSEric W. Biederman if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 18805a916be1STheodore Ts'o le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 188108cefc7aSEric W. Biederman SEQ_OPTS_PRINT("resuid=%u", 188208cefc7aSEric W. Biederman from_kuid_munged(&init_user_ns, sbi->s_resuid)); 188308cefc7aSEric W. Biederman if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 18845a916be1STheodore Ts'o le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 188508cefc7aSEric W. Biederman SEQ_OPTS_PRINT("resgid=%u", 188608cefc7aSEric W. Biederman from_kgid_munged(&init_user_ns, sbi->s_resgid)); 188766acdcf4STheodore Ts'o def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 18885a916be1STheodore Ts'o if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 18895a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=remount-ro"); 18902adf6da8STheodore Ts'o if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 18915a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=continue"); 18922adf6da8STheodore Ts'o if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 18935a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=panic"); 189466acdcf4STheodore Ts'o if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 18955a916be1STheodore Ts'o SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 189666acdcf4STheodore Ts'o if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 18975a916be1STheodore Ts'o SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 189866acdcf4STheodore Ts'o if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 18995a916be1STheodore Ts'o SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 19002adf6da8STheodore Ts'o if (sb->s_flags & MS_I_VERSION) 19015a916be1STheodore Ts'o SEQ_OPTS_PUTS("i_version"); 190266acdcf4STheodore Ts'o if (nodefs || sbi->s_stripe) 19035a916be1STheodore Ts'o SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 190466acdcf4STheodore Ts'o if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) { 19052adf6da8STheodore Ts'o if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 19065a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=journal"); 19072adf6da8STheodore Ts'o else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 19085a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=ordered"); 19092adf6da8STheodore Ts'o else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 19105a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=writeback"); 19115a916be1STheodore Ts'o } 191266acdcf4STheodore Ts'o if (nodefs || 191366acdcf4STheodore Ts'o sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 19145a916be1STheodore Ts'o SEQ_OPTS_PRINT("inode_readahead_blks=%u", 19152adf6da8STheodore Ts'o sbi->s_inode_readahead_blks); 19162adf6da8STheodore Ts'o 191766acdcf4STheodore Ts'o if (nodefs || (test_opt(sb, INIT_INODE_TABLE) && 191866acdcf4STheodore Ts'o (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 19195a916be1STheodore Ts'o SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 1920df981d03STheodore Ts'o if (nodefs || sbi->s_max_dir_size_kb) 1921df981d03STheodore Ts'o SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 19227915a861SAles Novak if (test_opt(sb, DATA_ERR_ABORT)) 19237915a861SAles Novak SEQ_OPTS_PUTS("data_err=abort"); 19242adf6da8STheodore Ts'o 19252adf6da8STheodore Ts'o ext4_show_quota_options(seq, sb); 19262adf6da8STheodore Ts'o return 0; 19272adf6da8STheodore Ts'o } 19282adf6da8STheodore Ts'o 192966acdcf4STheodore Ts'o static int ext4_show_options(struct seq_file *seq, struct dentry *root) 193066acdcf4STheodore Ts'o { 193166acdcf4STheodore Ts'o return _ext4_show_options(seq, root->d_sb, 0); 193266acdcf4STheodore Ts'o } 193366acdcf4STheodore Ts'o 1934ebd173beSTheodore Ts'o int ext4_seq_options_show(struct seq_file *seq, void *offset) 193566acdcf4STheodore Ts'o { 193666acdcf4STheodore Ts'o struct super_block *sb = seq->private; 193766acdcf4STheodore Ts'o int rc; 193866acdcf4STheodore Ts'o 193966acdcf4STheodore Ts'o seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw"); 194066acdcf4STheodore Ts'o rc = _ext4_show_options(seq, sb, 1); 194166acdcf4STheodore Ts'o seq_puts(seq, "\n"); 194266acdcf4STheodore Ts'o return rc; 194366acdcf4STheodore Ts'o } 194466acdcf4STheodore Ts'o 1945617ba13bSMingming Cao static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 1946ac27a0ecSDave Kleikamp int read_only) 1947ac27a0ecSDave Kleikamp { 1948617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 1949ac27a0ecSDave Kleikamp int res = 0; 1950ac27a0ecSDave Kleikamp 1951617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 1952b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "revision level too high, " 1953b31e1552SEric Sandeen "forcing read-only mode"); 1954ac27a0ecSDave Kleikamp res = MS_RDONLY; 1955ac27a0ecSDave Kleikamp } 1956ac27a0ecSDave Kleikamp if (read_only) 1957281b5995STheodore Ts'o goto done; 1958617ba13bSMingming Cao if (!(sbi->s_mount_state & EXT4_VALID_FS)) 1959b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 1960b31e1552SEric Sandeen "running e2fsck is recommended"); 1961c8b459f4SLukas Czerner else if (sbi->s_mount_state & EXT4_ERROR_FS) 1962b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 1963b31e1552SEric Sandeen "warning: mounting fs with errors, " 1964b31e1552SEric Sandeen "running e2fsck is recommended"); 1965ed3ce80aSTao Ma else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 1966ac27a0ecSDave Kleikamp le16_to_cpu(es->s_mnt_count) >= 1967ac27a0ecSDave Kleikamp (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 1968b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 1969b31e1552SEric Sandeen "warning: maximal mount count reached, " 1970b31e1552SEric Sandeen "running e2fsck is recommended"); 1971ac27a0ecSDave Kleikamp else if (le32_to_cpu(es->s_checkinterval) && 1972ac27a0ecSDave Kleikamp (le32_to_cpu(es->s_lastcheck) + 1973ac27a0ecSDave Kleikamp le32_to_cpu(es->s_checkinterval) <= get_seconds())) 1974b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 1975b31e1552SEric Sandeen "warning: checktime reached, " 1976b31e1552SEric Sandeen "running e2fsck is recommended"); 19770390131bSFrank Mayhar if (!sbi->s_journal) 1978216c34b2SMarcin Slusarz es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 1979ac27a0ecSDave Kleikamp if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 1980617ba13bSMingming Cao es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 1981e8546d06SMarcin Slusarz le16_add_cpu(&es->s_mnt_count, 1); 1982ac27a0ecSDave Kleikamp es->s_mtime = cpu_to_le32(get_seconds()); 1983617ba13bSMingming Cao ext4_update_dynamic_rev(sb); 19840390131bSFrank Mayhar if (sbi->s_journal) 1985e2b911c5SDarrick J. Wong ext4_set_feature_journal_needs_recovery(sb); 1986ac27a0ecSDave Kleikamp 1987e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 1988281b5995STheodore Ts'o done: 1989ac27a0ecSDave Kleikamp if (test_opt(sb, DEBUG)) 1990a9df9a49STheodore Ts'o printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 1991a2595b8aSTheodore Ts'o "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 1992ac27a0ecSDave Kleikamp sb->s_blocksize, 1993ac27a0ecSDave Kleikamp sbi->s_groups_count, 1994617ba13bSMingming Cao EXT4_BLOCKS_PER_GROUP(sb), 1995617ba13bSMingming Cao EXT4_INODES_PER_GROUP(sb), 1996a2595b8aSTheodore Ts'o sbi->s_mount_opt, sbi->s_mount_opt2); 1997ac27a0ecSDave Kleikamp 19987abc52c2SDan Magenheimer cleancache_init_fs(sb); 1999ac27a0ecSDave Kleikamp return res; 2000ac27a0ecSDave Kleikamp } 2001ac27a0ecSDave Kleikamp 2002117fff10STheodore Ts'o int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 2003117fff10STheodore Ts'o { 2004117fff10STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 2005117fff10STheodore Ts'o struct flex_groups *new_groups; 2006117fff10STheodore Ts'o int size; 2007117fff10STheodore Ts'o 2008117fff10STheodore Ts'o if (!sbi->s_log_groups_per_flex) 2009117fff10STheodore Ts'o return 0; 2010117fff10STheodore Ts'o 2011117fff10STheodore Ts'o size = ext4_flex_group(sbi, ngroup - 1) + 1; 2012117fff10STheodore Ts'o if (size <= sbi->s_flex_groups_allocated) 2013117fff10STheodore Ts'o return 0; 2014117fff10STheodore Ts'o 2015117fff10STheodore Ts'o size = roundup_pow_of_two(size * sizeof(struct flex_groups)); 2016117fff10STheodore Ts'o new_groups = ext4_kvzalloc(size, GFP_KERNEL); 2017117fff10STheodore Ts'o if (!new_groups) { 2018117fff10STheodore Ts'o ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups", 2019117fff10STheodore Ts'o size / (int) sizeof(struct flex_groups)); 2020117fff10STheodore Ts'o return -ENOMEM; 2021117fff10STheodore Ts'o } 2022117fff10STheodore Ts'o 2023117fff10STheodore Ts'o if (sbi->s_flex_groups) { 2024117fff10STheodore Ts'o memcpy(new_groups, sbi->s_flex_groups, 2025117fff10STheodore Ts'o (sbi->s_flex_groups_allocated * 2026117fff10STheodore Ts'o sizeof(struct flex_groups))); 2027b93b41d4SAl Viro kvfree(sbi->s_flex_groups); 2028117fff10STheodore Ts'o } 2029117fff10STheodore Ts'o sbi->s_flex_groups = new_groups; 2030117fff10STheodore Ts'o sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups); 2031117fff10STheodore Ts'o return 0; 2032117fff10STheodore Ts'o } 2033117fff10STheodore Ts'o 2034772cb7c8SJose R. Santos static int ext4_fill_flex_info(struct super_block *sb) 2035772cb7c8SJose R. Santos { 2036772cb7c8SJose R. Santos struct ext4_sb_info *sbi = EXT4_SB(sb); 2037772cb7c8SJose R. Santos struct ext4_group_desc *gdp = NULL; 2038772cb7c8SJose R. Santos ext4_group_t flex_group; 2039117fff10STheodore Ts'o int i, err; 2040772cb7c8SJose R. Santos 2041503358aeSTheodore Ts'o sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 2042d50f2ab6SXi Wang if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 2043772cb7c8SJose R. Santos sbi->s_log_groups_per_flex = 0; 2044772cb7c8SJose R. Santos return 1; 2045772cb7c8SJose R. Santos } 2046772cb7c8SJose R. Santos 2047117fff10STheodore Ts'o err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 2048117fff10STheodore Ts'o if (err) 2049772cb7c8SJose R. Santos goto failed; 2050772cb7c8SJose R. Santos 2051772cb7c8SJose R. Santos for (i = 0; i < sbi->s_groups_count; i++) { 205288b6edd1STheodore Ts'o gdp = ext4_get_group_desc(sb, i, NULL); 2053772cb7c8SJose R. Santos 2054772cb7c8SJose R. Santos flex_group = ext4_flex_group(sbi, i); 20557ad9bb65STheodore Ts'o atomic_add(ext4_free_inodes_count(sb, gdp), 20567ad9bb65STheodore Ts'o &sbi->s_flex_groups[flex_group].free_inodes); 205790ba983fSTheodore Ts'o atomic64_add(ext4_free_group_clusters(sb, gdp), 205824aaa8efSTheodore Ts'o &sbi->s_flex_groups[flex_group].free_clusters); 20597ad9bb65STheodore Ts'o atomic_add(ext4_used_dirs_count(sb, gdp), 20607ad9bb65STheodore Ts'o &sbi->s_flex_groups[flex_group].used_dirs); 2061772cb7c8SJose R. Santos } 2062772cb7c8SJose R. Santos 2063772cb7c8SJose R. Santos return 1; 2064772cb7c8SJose R. Santos failed: 2065772cb7c8SJose R. Santos return 0; 2066772cb7c8SJose R. Santos } 2067772cb7c8SJose R. Santos 2068e2b911c5SDarrick J. Wong static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 2069717d50e4SAndreas Dilger struct ext4_group_desc *gdp) 2070717d50e4SAndreas Dilger { 2071*b47820edSDaeho Jeong int offset = offsetof(struct ext4_group_desc, bg_checksum); 2072717d50e4SAndreas Dilger __u16 crc = 0; 2073717d50e4SAndreas Dilger __le32 le_group = cpu_to_le32(block_group); 2074e2b911c5SDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 2075717d50e4SAndreas Dilger 20769aa5d32bSDmitry Monakhov if (ext4_has_metadata_csum(sbi->s_sb)) { 2077feb0ab32SDarrick J. Wong /* Use new metadata_csum algorithm */ 2078feb0ab32SDarrick J. Wong __u32 csum32; 2079*b47820edSDaeho Jeong __u16 dummy_csum = 0; 2080feb0ab32SDarrick J. Wong 2081feb0ab32SDarrick J. Wong csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 2082feb0ab32SDarrick J. Wong sizeof(le_group)); 2083*b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 2084*b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 2085*b47820edSDaeho Jeong sizeof(dummy_csum)); 2086*b47820edSDaeho Jeong offset += sizeof(dummy_csum); 2087*b47820edSDaeho Jeong if (offset < sbi->s_desc_size) 2088*b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 2089*b47820edSDaeho Jeong sbi->s_desc_size - offset); 2090feb0ab32SDarrick J. Wong 2091feb0ab32SDarrick J. Wong crc = csum32 & 0xFFFF; 2092feb0ab32SDarrick J. Wong goto out; 2093feb0ab32SDarrick J. Wong } 2094feb0ab32SDarrick J. Wong 2095feb0ab32SDarrick J. Wong /* old crc16 code */ 2096e2b911c5SDarrick J. Wong if (!ext4_has_feature_gdt_csum(sb)) 2097813d32f9SDarrick J. Wong return 0; 2098813d32f9SDarrick J. Wong 2099717d50e4SAndreas Dilger crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2100717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2101717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)gdp, offset); 2102717d50e4SAndreas Dilger offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2103717d50e4SAndreas Dilger /* for checksum of struct ext4_group_desc do the rest...*/ 2104e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb) && 2105717d50e4SAndreas Dilger offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2106717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)gdp + offset, 2107717d50e4SAndreas Dilger le16_to_cpu(sbi->s_es->s_desc_size) - 2108717d50e4SAndreas Dilger offset); 2109717d50e4SAndreas Dilger 2110feb0ab32SDarrick J. Wong out: 2111717d50e4SAndreas Dilger return cpu_to_le16(crc); 2112717d50e4SAndreas Dilger } 2113717d50e4SAndreas Dilger 2114feb0ab32SDarrick J. Wong int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2115717d50e4SAndreas Dilger struct ext4_group_desc *gdp) 2116717d50e4SAndreas Dilger { 2117feb0ab32SDarrick J. Wong if (ext4_has_group_desc_csum(sb) && 2118e2b911c5SDarrick J. Wong (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 2119717d50e4SAndreas Dilger return 0; 2120717d50e4SAndreas Dilger 2121717d50e4SAndreas Dilger return 1; 2122717d50e4SAndreas Dilger } 2123717d50e4SAndreas Dilger 2124feb0ab32SDarrick J. Wong void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2125feb0ab32SDarrick J. Wong struct ext4_group_desc *gdp) 2126feb0ab32SDarrick J. Wong { 2127feb0ab32SDarrick J. Wong if (!ext4_has_group_desc_csum(sb)) 2128feb0ab32SDarrick J. Wong return; 2129e2b911c5SDarrick J. Wong gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 2130feb0ab32SDarrick J. Wong } 2131feb0ab32SDarrick J. Wong 2132ac27a0ecSDave Kleikamp /* Called at mount-time, super-block is locked */ 2133bfff6873SLukas Czerner static int ext4_check_descriptors(struct super_block *sb, 2134bfff6873SLukas Czerner ext4_group_t *first_not_zeroed) 2135ac27a0ecSDave Kleikamp { 2136617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 2137617ba13bSMingming Cao ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2138617ba13bSMingming Cao ext4_fsblk_t last_block; 2139bd81d8eeSLaurent Vivier ext4_fsblk_t block_bitmap; 2140bd81d8eeSLaurent Vivier ext4_fsblk_t inode_bitmap; 2141bd81d8eeSLaurent Vivier ext4_fsblk_t inode_table; 2142ce421581SJose R. Santos int flexbg_flag = 0; 2143bfff6873SLukas Czerner ext4_group_t i, grp = sbi->s_groups_count; 2144ac27a0ecSDave Kleikamp 2145e2b911c5SDarrick J. Wong if (ext4_has_feature_flex_bg(sb)) 2146ce421581SJose R. Santos flexbg_flag = 1; 2147ce421581SJose R. Santos 2148617ba13bSMingming Cao ext4_debug("Checking group descriptors"); 2149ac27a0ecSDave Kleikamp 2150197cd65aSAkinobu Mita for (i = 0; i < sbi->s_groups_count; i++) { 2151197cd65aSAkinobu Mita struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2152197cd65aSAkinobu Mita 2153ce421581SJose R. Santos if (i == sbi->s_groups_count - 1 || flexbg_flag) 2154bd81d8eeSLaurent Vivier last_block = ext4_blocks_count(sbi->s_es) - 1; 2155ac27a0ecSDave Kleikamp else 2156ac27a0ecSDave Kleikamp last_block = first_block + 2157617ba13bSMingming Cao (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2158ac27a0ecSDave Kleikamp 2159bfff6873SLukas Czerner if ((grp == sbi->s_groups_count) && 2160bfff6873SLukas Czerner !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2161bfff6873SLukas Czerner grp = i; 2162bfff6873SLukas Czerner 21638fadc143SAlexandre Ratchov block_bitmap = ext4_block_bitmap(sb, gdp); 21642b2d6d01STheodore Ts'o if (block_bitmap < first_block || block_bitmap > last_block) { 2165b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2166a9df9a49STheodore Ts'o "Block bitmap for group %u not in group " 2167b31e1552SEric Sandeen "(block %llu)!", i, block_bitmap); 2168ac27a0ecSDave Kleikamp return 0; 2169ac27a0ecSDave Kleikamp } 21708fadc143SAlexandre Ratchov inode_bitmap = ext4_inode_bitmap(sb, gdp); 21712b2d6d01STheodore Ts'o if (inode_bitmap < first_block || inode_bitmap > last_block) { 2172b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2173a9df9a49STheodore Ts'o "Inode bitmap for group %u not in group " 2174b31e1552SEric Sandeen "(block %llu)!", i, inode_bitmap); 2175ac27a0ecSDave Kleikamp return 0; 2176ac27a0ecSDave Kleikamp } 21778fadc143SAlexandre Ratchov inode_table = ext4_inode_table(sb, gdp); 2178bd81d8eeSLaurent Vivier if (inode_table < first_block || 21792b2d6d01STheodore Ts'o inode_table + sbi->s_itb_per_group - 1 > last_block) { 2180b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2181a9df9a49STheodore Ts'o "Inode table for group %u not in group " 2182b31e1552SEric Sandeen "(block %llu)!", i, inode_table); 2183ac27a0ecSDave Kleikamp return 0; 2184ac27a0ecSDave Kleikamp } 2185955ce5f5SAneesh Kumar K.V ext4_lock_group(sb, i); 2186feb0ab32SDarrick J. Wong if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2187b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2188b31e1552SEric Sandeen "Checksum for group %u failed (%u!=%u)", 2189e2b911c5SDarrick J. Wong i, le16_to_cpu(ext4_group_desc_csum(sb, i, 2190fd2d4291SAvantika Mathur gdp)), le16_to_cpu(gdp->bg_checksum)); 21917ee1ec4cSLi Zefan if (!(sb->s_flags & MS_RDONLY)) { 2192955ce5f5SAneesh Kumar K.V ext4_unlock_group(sb, i); 2193717d50e4SAndreas Dilger return 0; 2194717d50e4SAndreas Dilger } 21957ee1ec4cSLi Zefan } 2196955ce5f5SAneesh Kumar K.V ext4_unlock_group(sb, i); 2197ce421581SJose R. Santos if (!flexbg_flag) 2198617ba13bSMingming Cao first_block += EXT4_BLOCKS_PER_GROUP(sb); 2199ac27a0ecSDave Kleikamp } 2200bfff6873SLukas Czerner if (NULL != first_not_zeroed) 2201bfff6873SLukas Czerner *first_not_zeroed = grp; 2202ac27a0ecSDave Kleikamp return 1; 2203ac27a0ecSDave Kleikamp } 2204ac27a0ecSDave Kleikamp 2205617ba13bSMingming Cao /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2206ac27a0ecSDave Kleikamp * the superblock) which were deleted from all directories, but held open by 2207ac27a0ecSDave Kleikamp * a process at the time of a crash. We walk the list and try to delete these 2208ac27a0ecSDave Kleikamp * inodes at recovery time (only with a read-write filesystem). 2209ac27a0ecSDave Kleikamp * 2210ac27a0ecSDave Kleikamp * In order to keep the orphan inode chain consistent during traversal (in 2211ac27a0ecSDave Kleikamp * case of crash during recovery), we link each inode into the superblock 2212ac27a0ecSDave Kleikamp * orphan list_head and handle it the same way as an inode deletion during 2213ac27a0ecSDave Kleikamp * normal operation (which journals the operations for us). 2214ac27a0ecSDave Kleikamp * 2215ac27a0ecSDave Kleikamp * We only do an iget() and an iput() on each inode, which is very safe if we 2216ac27a0ecSDave Kleikamp * accidentally point at an in-use or already deleted inode. The worst that 2217ac27a0ecSDave Kleikamp * can happen in this case is that we get a "bit already cleared" message from 2218617ba13bSMingming Cao * ext4_free_inode(). The only reason we would point at a wrong inode is if 2219ac27a0ecSDave Kleikamp * e2fsck was run on this filesystem, and it must have already done the orphan 2220ac27a0ecSDave Kleikamp * inode cleanup for us, so we can safely abort without any further action. 2221ac27a0ecSDave Kleikamp */ 2222617ba13bSMingming Cao static void ext4_orphan_cleanup(struct super_block *sb, 2223617ba13bSMingming Cao struct ext4_super_block *es) 2224ac27a0ecSDave Kleikamp { 2225ac27a0ecSDave Kleikamp unsigned int s_flags = sb->s_flags; 2226ac27a0ecSDave Kleikamp int nr_orphans = 0, nr_truncates = 0; 2227ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 2228ac27a0ecSDave Kleikamp int i; 2229ac27a0ecSDave Kleikamp #endif 2230ac27a0ecSDave Kleikamp if (!es->s_last_orphan) { 2231ac27a0ecSDave Kleikamp jbd_debug(4, "no orphan inodes to clean up\n"); 2232ac27a0ecSDave Kleikamp return; 2233ac27a0ecSDave Kleikamp } 2234ac27a0ecSDave Kleikamp 2235a8f48a95SEric Sandeen if (bdev_read_only(sb->s_bdev)) { 2236b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "write access " 2237b31e1552SEric Sandeen "unavailable, skipping orphan cleanup"); 2238a8f48a95SEric Sandeen return; 2239a8f48a95SEric Sandeen } 2240a8f48a95SEric Sandeen 2241d39195c3SAmir Goldstein /* Check if feature set would not allow a r/w mount */ 2242d39195c3SAmir Goldstein if (!ext4_feature_set_ok(sb, 0)) { 2243d39195c3SAmir Goldstein ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2244d39195c3SAmir Goldstein "unknown ROCOMPAT features"); 2245d39195c3SAmir Goldstein return; 2246d39195c3SAmir Goldstein } 2247d39195c3SAmir Goldstein 2248617ba13bSMingming Cao if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2249c25f9bc6SEric Sandeen /* don't clear list on RO mount w/ errors */ 2250c25f9bc6SEric Sandeen if (es->s_last_orphan && !(s_flags & MS_RDONLY)) { 225184474976SDmitry Monakhov ext4_msg(sb, KERN_INFO, "Errors on filesystem, " 2252ac27a0ecSDave Kleikamp "clearing orphan list.\n"); 2253ac27a0ecSDave Kleikamp es->s_last_orphan = 0; 2254c25f9bc6SEric Sandeen } 2255ac27a0ecSDave Kleikamp jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2256ac27a0ecSDave Kleikamp return; 2257ac27a0ecSDave Kleikamp } 2258ac27a0ecSDave Kleikamp 2259ac27a0ecSDave Kleikamp if (s_flags & MS_RDONLY) { 2260b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 2261ac27a0ecSDave Kleikamp sb->s_flags &= ~MS_RDONLY; 2262ac27a0ecSDave Kleikamp } 2263ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 2264ac27a0ecSDave Kleikamp /* Needed for iput() to work correctly and not trash data */ 2265ac27a0ecSDave Kleikamp sb->s_flags |= MS_ACTIVE; 2266ac27a0ecSDave Kleikamp /* Turn on quotas so that they are updated correctly */ 2267a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2268617ba13bSMingming Cao if (EXT4_SB(sb)->s_qf_names[i]) { 2269617ba13bSMingming Cao int ret = ext4_quota_on_mount(sb, i); 2270ac27a0ecSDave Kleikamp if (ret < 0) 2271b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 2272b31e1552SEric Sandeen "Cannot turn on journaled " 2273b31e1552SEric Sandeen "quota: error %d", ret); 2274ac27a0ecSDave Kleikamp } 2275ac27a0ecSDave Kleikamp } 2276ac27a0ecSDave Kleikamp #endif 2277ac27a0ecSDave Kleikamp 2278ac27a0ecSDave Kleikamp while (es->s_last_orphan) { 2279ac27a0ecSDave Kleikamp struct inode *inode; 2280ac27a0ecSDave Kleikamp 228197bd42b9SJosef Bacik inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 228297bd42b9SJosef Bacik if (IS_ERR(inode)) { 2283ac27a0ecSDave Kleikamp es->s_last_orphan = 0; 2284ac27a0ecSDave Kleikamp break; 2285ac27a0ecSDave Kleikamp } 2286ac27a0ecSDave Kleikamp 2287617ba13bSMingming Cao list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2288871a2931SChristoph Hellwig dquot_initialize(inode); 2289ac27a0ecSDave Kleikamp if (inode->i_nlink) { 2290566370a2SPaul Taysom if (test_opt(sb, DEBUG)) 2291b31e1552SEric Sandeen ext4_msg(sb, KERN_DEBUG, 2292b31e1552SEric Sandeen "%s: truncating inode %lu to %lld bytes", 229346e665e9SHarvey Harrison __func__, inode->i_ino, inode->i_size); 2294e5f8eab8STheodore Ts'o jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2295ac27a0ecSDave Kleikamp inode->i_ino, inode->i_size); 22965955102cSAl Viro inode_lock(inode); 229755f252c9SLukas Czerner truncate_inode_pages(inode->i_mapping, inode->i_size); 2298617ba13bSMingming Cao ext4_truncate(inode); 22995955102cSAl Viro inode_unlock(inode); 2300ac27a0ecSDave Kleikamp nr_truncates++; 2301ac27a0ecSDave Kleikamp } else { 2302566370a2SPaul Taysom if (test_opt(sb, DEBUG)) 2303b31e1552SEric Sandeen ext4_msg(sb, KERN_DEBUG, 2304b31e1552SEric Sandeen "%s: deleting unreferenced inode %lu", 230546e665e9SHarvey Harrison __func__, inode->i_ino); 2306ac27a0ecSDave Kleikamp jbd_debug(2, "deleting unreferenced inode %lu\n", 2307ac27a0ecSDave Kleikamp inode->i_ino); 2308ac27a0ecSDave Kleikamp nr_orphans++; 2309ac27a0ecSDave Kleikamp } 2310ac27a0ecSDave Kleikamp iput(inode); /* The delete magic happens here! */ 2311ac27a0ecSDave Kleikamp } 2312ac27a0ecSDave Kleikamp 2313ac27a0ecSDave Kleikamp #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 2314ac27a0ecSDave Kleikamp 2315ac27a0ecSDave Kleikamp if (nr_orphans) 2316b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 2317b31e1552SEric Sandeen PLURAL(nr_orphans)); 2318ac27a0ecSDave Kleikamp if (nr_truncates) 2319b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 2320b31e1552SEric Sandeen PLURAL(nr_truncates)); 2321ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 2322ac27a0ecSDave Kleikamp /* Turn quotas off */ 2323a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2324ac27a0ecSDave Kleikamp if (sb_dqopt(sb)->files[i]) 2325287a8095SChristoph Hellwig dquot_quota_off(sb, i); 2326ac27a0ecSDave Kleikamp } 2327ac27a0ecSDave Kleikamp #endif 2328ac27a0ecSDave Kleikamp sb->s_flags = s_flags; /* Restore MS_RDONLY status */ 2329ac27a0ecSDave Kleikamp } 23300b8e58a1SAndreas Dilger 2331cd2291a4SEric Sandeen /* 2332cd2291a4SEric Sandeen * Maximal extent format file size. 2333cd2291a4SEric Sandeen * Resulting logical blkno at s_maxbytes must fit in our on-disk 2334cd2291a4SEric Sandeen * extent format containers, within a sector_t, and within i_blocks 2335cd2291a4SEric Sandeen * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2336cd2291a4SEric Sandeen * so that won't be a limiting factor. 2337cd2291a4SEric Sandeen * 2338f17722f9SLukas Czerner * However there is other limiting factor. We do store extents in the form 2339f17722f9SLukas Czerner * of starting block and length, hence the resulting length of the extent 2340f17722f9SLukas Czerner * covering maximum file size must fit into on-disk format containers as 2341f17722f9SLukas Czerner * well. Given that length is always by 1 unit bigger than max unit (because 2342f17722f9SLukas Czerner * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2343f17722f9SLukas Czerner * 2344cd2291a4SEric Sandeen * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2345cd2291a4SEric Sandeen */ 2346f287a1a5STheodore Ts'o static loff_t ext4_max_size(int blkbits, int has_huge_files) 2347cd2291a4SEric Sandeen { 2348cd2291a4SEric Sandeen loff_t res; 2349cd2291a4SEric Sandeen loff_t upper_limit = MAX_LFS_FILESIZE; 2350cd2291a4SEric Sandeen 2351cd2291a4SEric Sandeen /* small i_blocks in vfs inode? */ 2352f287a1a5STheodore Ts'o if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 2353cd2291a4SEric Sandeen /* 235490c699a9SBartlomiej Zolnierkiewicz * CONFIG_LBDAF is not enabled implies the inode 2355cd2291a4SEric Sandeen * i_block represent total blocks in 512 bytes 2356cd2291a4SEric Sandeen * 32 == size of vfs inode i_blocks * 8 2357cd2291a4SEric Sandeen */ 2358cd2291a4SEric Sandeen upper_limit = (1LL << 32) - 1; 2359cd2291a4SEric Sandeen 2360cd2291a4SEric Sandeen /* total blocks in file system block size */ 2361cd2291a4SEric Sandeen upper_limit >>= (blkbits - 9); 2362cd2291a4SEric Sandeen upper_limit <<= blkbits; 2363cd2291a4SEric Sandeen } 2364cd2291a4SEric Sandeen 2365f17722f9SLukas Czerner /* 2366f17722f9SLukas Czerner * 32-bit extent-start container, ee_block. We lower the maxbytes 2367f17722f9SLukas Czerner * by one fs block, so ee_len can cover the extent of maximum file 2368f17722f9SLukas Czerner * size 2369f17722f9SLukas Czerner */ 2370f17722f9SLukas Czerner res = (1LL << 32) - 1; 2371cd2291a4SEric Sandeen res <<= blkbits; 2372cd2291a4SEric Sandeen 2373cd2291a4SEric Sandeen /* Sanity check against vm- & vfs- imposed limits */ 2374cd2291a4SEric Sandeen if (res > upper_limit) 2375cd2291a4SEric Sandeen res = upper_limit; 2376cd2291a4SEric Sandeen 2377cd2291a4SEric Sandeen return res; 2378cd2291a4SEric Sandeen } 2379ac27a0ecSDave Kleikamp 2380ac27a0ecSDave Kleikamp /* 2381cd2291a4SEric Sandeen * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 23820fc1b451SAneesh Kumar K.V * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 23830fc1b451SAneesh Kumar K.V * We need to be 1 filesystem block less than the 2^48 sector limit. 2384ac27a0ecSDave Kleikamp */ 2385f287a1a5STheodore Ts'o static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 2386ac27a0ecSDave Kleikamp { 2387617ba13bSMingming Cao loff_t res = EXT4_NDIR_BLOCKS; 23880fc1b451SAneesh Kumar K.V int meta_blocks; 23890fc1b451SAneesh Kumar K.V loff_t upper_limit; 23900b8e58a1SAndreas Dilger /* This is calculated to be the largest file size for a dense, block 23910b8e58a1SAndreas Dilger * mapped file such that the file's total number of 512-byte sectors, 23920b8e58a1SAndreas Dilger * including data and all indirect blocks, does not exceed (2^48 - 1). 23930b8e58a1SAndreas Dilger * 23940b8e58a1SAndreas Dilger * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 23950b8e58a1SAndreas Dilger * number of 512-byte sectors of the file. 23960fc1b451SAneesh Kumar K.V */ 23970fc1b451SAneesh Kumar K.V 2398f287a1a5STheodore Ts'o if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 23990fc1b451SAneesh Kumar K.V /* 240090c699a9SBartlomiej Zolnierkiewicz * !has_huge_files or CONFIG_LBDAF not enabled implies that 24010b8e58a1SAndreas Dilger * the inode i_block field represents total file blocks in 24020b8e58a1SAndreas Dilger * 2^32 512-byte sectors == size of vfs inode i_blocks * 8 24030fc1b451SAneesh Kumar K.V */ 24040fc1b451SAneesh Kumar K.V upper_limit = (1LL << 32) - 1; 24050fc1b451SAneesh Kumar K.V 24060fc1b451SAneesh Kumar K.V /* total blocks in file system block size */ 24070fc1b451SAneesh Kumar K.V upper_limit >>= (bits - 9); 24080fc1b451SAneesh Kumar K.V 24090fc1b451SAneesh Kumar K.V } else { 24108180a562SAneesh Kumar K.V /* 24118180a562SAneesh Kumar K.V * We use 48 bit ext4_inode i_blocks 24128180a562SAneesh Kumar K.V * With EXT4_HUGE_FILE_FL set the i_blocks 24138180a562SAneesh Kumar K.V * represent total number of blocks in 24148180a562SAneesh Kumar K.V * file system block size 24158180a562SAneesh Kumar K.V */ 24160fc1b451SAneesh Kumar K.V upper_limit = (1LL << 48) - 1; 24170fc1b451SAneesh Kumar K.V 24180fc1b451SAneesh Kumar K.V } 24190fc1b451SAneesh Kumar K.V 24200fc1b451SAneesh Kumar K.V /* indirect blocks */ 24210fc1b451SAneesh Kumar K.V meta_blocks = 1; 24220fc1b451SAneesh Kumar K.V /* double indirect blocks */ 24230fc1b451SAneesh Kumar K.V meta_blocks += 1 + (1LL << (bits-2)); 24240fc1b451SAneesh Kumar K.V /* tripple indirect blocks */ 24250fc1b451SAneesh Kumar K.V meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 24260fc1b451SAneesh Kumar K.V 24270fc1b451SAneesh Kumar K.V upper_limit -= meta_blocks; 24280fc1b451SAneesh Kumar K.V upper_limit <<= bits; 2429ac27a0ecSDave Kleikamp 2430ac27a0ecSDave Kleikamp res += 1LL << (bits-2); 2431ac27a0ecSDave Kleikamp res += 1LL << (2*(bits-2)); 2432ac27a0ecSDave Kleikamp res += 1LL << (3*(bits-2)); 2433ac27a0ecSDave Kleikamp res <<= bits; 2434ac27a0ecSDave Kleikamp if (res > upper_limit) 2435ac27a0ecSDave Kleikamp res = upper_limit; 24360fc1b451SAneesh Kumar K.V 24370fc1b451SAneesh Kumar K.V if (res > MAX_LFS_FILESIZE) 24380fc1b451SAneesh Kumar K.V res = MAX_LFS_FILESIZE; 24390fc1b451SAneesh Kumar K.V 2440ac27a0ecSDave Kleikamp return res; 2441ac27a0ecSDave Kleikamp } 2442ac27a0ecSDave Kleikamp 2443617ba13bSMingming Cao static ext4_fsblk_t descriptor_loc(struct super_block *sb, 244470bbb3e0SAndrew Morton ext4_fsblk_t logical_sb_block, int nr) 2445ac27a0ecSDave Kleikamp { 2446617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 2447fd2d4291SAvantika Mathur ext4_group_t bg, first_meta_bg; 2448ac27a0ecSDave Kleikamp int has_super = 0; 2449ac27a0ecSDave Kleikamp 2450ac27a0ecSDave Kleikamp first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 2451ac27a0ecSDave Kleikamp 2452e2b911c5SDarrick J. Wong if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 245370bbb3e0SAndrew Morton return logical_sb_block + nr + 1; 2454ac27a0ecSDave Kleikamp bg = sbi->s_desc_per_block * nr; 2455617ba13bSMingming Cao if (ext4_bg_has_super(sb, bg)) 2456ac27a0ecSDave Kleikamp has_super = 1; 24570b8e58a1SAndreas Dilger 2458bd63f6b0SDarrick J. Wong /* 2459bd63f6b0SDarrick J. Wong * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 2460bd63f6b0SDarrick J. Wong * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 2461bd63f6b0SDarrick J. Wong * on modern mke2fs or blksize > 1k on older mke2fs) then we must 2462bd63f6b0SDarrick J. Wong * compensate. 2463bd63f6b0SDarrick J. Wong */ 2464bd63f6b0SDarrick J. Wong if (sb->s_blocksize == 1024 && nr == 0 && 2465bd63f6b0SDarrick J. Wong le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0) 2466bd63f6b0SDarrick J. Wong has_super++; 2467bd63f6b0SDarrick J. Wong 2468617ba13bSMingming Cao return (has_super + ext4_group_first_block_no(sb, bg)); 2469ac27a0ecSDave Kleikamp } 2470ac27a0ecSDave Kleikamp 2471c9de560dSAlex Tomas /** 2472c9de560dSAlex Tomas * ext4_get_stripe_size: Get the stripe size. 2473c9de560dSAlex Tomas * @sbi: In memory super block info 2474c9de560dSAlex Tomas * 2475c9de560dSAlex Tomas * If we have specified it via mount option, then 2476c9de560dSAlex Tomas * use the mount option value. If the value specified at mount time is 2477c9de560dSAlex Tomas * greater than the blocks per group use the super block value. 2478c9de560dSAlex Tomas * If the super block value is greater than blocks per group return 0. 2479c9de560dSAlex Tomas * Allocator needs it be less than blocks per group. 2480c9de560dSAlex Tomas * 2481c9de560dSAlex Tomas */ 2482c9de560dSAlex Tomas static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 2483c9de560dSAlex Tomas { 2484c9de560dSAlex Tomas unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 2485c9de560dSAlex Tomas unsigned long stripe_width = 2486c9de560dSAlex Tomas le32_to_cpu(sbi->s_es->s_raid_stripe_width); 24873eb08658SDan Ehrenberg int ret; 2488c9de560dSAlex Tomas 2489c9de560dSAlex Tomas if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 24903eb08658SDan Ehrenberg ret = sbi->s_stripe; 24913eb08658SDan Ehrenberg else if (stripe_width <= sbi->s_blocks_per_group) 24923eb08658SDan Ehrenberg ret = stripe_width; 24933eb08658SDan Ehrenberg else if (stride <= sbi->s_blocks_per_group) 24943eb08658SDan Ehrenberg ret = stride; 24953eb08658SDan Ehrenberg else 24963eb08658SDan Ehrenberg ret = 0; 2497c9de560dSAlex Tomas 24983eb08658SDan Ehrenberg /* 24993eb08658SDan Ehrenberg * If the stripe width is 1, this makes no sense and 25003eb08658SDan Ehrenberg * we set it to 0 to turn off stripe handling code. 25013eb08658SDan Ehrenberg */ 25023eb08658SDan Ehrenberg if (ret <= 1) 25033eb08658SDan Ehrenberg ret = 0; 2504c9de560dSAlex Tomas 25053eb08658SDan Ehrenberg return ret; 2506c9de560dSAlex Tomas } 2507ac27a0ecSDave Kleikamp 2508a13fb1a4SEric Sandeen /* 2509a13fb1a4SEric Sandeen * Check whether this filesystem can be mounted based on 2510a13fb1a4SEric Sandeen * the features present and the RDONLY/RDWR mount requested. 2511a13fb1a4SEric Sandeen * Returns 1 if this filesystem can be mounted as requested, 2512a13fb1a4SEric Sandeen * 0 if it cannot be. 2513a13fb1a4SEric Sandeen */ 2514a13fb1a4SEric Sandeen static int ext4_feature_set_ok(struct super_block *sb, int readonly) 2515a13fb1a4SEric Sandeen { 2516e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext4_incompat_features(sb)) { 2517a13fb1a4SEric Sandeen ext4_msg(sb, KERN_ERR, 2518a13fb1a4SEric Sandeen "Couldn't mount because of " 2519a13fb1a4SEric Sandeen "unsupported optional features (%x)", 2520a13fb1a4SEric Sandeen (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 2521a13fb1a4SEric Sandeen ~EXT4_FEATURE_INCOMPAT_SUPP)); 2522a13fb1a4SEric Sandeen return 0; 2523a13fb1a4SEric Sandeen } 2524a13fb1a4SEric Sandeen 2525a13fb1a4SEric Sandeen if (readonly) 2526a13fb1a4SEric Sandeen return 1; 2527a13fb1a4SEric Sandeen 2528e2b911c5SDarrick J. Wong if (ext4_has_feature_readonly(sb)) { 25292cb5cc8bSDarrick J. Wong ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 25302cb5cc8bSDarrick J. Wong sb->s_flags |= MS_RDONLY; 25312cb5cc8bSDarrick J. Wong return 1; 25322cb5cc8bSDarrick J. Wong } 25332cb5cc8bSDarrick J. Wong 2534a13fb1a4SEric Sandeen /* Check that feature set is OK for a read-write mount */ 2535e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 2536a13fb1a4SEric Sandeen ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 2537a13fb1a4SEric Sandeen "unsupported optional features (%x)", 2538a13fb1a4SEric Sandeen (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 2539a13fb1a4SEric Sandeen ~EXT4_FEATURE_RO_COMPAT_SUPP)); 2540a13fb1a4SEric Sandeen return 0; 2541a13fb1a4SEric Sandeen } 2542a13fb1a4SEric Sandeen /* 2543a13fb1a4SEric Sandeen * Large file size enabled file system can only be mounted 2544a13fb1a4SEric Sandeen * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF 2545a13fb1a4SEric Sandeen */ 2546e2b911c5SDarrick J. Wong if (ext4_has_feature_huge_file(sb)) { 2547a13fb1a4SEric Sandeen if (sizeof(blkcnt_t) < sizeof(u64)) { 2548a13fb1a4SEric Sandeen ext4_msg(sb, KERN_ERR, "Filesystem with huge files " 2549a13fb1a4SEric Sandeen "cannot be mounted RDWR without " 2550a13fb1a4SEric Sandeen "CONFIG_LBDAF"); 2551a13fb1a4SEric Sandeen return 0; 2552a13fb1a4SEric Sandeen } 2553a13fb1a4SEric Sandeen } 2554e2b911c5SDarrick J. Wong if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 2555bab08ab9STheodore Ts'o ext4_msg(sb, KERN_ERR, 2556bab08ab9STheodore Ts'o "Can't support bigalloc feature without " 2557bab08ab9STheodore Ts'o "extents feature\n"); 2558bab08ab9STheodore Ts'o return 0; 2559bab08ab9STheodore Ts'o } 25607c319d32SAditya Kali 25617c319d32SAditya Kali #ifndef CONFIG_QUOTA 2562e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb) && !readonly) { 25637c319d32SAditya Kali ext4_msg(sb, KERN_ERR, 25647c319d32SAditya Kali "Filesystem with quota feature cannot be mounted RDWR " 25657c319d32SAditya Kali "without CONFIG_QUOTA"); 25667c319d32SAditya Kali return 0; 25677c319d32SAditya Kali } 2568689c958cSLi Xi if (ext4_has_feature_project(sb) && !readonly) { 2569689c958cSLi Xi ext4_msg(sb, KERN_ERR, 2570689c958cSLi Xi "Filesystem with project quota feature cannot be mounted RDWR " 2571689c958cSLi Xi "without CONFIG_QUOTA"); 2572689c958cSLi Xi return 0; 2573689c958cSLi Xi } 25747c319d32SAditya Kali #endif /* CONFIG_QUOTA */ 2575a13fb1a4SEric Sandeen return 1; 2576a13fb1a4SEric Sandeen } 2577a13fb1a4SEric Sandeen 257866e61a9eSTheodore Ts'o /* 257966e61a9eSTheodore Ts'o * This function is called once a day if we have errors logged 258066e61a9eSTheodore Ts'o * on the file system 258166e61a9eSTheodore Ts'o */ 258266e61a9eSTheodore Ts'o static void print_daily_error_info(unsigned long arg) 258366e61a9eSTheodore Ts'o { 258466e61a9eSTheodore Ts'o struct super_block *sb = (struct super_block *) arg; 258566e61a9eSTheodore Ts'o struct ext4_sb_info *sbi; 258666e61a9eSTheodore Ts'o struct ext4_super_block *es; 258766e61a9eSTheodore Ts'o 258866e61a9eSTheodore Ts'o sbi = EXT4_SB(sb); 258966e61a9eSTheodore Ts'o es = sbi->s_es; 259066e61a9eSTheodore Ts'o 259166e61a9eSTheodore Ts'o if (es->s_error_count) 2592ae0f78deSTheodore Ts'o /* fsck newer than v1.41.13 is needed to clean this condition. */ 2593ae0f78deSTheodore Ts'o ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 259466e61a9eSTheodore Ts'o le32_to_cpu(es->s_error_count)); 259566e61a9eSTheodore Ts'o if (es->s_first_error_time) { 2596ae0f78deSTheodore Ts'o printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d", 259766e61a9eSTheodore Ts'o sb->s_id, le32_to_cpu(es->s_first_error_time), 259866e61a9eSTheodore Ts'o (int) sizeof(es->s_first_error_func), 259966e61a9eSTheodore Ts'o es->s_first_error_func, 260066e61a9eSTheodore Ts'o le32_to_cpu(es->s_first_error_line)); 260166e61a9eSTheodore Ts'o if (es->s_first_error_ino) 260266e61a9eSTheodore Ts'o printk(": inode %u", 260366e61a9eSTheodore Ts'o le32_to_cpu(es->s_first_error_ino)); 260466e61a9eSTheodore Ts'o if (es->s_first_error_block) 260566e61a9eSTheodore Ts'o printk(": block %llu", (unsigned long long) 260666e61a9eSTheodore Ts'o le64_to_cpu(es->s_first_error_block)); 260766e61a9eSTheodore Ts'o printk("\n"); 260866e61a9eSTheodore Ts'o } 260966e61a9eSTheodore Ts'o if (es->s_last_error_time) { 2610ae0f78deSTheodore Ts'o printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d", 261166e61a9eSTheodore Ts'o sb->s_id, le32_to_cpu(es->s_last_error_time), 261266e61a9eSTheodore Ts'o (int) sizeof(es->s_last_error_func), 261366e61a9eSTheodore Ts'o es->s_last_error_func, 261466e61a9eSTheodore Ts'o le32_to_cpu(es->s_last_error_line)); 261566e61a9eSTheodore Ts'o if (es->s_last_error_ino) 261666e61a9eSTheodore Ts'o printk(": inode %u", 261766e61a9eSTheodore Ts'o le32_to_cpu(es->s_last_error_ino)); 261866e61a9eSTheodore Ts'o if (es->s_last_error_block) 261966e61a9eSTheodore Ts'o printk(": block %llu", (unsigned long long) 262066e61a9eSTheodore Ts'o le64_to_cpu(es->s_last_error_block)); 262166e61a9eSTheodore Ts'o printk("\n"); 262266e61a9eSTheodore Ts'o } 262366e61a9eSTheodore Ts'o mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 262466e61a9eSTheodore Ts'o } 262566e61a9eSTheodore Ts'o 2626bfff6873SLukas Czerner /* Find next suitable group and run ext4_init_inode_table */ 2627bfff6873SLukas Czerner static int ext4_run_li_request(struct ext4_li_request *elr) 2628bfff6873SLukas Czerner { 2629bfff6873SLukas Czerner struct ext4_group_desc *gdp = NULL; 2630bfff6873SLukas Czerner ext4_group_t group, ngroups; 2631bfff6873SLukas Czerner struct super_block *sb; 2632bfff6873SLukas Czerner unsigned long timeout = 0; 2633bfff6873SLukas Czerner int ret = 0; 2634bfff6873SLukas Czerner 2635bfff6873SLukas Czerner sb = elr->lr_super; 2636bfff6873SLukas Czerner ngroups = EXT4_SB(sb)->s_groups_count; 2637bfff6873SLukas Czerner 26388e8ad8a5SJan Kara sb_start_write(sb); 2639bfff6873SLukas Czerner for (group = elr->lr_next_group; group < ngroups; group++) { 2640bfff6873SLukas Czerner gdp = ext4_get_group_desc(sb, group, NULL); 2641bfff6873SLukas Czerner if (!gdp) { 2642bfff6873SLukas Czerner ret = 1; 2643bfff6873SLukas Czerner break; 2644bfff6873SLukas Czerner } 2645bfff6873SLukas Czerner 2646bfff6873SLukas Czerner if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2647bfff6873SLukas Czerner break; 2648bfff6873SLukas Czerner } 2649bfff6873SLukas Czerner 26507f511862STheodore Ts'o if (group >= ngroups) 2651bfff6873SLukas Czerner ret = 1; 2652bfff6873SLukas Czerner 2653bfff6873SLukas Czerner if (!ret) { 2654bfff6873SLukas Czerner timeout = jiffies; 2655bfff6873SLukas Czerner ret = ext4_init_inode_table(sb, group, 2656bfff6873SLukas Czerner elr->lr_timeout ? 0 : 1); 2657bfff6873SLukas Czerner if (elr->lr_timeout == 0) { 265851ce6511SLukas Czerner timeout = (jiffies - timeout) * 265951ce6511SLukas Czerner elr->lr_sbi->s_li_wait_mult; 2660bfff6873SLukas Czerner elr->lr_timeout = timeout; 2661bfff6873SLukas Czerner } 2662bfff6873SLukas Czerner elr->lr_next_sched = jiffies + elr->lr_timeout; 2663bfff6873SLukas Czerner elr->lr_next_group = group + 1; 2664bfff6873SLukas Czerner } 26658e8ad8a5SJan Kara sb_end_write(sb); 2666bfff6873SLukas Czerner 2667bfff6873SLukas Czerner return ret; 2668bfff6873SLukas Czerner } 2669bfff6873SLukas Czerner 2670bfff6873SLukas Czerner /* 2671bfff6873SLukas Czerner * Remove lr_request from the list_request and free the 26724ed5c033SLukas Czerner * request structure. Should be called with li_list_mtx held 2673bfff6873SLukas Czerner */ 2674bfff6873SLukas Czerner static void ext4_remove_li_request(struct ext4_li_request *elr) 2675bfff6873SLukas Czerner { 2676bfff6873SLukas Czerner struct ext4_sb_info *sbi; 2677bfff6873SLukas Czerner 2678bfff6873SLukas Czerner if (!elr) 2679bfff6873SLukas Czerner return; 2680bfff6873SLukas Czerner 2681bfff6873SLukas Czerner sbi = elr->lr_sbi; 2682bfff6873SLukas Czerner 2683bfff6873SLukas Czerner list_del(&elr->lr_request); 2684bfff6873SLukas Czerner sbi->s_li_request = NULL; 2685bfff6873SLukas Czerner kfree(elr); 2686bfff6873SLukas Czerner } 2687bfff6873SLukas Czerner 2688bfff6873SLukas Czerner static void ext4_unregister_li_request(struct super_block *sb) 2689bfff6873SLukas Czerner { 26901bb933fbSLukas Czerner mutex_lock(&ext4_li_mtx); 26911bb933fbSLukas Czerner if (!ext4_li_info) { 26921bb933fbSLukas Czerner mutex_unlock(&ext4_li_mtx); 2693bfff6873SLukas Czerner return; 26941bb933fbSLukas Czerner } 2695bfff6873SLukas Czerner 2696bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 26971bb933fbSLukas Czerner ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 2698bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 26991bb933fbSLukas Czerner mutex_unlock(&ext4_li_mtx); 2700bfff6873SLukas Czerner } 2701bfff6873SLukas Czerner 27028f1f7453SEric Sandeen static struct task_struct *ext4_lazyinit_task; 27038f1f7453SEric Sandeen 2704bfff6873SLukas Czerner /* 2705bfff6873SLukas Czerner * This is the function where ext4lazyinit thread lives. It walks 2706bfff6873SLukas Czerner * through the request list searching for next scheduled filesystem. 2707bfff6873SLukas Czerner * When such a fs is found, run the lazy initialization request 2708bfff6873SLukas Czerner * (ext4_rn_li_request) and keep track of the time spend in this 2709bfff6873SLukas Czerner * function. Based on that time we compute next schedule time of 2710bfff6873SLukas Czerner * the request. When walking through the list is complete, compute 2711bfff6873SLukas Czerner * next waking time and put itself into sleep. 2712bfff6873SLukas Czerner */ 2713bfff6873SLukas Czerner static int ext4_lazyinit_thread(void *arg) 2714bfff6873SLukas Czerner { 2715bfff6873SLukas Czerner struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 2716bfff6873SLukas Czerner struct list_head *pos, *n; 2717bfff6873SLukas Czerner struct ext4_li_request *elr; 27184ed5c033SLukas Czerner unsigned long next_wakeup, cur; 2719bfff6873SLukas Czerner 2720bfff6873SLukas Czerner BUG_ON(NULL == eli); 2721bfff6873SLukas Czerner 2722bfff6873SLukas Czerner cont_thread: 2723bfff6873SLukas Czerner while (true) { 2724bfff6873SLukas Czerner next_wakeup = MAX_JIFFY_OFFSET; 2725bfff6873SLukas Czerner 2726bfff6873SLukas Czerner mutex_lock(&eli->li_list_mtx); 2727bfff6873SLukas Czerner if (list_empty(&eli->li_request_list)) { 2728bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 2729bfff6873SLukas Czerner goto exit_thread; 2730bfff6873SLukas Czerner } 2731bfff6873SLukas Czerner 2732bfff6873SLukas Czerner list_for_each_safe(pos, n, &eli->li_request_list) { 2733bfff6873SLukas Czerner elr = list_entry(pos, struct ext4_li_request, 2734bfff6873SLukas Czerner lr_request); 2735bfff6873SLukas Czerner 2736b2c78cd0STheodore Ts'o if (time_after_eq(jiffies, elr->lr_next_sched)) { 2737b2c78cd0STheodore Ts'o if (ext4_run_li_request(elr) != 0) { 2738b2c78cd0STheodore Ts'o /* error, remove the lazy_init job */ 2739bfff6873SLukas Czerner ext4_remove_li_request(elr); 2740bfff6873SLukas Czerner continue; 2741bfff6873SLukas Czerner } 2742b2c78cd0STheodore Ts'o } 2743bfff6873SLukas Czerner 2744bfff6873SLukas Czerner if (time_before(elr->lr_next_sched, next_wakeup)) 2745bfff6873SLukas Czerner next_wakeup = elr->lr_next_sched; 2746bfff6873SLukas Czerner } 2747bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 2748bfff6873SLukas Czerner 2749a0acae0eSTejun Heo try_to_freeze(); 2750bfff6873SLukas Czerner 27514ed5c033SLukas Czerner cur = jiffies; 27524ed5c033SLukas Czerner if ((time_after_eq(cur, next_wakeup)) || 2753f4245bd4SLukas Czerner (MAX_JIFFY_OFFSET == next_wakeup)) { 2754bfff6873SLukas Czerner cond_resched(); 2755bfff6873SLukas Czerner continue; 2756bfff6873SLukas Czerner } 2757bfff6873SLukas Czerner 27584ed5c033SLukas Czerner schedule_timeout_interruptible(next_wakeup - cur); 27594ed5c033SLukas Czerner 27608f1f7453SEric Sandeen if (kthread_should_stop()) { 27618f1f7453SEric Sandeen ext4_clear_request_list(); 27628f1f7453SEric Sandeen goto exit_thread; 27638f1f7453SEric Sandeen } 2764bfff6873SLukas Czerner } 2765bfff6873SLukas Czerner 2766bfff6873SLukas Czerner exit_thread: 2767bfff6873SLukas Czerner /* 2768bfff6873SLukas Czerner * It looks like the request list is empty, but we need 2769bfff6873SLukas Czerner * to check it under the li_list_mtx lock, to prevent any 2770bfff6873SLukas Czerner * additions into it, and of course we should lock ext4_li_mtx 2771bfff6873SLukas Czerner * to atomically free the list and ext4_li_info, because at 2772bfff6873SLukas Czerner * this point another ext4 filesystem could be registering 2773bfff6873SLukas Czerner * new one. 2774bfff6873SLukas Czerner */ 2775bfff6873SLukas Czerner mutex_lock(&ext4_li_mtx); 2776bfff6873SLukas Czerner mutex_lock(&eli->li_list_mtx); 2777bfff6873SLukas Czerner if (!list_empty(&eli->li_request_list)) { 2778bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 2779bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 2780bfff6873SLukas Czerner goto cont_thread; 2781bfff6873SLukas Czerner } 2782bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 2783bfff6873SLukas Czerner kfree(ext4_li_info); 2784bfff6873SLukas Czerner ext4_li_info = NULL; 2785bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 2786bfff6873SLukas Czerner 2787bfff6873SLukas Czerner return 0; 2788bfff6873SLukas Czerner } 2789bfff6873SLukas Czerner 2790bfff6873SLukas Czerner static void ext4_clear_request_list(void) 2791bfff6873SLukas Czerner { 2792bfff6873SLukas Czerner struct list_head *pos, *n; 2793bfff6873SLukas Czerner struct ext4_li_request *elr; 2794bfff6873SLukas Czerner 2795bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 2796bfff6873SLukas Czerner list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 2797bfff6873SLukas Czerner elr = list_entry(pos, struct ext4_li_request, 2798bfff6873SLukas Czerner lr_request); 2799bfff6873SLukas Czerner ext4_remove_li_request(elr); 2800bfff6873SLukas Czerner } 2801bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 2802bfff6873SLukas Czerner } 2803bfff6873SLukas Czerner 2804bfff6873SLukas Czerner static int ext4_run_lazyinit_thread(void) 2805bfff6873SLukas Czerner { 28068f1f7453SEric Sandeen ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 28078f1f7453SEric Sandeen ext4_li_info, "ext4lazyinit"); 28088f1f7453SEric Sandeen if (IS_ERR(ext4_lazyinit_task)) { 28098f1f7453SEric Sandeen int err = PTR_ERR(ext4_lazyinit_task); 2810bfff6873SLukas Czerner ext4_clear_request_list(); 2811bfff6873SLukas Czerner kfree(ext4_li_info); 2812bfff6873SLukas Czerner ext4_li_info = NULL; 281392b97816STheodore Ts'o printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 2814bfff6873SLukas Czerner "initialization thread\n", 2815bfff6873SLukas Czerner err); 2816bfff6873SLukas Czerner return err; 2817bfff6873SLukas Czerner } 2818bfff6873SLukas Czerner ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 2819bfff6873SLukas Czerner return 0; 2820bfff6873SLukas Czerner } 2821bfff6873SLukas Czerner 2822bfff6873SLukas Czerner /* 2823bfff6873SLukas Czerner * Check whether it make sense to run itable init. thread or not. 2824bfff6873SLukas Czerner * If there is at least one uninitialized inode table, return 2825bfff6873SLukas Czerner * corresponding group number, else the loop goes through all 2826bfff6873SLukas Czerner * groups and return total number of groups. 2827bfff6873SLukas Czerner */ 2828bfff6873SLukas Czerner static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 2829bfff6873SLukas Czerner { 2830bfff6873SLukas Czerner ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 2831bfff6873SLukas Czerner struct ext4_group_desc *gdp = NULL; 2832bfff6873SLukas Czerner 2833bfff6873SLukas Czerner for (group = 0; group < ngroups; group++) { 2834bfff6873SLukas Czerner gdp = ext4_get_group_desc(sb, group, NULL); 2835bfff6873SLukas Czerner if (!gdp) 2836bfff6873SLukas Czerner continue; 2837bfff6873SLukas Czerner 2838bfff6873SLukas Czerner if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2839bfff6873SLukas Czerner break; 2840bfff6873SLukas Czerner } 2841bfff6873SLukas Czerner 2842bfff6873SLukas Czerner return group; 2843bfff6873SLukas Czerner } 2844bfff6873SLukas Czerner 2845bfff6873SLukas Czerner static int ext4_li_info_new(void) 2846bfff6873SLukas Czerner { 2847bfff6873SLukas Czerner struct ext4_lazy_init *eli = NULL; 2848bfff6873SLukas Czerner 2849bfff6873SLukas Czerner eli = kzalloc(sizeof(*eli), GFP_KERNEL); 2850bfff6873SLukas Czerner if (!eli) 2851bfff6873SLukas Czerner return -ENOMEM; 2852bfff6873SLukas Czerner 2853bfff6873SLukas Czerner INIT_LIST_HEAD(&eli->li_request_list); 2854bfff6873SLukas Czerner mutex_init(&eli->li_list_mtx); 2855bfff6873SLukas Czerner 2856bfff6873SLukas Czerner eli->li_state |= EXT4_LAZYINIT_QUIT; 2857bfff6873SLukas Czerner 2858bfff6873SLukas Czerner ext4_li_info = eli; 2859bfff6873SLukas Czerner 2860bfff6873SLukas Czerner return 0; 2861bfff6873SLukas Czerner } 2862bfff6873SLukas Czerner 2863bfff6873SLukas Czerner static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 2864bfff6873SLukas Czerner ext4_group_t start) 2865bfff6873SLukas Czerner { 2866bfff6873SLukas Czerner struct ext4_sb_info *sbi = EXT4_SB(sb); 2867bfff6873SLukas Czerner struct ext4_li_request *elr; 2868bfff6873SLukas Czerner 2869bfff6873SLukas Czerner elr = kzalloc(sizeof(*elr), GFP_KERNEL); 2870bfff6873SLukas Czerner if (!elr) 2871bfff6873SLukas Czerner return NULL; 2872bfff6873SLukas Czerner 2873bfff6873SLukas Czerner elr->lr_super = sb; 2874bfff6873SLukas Czerner elr->lr_sbi = sbi; 2875bfff6873SLukas Czerner elr->lr_next_group = start; 2876bfff6873SLukas Czerner 2877bfff6873SLukas Czerner /* 2878bfff6873SLukas Czerner * Randomize first schedule time of the request to 2879bfff6873SLukas Czerner * spread the inode table initialization requests 2880bfff6873SLukas Czerner * better. 2881bfff6873SLukas Czerner */ 2882dd1f723bSTheodore Ts'o elr->lr_next_sched = jiffies + (prandom_u32() % 2883dd1f723bSTheodore Ts'o (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 2884bfff6873SLukas Czerner return elr; 2885bfff6873SLukas Czerner } 2886bfff6873SLukas Czerner 28877f511862STheodore Ts'o int ext4_register_li_request(struct super_block *sb, 2888bfff6873SLukas Czerner ext4_group_t first_not_zeroed) 2889bfff6873SLukas Czerner { 2890bfff6873SLukas Czerner struct ext4_sb_info *sbi = EXT4_SB(sb); 28917f511862STheodore Ts'o struct ext4_li_request *elr = NULL; 2892bfff6873SLukas Czerner ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 28936c5a6cb9SAndrew Morton int ret = 0; 2894bfff6873SLukas Czerner 28957f511862STheodore Ts'o mutex_lock(&ext4_li_mtx); 289651ce6511SLukas Czerner if (sbi->s_li_request != NULL) { 289751ce6511SLukas Czerner /* 289851ce6511SLukas Czerner * Reset timeout so it can be computed again, because 289951ce6511SLukas Czerner * s_li_wait_mult might have changed. 290051ce6511SLukas Czerner */ 290151ce6511SLukas Czerner sbi->s_li_request->lr_timeout = 0; 29027f511862STheodore Ts'o goto out; 290351ce6511SLukas Czerner } 2904bfff6873SLukas Czerner 2905bfff6873SLukas Czerner if (first_not_zeroed == ngroups || 2906bfff6873SLukas Czerner (sb->s_flags & MS_RDONLY) || 290755ff3840STao Ma !test_opt(sb, INIT_INODE_TABLE)) 29087f511862STheodore Ts'o goto out; 2909bfff6873SLukas Czerner 2910bfff6873SLukas Czerner elr = ext4_li_request_new(sb, first_not_zeroed); 29117f511862STheodore Ts'o if (!elr) { 29127f511862STheodore Ts'o ret = -ENOMEM; 29137f511862STheodore Ts'o goto out; 29147f511862STheodore Ts'o } 2915bfff6873SLukas Czerner 2916bfff6873SLukas Czerner if (NULL == ext4_li_info) { 2917bfff6873SLukas Czerner ret = ext4_li_info_new(); 2918bfff6873SLukas Czerner if (ret) 2919bfff6873SLukas Czerner goto out; 2920bfff6873SLukas Czerner } 2921bfff6873SLukas Czerner 2922bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 2923bfff6873SLukas Czerner list_add(&elr->lr_request, &ext4_li_info->li_request_list); 2924bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 2925bfff6873SLukas Czerner 2926bfff6873SLukas Czerner sbi->s_li_request = elr; 292746e4690bSTao Ma /* 292846e4690bSTao Ma * set elr to NULL here since it has been inserted to 292946e4690bSTao Ma * the request_list and the removal and free of it is 293046e4690bSTao Ma * handled by ext4_clear_request_list from now on. 293146e4690bSTao Ma */ 293246e4690bSTao Ma elr = NULL; 2933bfff6873SLukas Czerner 2934bfff6873SLukas Czerner if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 2935bfff6873SLukas Czerner ret = ext4_run_lazyinit_thread(); 2936bfff6873SLukas Czerner if (ret) 2937bfff6873SLukas Czerner goto out; 2938bfff6873SLukas Czerner } 2939bfff6873SLukas Czerner out: 2940bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 2941beed5ecbSNicolas Kaiser if (ret) 2942bfff6873SLukas Czerner kfree(elr); 2943bfff6873SLukas Czerner return ret; 2944bfff6873SLukas Czerner } 2945bfff6873SLukas Czerner 2946bfff6873SLukas Czerner /* 2947bfff6873SLukas Czerner * We do not need to lock anything since this is called on 2948bfff6873SLukas Czerner * module unload. 2949bfff6873SLukas Czerner */ 2950bfff6873SLukas Czerner static void ext4_destroy_lazyinit_thread(void) 2951bfff6873SLukas Czerner { 2952bfff6873SLukas Czerner /* 2953bfff6873SLukas Czerner * If thread exited earlier 2954bfff6873SLukas Czerner * there's nothing to be done. 2955bfff6873SLukas Czerner */ 29568f1f7453SEric Sandeen if (!ext4_li_info || !ext4_lazyinit_task) 2957bfff6873SLukas Czerner return; 2958bfff6873SLukas Czerner 29598f1f7453SEric Sandeen kthread_stop(ext4_lazyinit_task); 2960bfff6873SLukas Czerner } 2961bfff6873SLukas Czerner 296225ed6e8aSDarrick J. Wong static int set_journal_csum_feature_set(struct super_block *sb) 296325ed6e8aSDarrick J. Wong { 296425ed6e8aSDarrick J. Wong int ret = 1; 296525ed6e8aSDarrick J. Wong int compat, incompat; 296625ed6e8aSDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 296725ed6e8aSDarrick J. Wong 29689aa5d32bSDmitry Monakhov if (ext4_has_metadata_csum(sb)) { 2969db9ee220SDarrick J. Wong /* journal checksum v3 */ 297025ed6e8aSDarrick J. Wong compat = 0; 2971db9ee220SDarrick J. Wong incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 297225ed6e8aSDarrick J. Wong } else { 297325ed6e8aSDarrick J. Wong /* journal checksum v1 */ 297425ed6e8aSDarrick J. Wong compat = JBD2_FEATURE_COMPAT_CHECKSUM; 297525ed6e8aSDarrick J. Wong incompat = 0; 297625ed6e8aSDarrick J. Wong } 297725ed6e8aSDarrick J. Wong 2978feb8c6d3SDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 2979feb8c6d3SDarrick J. Wong JBD2_FEATURE_COMPAT_CHECKSUM, 0, 2980feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_CSUM_V3 | 2981feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_CSUM_V2); 298225ed6e8aSDarrick J. Wong if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 298325ed6e8aSDarrick J. Wong ret = jbd2_journal_set_features(sbi->s_journal, 298425ed6e8aSDarrick J. Wong compat, 0, 298525ed6e8aSDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 298625ed6e8aSDarrick J. Wong incompat); 298725ed6e8aSDarrick J. Wong } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 298825ed6e8aSDarrick J. Wong ret = jbd2_journal_set_features(sbi->s_journal, 298925ed6e8aSDarrick J. Wong compat, 0, 299025ed6e8aSDarrick J. Wong incompat); 299125ed6e8aSDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 0, 0, 299225ed6e8aSDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 299325ed6e8aSDarrick J. Wong } else { 2994feb8c6d3SDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 0, 0, 2995feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 299625ed6e8aSDarrick J. Wong } 299725ed6e8aSDarrick J. Wong 299825ed6e8aSDarrick J. Wong return ret; 299925ed6e8aSDarrick J. Wong } 300025ed6e8aSDarrick J. Wong 3001952fc18eSTheodore Ts'o /* 3002952fc18eSTheodore Ts'o * Note: calculating the overhead so we can be compatible with 3003952fc18eSTheodore Ts'o * historical BSD practice is quite difficult in the face of 3004952fc18eSTheodore Ts'o * clusters/bigalloc. This is because multiple metadata blocks from 3005952fc18eSTheodore Ts'o * different block group can end up in the same allocation cluster. 3006952fc18eSTheodore Ts'o * Calculating the exact overhead in the face of clustered allocation 3007952fc18eSTheodore Ts'o * requires either O(all block bitmaps) in memory or O(number of block 3008952fc18eSTheodore Ts'o * groups**2) in time. We will still calculate the superblock for 3009952fc18eSTheodore Ts'o * older file systems --- and if we come across with a bigalloc file 3010952fc18eSTheodore Ts'o * system with zero in s_overhead_clusters the estimate will be close to 3011952fc18eSTheodore Ts'o * correct especially for very large cluster sizes --- but for newer 3012952fc18eSTheodore Ts'o * file systems, it's better to calculate this figure once at mkfs 3013952fc18eSTheodore Ts'o * time, and store it in the superblock. If the superblock value is 3014952fc18eSTheodore Ts'o * present (even for non-bigalloc file systems), we will use it. 3015952fc18eSTheodore Ts'o */ 3016952fc18eSTheodore Ts'o static int count_overhead(struct super_block *sb, ext4_group_t grp, 3017952fc18eSTheodore Ts'o char *buf) 3018952fc18eSTheodore Ts'o { 3019952fc18eSTheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 3020952fc18eSTheodore Ts'o struct ext4_group_desc *gdp; 3021952fc18eSTheodore Ts'o ext4_fsblk_t first_block, last_block, b; 3022952fc18eSTheodore Ts'o ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3023952fc18eSTheodore Ts'o int s, j, count = 0; 3024952fc18eSTheodore Ts'o 3025e2b911c5SDarrick J. Wong if (!ext4_has_feature_bigalloc(sb)) 30260548bbb8STheodore Ts'o return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 30270548bbb8STheodore Ts'o sbi->s_itb_per_group + 2); 30280548bbb8STheodore Ts'o 3029952fc18eSTheodore Ts'o first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3030952fc18eSTheodore Ts'o (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3031952fc18eSTheodore Ts'o last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3032952fc18eSTheodore Ts'o for (i = 0; i < ngroups; i++) { 3033952fc18eSTheodore Ts'o gdp = ext4_get_group_desc(sb, i, NULL); 3034952fc18eSTheodore Ts'o b = ext4_block_bitmap(sb, gdp); 3035952fc18eSTheodore Ts'o if (b >= first_block && b <= last_block) { 3036952fc18eSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3037952fc18eSTheodore Ts'o count++; 3038952fc18eSTheodore Ts'o } 3039952fc18eSTheodore Ts'o b = ext4_inode_bitmap(sb, gdp); 3040952fc18eSTheodore Ts'o if (b >= first_block && b <= last_block) { 3041952fc18eSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3042952fc18eSTheodore Ts'o count++; 3043952fc18eSTheodore Ts'o } 3044952fc18eSTheodore Ts'o b = ext4_inode_table(sb, gdp); 3045952fc18eSTheodore Ts'o if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3046952fc18eSTheodore Ts'o for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3047952fc18eSTheodore Ts'o int c = EXT4_B2C(sbi, b - first_block); 3048952fc18eSTheodore Ts'o ext4_set_bit(c, buf); 3049952fc18eSTheodore Ts'o count++; 3050952fc18eSTheodore Ts'o } 3051952fc18eSTheodore Ts'o if (i != grp) 3052952fc18eSTheodore Ts'o continue; 3053952fc18eSTheodore Ts'o s = 0; 3054952fc18eSTheodore Ts'o if (ext4_bg_has_super(sb, grp)) { 3055952fc18eSTheodore Ts'o ext4_set_bit(s++, buf); 3056952fc18eSTheodore Ts'o count++; 3057952fc18eSTheodore Ts'o } 3058952fc18eSTheodore Ts'o for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) { 3059952fc18eSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3060952fc18eSTheodore Ts'o count++; 3061952fc18eSTheodore Ts'o } 3062952fc18eSTheodore Ts'o } 3063952fc18eSTheodore Ts'o if (!count) 3064952fc18eSTheodore Ts'o return 0; 3065952fc18eSTheodore Ts'o return EXT4_CLUSTERS_PER_GROUP(sb) - 3066952fc18eSTheodore Ts'o ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3067952fc18eSTheodore Ts'o } 3068952fc18eSTheodore Ts'o 3069952fc18eSTheodore Ts'o /* 3070952fc18eSTheodore Ts'o * Compute the overhead and stash it in sbi->s_overhead 3071952fc18eSTheodore Ts'o */ 3072952fc18eSTheodore Ts'o int ext4_calculate_overhead(struct super_block *sb) 3073952fc18eSTheodore Ts'o { 3074952fc18eSTheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 3075952fc18eSTheodore Ts'o struct ext4_super_block *es = sbi->s_es; 3076952fc18eSTheodore Ts'o ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3077952fc18eSTheodore Ts'o ext4_fsblk_t overhead = 0; 30784fdb5543SDmitry Monakhov char *buf = (char *) get_zeroed_page(GFP_NOFS); 3079952fc18eSTheodore Ts'o 3080952fc18eSTheodore Ts'o if (!buf) 3081952fc18eSTheodore Ts'o return -ENOMEM; 3082952fc18eSTheodore Ts'o 3083952fc18eSTheodore Ts'o /* 3084952fc18eSTheodore Ts'o * Compute the overhead (FS structures). This is constant 3085952fc18eSTheodore Ts'o * for a given filesystem unless the number of block groups 3086952fc18eSTheodore Ts'o * changes so we cache the previous value until it does. 3087952fc18eSTheodore Ts'o */ 3088952fc18eSTheodore Ts'o 3089952fc18eSTheodore Ts'o /* 3090952fc18eSTheodore Ts'o * All of the blocks before first_data_block are overhead 3091952fc18eSTheodore Ts'o */ 3092952fc18eSTheodore Ts'o overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3093952fc18eSTheodore Ts'o 3094952fc18eSTheodore Ts'o /* 3095952fc18eSTheodore Ts'o * Add the overhead found in each block group 3096952fc18eSTheodore Ts'o */ 3097952fc18eSTheodore Ts'o for (i = 0; i < ngroups; i++) { 3098952fc18eSTheodore Ts'o int blks; 3099952fc18eSTheodore Ts'o 3100952fc18eSTheodore Ts'o blks = count_overhead(sb, i, buf); 3101952fc18eSTheodore Ts'o overhead += blks; 3102952fc18eSTheodore Ts'o if (blks) 3103952fc18eSTheodore Ts'o memset(buf, 0, PAGE_SIZE); 3104952fc18eSTheodore Ts'o cond_resched(); 3105952fc18eSTheodore Ts'o } 3106b003b524SEric Sandeen /* Add the internal journal blocks as well */ 3107b003b524SEric Sandeen if (sbi->s_journal && !sbi->journal_bdev) 3108810da240SLukas Czerner overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen); 31090875a2b4SEric Sandeen 3110952fc18eSTheodore Ts'o sbi->s_overhead = overhead; 3111952fc18eSTheodore Ts'o smp_wmb(); 3112952fc18eSTheodore Ts'o free_page((unsigned long) buf); 3113952fc18eSTheodore Ts'o return 0; 3114952fc18eSTheodore Ts'o } 3115952fc18eSTheodore Ts'o 3116b5799018STheodore Ts'o static void ext4_set_resv_clusters(struct super_block *sb) 311727dd4385SLukas Czerner { 311827dd4385SLukas Czerner ext4_fsblk_t resv_clusters; 3119b5799018STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 312027dd4385SLukas Czerner 312127dd4385SLukas Czerner /* 312230fac0f7SJan Kara * There's no need to reserve anything when we aren't using extents. 312330fac0f7SJan Kara * The space estimates are exact, there are no unwritten extents, 312430fac0f7SJan Kara * hole punching doesn't need new metadata... This is needed especially 312530fac0f7SJan Kara * to keep ext2/3 backward compatibility. 312630fac0f7SJan Kara */ 3127e2b911c5SDarrick J. Wong if (!ext4_has_feature_extents(sb)) 3128b5799018STheodore Ts'o return; 312930fac0f7SJan Kara /* 313027dd4385SLukas Czerner * By default we reserve 2% or 4096 clusters, whichever is smaller. 313127dd4385SLukas Czerner * This should cover the situations where we can not afford to run 313227dd4385SLukas Czerner * out of space like for example punch hole, or converting 3133556615dcSLukas Czerner * unwritten extents in delalloc path. In most cases such 313427dd4385SLukas Czerner * allocation would require 1, or 2 blocks, higher numbers are 313527dd4385SLukas Czerner * very rare. 313627dd4385SLukas Czerner */ 3137b5799018STheodore Ts'o resv_clusters = (ext4_blocks_count(sbi->s_es) >> 3138b5799018STheodore Ts'o sbi->s_cluster_bits); 313927dd4385SLukas Czerner 314027dd4385SLukas Czerner do_div(resv_clusters, 50); 314127dd4385SLukas Czerner resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 314227dd4385SLukas Czerner 3143b5799018STheodore Ts'o atomic64_set(&sbi->s_resv_clusters, resv_clusters); 314427dd4385SLukas Czerner } 314527dd4385SLukas Czerner 3146617ba13bSMingming Cao static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3147ac27a0ecSDave Kleikamp { 3148d4c402d9SCurt Wohlgemuth char *orig_data = kstrdup(data, GFP_KERNEL); 3149ac27a0ecSDave Kleikamp struct buffer_head *bh; 3150617ba13bSMingming Cao struct ext4_super_block *es = NULL; 3151617ba13bSMingming Cao struct ext4_sb_info *sbi; 3152617ba13bSMingming Cao ext4_fsblk_t block; 3153617ba13bSMingming Cao ext4_fsblk_t sb_block = get_sb_block(&data); 315470bbb3e0SAndrew Morton ext4_fsblk_t logical_sb_block; 3155ac27a0ecSDave Kleikamp unsigned long offset = 0; 3156ac27a0ecSDave Kleikamp unsigned long journal_devnum = 0; 3157ac27a0ecSDave Kleikamp unsigned long def_mount_opts; 3158ac27a0ecSDave Kleikamp struct inode *root; 31590390131bSFrank Mayhar const char *descr; 3160dcc7dae3SCyrill Gorcunov int ret = -ENOMEM; 3161281b5995STheodore Ts'o int blocksize, clustersize; 31624ec11028STheodore Ts'o unsigned int db_count; 31634ec11028STheodore Ts'o unsigned int i; 3164281b5995STheodore Ts'o int needs_recovery, has_huge_files, has_bigalloc; 3165bd81d8eeSLaurent Vivier __u64 blocks_count; 316607aa2ea1SLukas Czerner int err = 0; 3167b3881f74STheodore Ts'o unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3168bfff6873SLukas Czerner ext4_group_t first_not_zeroed; 3169ac27a0ecSDave Kleikamp 3170ac27a0ecSDave Kleikamp sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 3171ac27a0ecSDave Kleikamp if (!sbi) 3172dcc7dae3SCyrill Gorcunov goto out_free_orig; 3173705895b6SPekka Enberg 3174705895b6SPekka Enberg sbi->s_blockgroup_lock = 3175705895b6SPekka Enberg kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 3176705895b6SPekka Enberg if (!sbi->s_blockgroup_lock) { 3177705895b6SPekka Enberg kfree(sbi); 3178dcc7dae3SCyrill Gorcunov goto out_free_orig; 3179705895b6SPekka Enberg } 3180ac27a0ecSDave Kleikamp sb->s_fs_info = sbi; 31812c0544b2STheodore Ts'o sbi->s_sb = sb; 3182240799cdSTheodore Ts'o sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3183d9c9bef1SMiklos Szeredi sbi->s_sb_block = sb_block; 3184f613dfcbSTheodore Ts'o if (sb->s_bdev->bd_part) 3185f613dfcbSTheodore Ts'o sbi->s_sectors_written_start = 3186f613dfcbSTheodore Ts'o part_stat_read(sb->s_bdev->bd_part, sectors[1]); 3187ac27a0ecSDave Kleikamp 31889f6200bbSTheodore Ts'o /* Cleanup superblock name */ 3189ec3904dcSRasmus Villemoes strreplace(sb->s_id, '/', '!'); 31909f6200bbSTheodore Ts'o 319107aa2ea1SLukas Czerner /* -EINVAL is default */ 3192dcc7dae3SCyrill Gorcunov ret = -EINVAL; 3193617ba13bSMingming Cao blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3194ac27a0ecSDave Kleikamp if (!blocksize) { 3195b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3196ac27a0ecSDave Kleikamp goto out_fail; 3197ac27a0ecSDave Kleikamp } 3198ac27a0ecSDave Kleikamp 3199ac27a0ecSDave Kleikamp /* 3200617ba13bSMingming Cao * The ext4 superblock will not be buffer aligned for other than 1kB 3201ac27a0ecSDave Kleikamp * block sizes. We need to calculate the offset from buffer start. 3202ac27a0ecSDave Kleikamp */ 3203617ba13bSMingming Cao if (blocksize != EXT4_MIN_BLOCK_SIZE) { 320470bbb3e0SAndrew Morton logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 320570bbb3e0SAndrew Morton offset = do_div(logical_sb_block, blocksize); 3206ac27a0ecSDave Kleikamp } else { 320770bbb3e0SAndrew Morton logical_sb_block = sb_block; 3208ac27a0ecSDave Kleikamp } 3209ac27a0ecSDave Kleikamp 3210a8ac900bSGioh Kim if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) { 3211b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3212ac27a0ecSDave Kleikamp goto out_fail; 3213ac27a0ecSDave Kleikamp } 3214ac27a0ecSDave Kleikamp /* 3215ac27a0ecSDave Kleikamp * Note: s_es must be initialized as soon as possible because 3216617ba13bSMingming Cao * some ext4 macro-instructions depend on its value 3217ac27a0ecSDave Kleikamp */ 32182716b802STheodore Ts'o es = (struct ext4_super_block *) (bh->b_data + offset); 3219ac27a0ecSDave Kleikamp sbi->s_es = es; 3220ac27a0ecSDave Kleikamp sb->s_magic = le16_to_cpu(es->s_magic); 3221617ba13bSMingming Cao if (sb->s_magic != EXT4_SUPER_MAGIC) 3222617ba13bSMingming Cao goto cantfind_ext4; 3223afc32f7eSTheodore Ts'o sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3224ac27a0ecSDave Kleikamp 3225feb0ab32SDarrick J. Wong /* Warn if metadata_csum and gdt_csum are both set. */ 3226e2b911c5SDarrick J. Wong if (ext4_has_feature_metadata_csum(sb) && 3227e2b911c5SDarrick J. Wong ext4_has_feature_gdt_csum(sb)) 3228363307e6SJakub Wilk ext4_warning(sb, "metadata_csum and uninit_bg are " 3229feb0ab32SDarrick J. Wong "redundant flags; please run fsck."); 3230feb0ab32SDarrick J. Wong 3231d25425f8SDarrick J. Wong /* Check for a known checksum algorithm */ 3232d25425f8SDarrick J. Wong if (!ext4_verify_csum_type(sb, es)) { 3233d25425f8SDarrick J. Wong ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3234d25425f8SDarrick J. Wong "unknown checksum algorithm."); 3235d25425f8SDarrick J. Wong silent = 1; 3236d25425f8SDarrick J. Wong goto cantfind_ext4; 3237d25425f8SDarrick J. Wong } 3238d25425f8SDarrick J. Wong 32390441984aSDarrick J. Wong /* Load the checksum driver */ 3240e2b911c5SDarrick J. Wong if (ext4_has_feature_metadata_csum(sb)) { 32410441984aSDarrick J. Wong sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 32420441984aSDarrick J. Wong if (IS_ERR(sbi->s_chksum_driver)) { 32430441984aSDarrick J. Wong ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 32440441984aSDarrick J. Wong ret = PTR_ERR(sbi->s_chksum_driver); 32450441984aSDarrick J. Wong sbi->s_chksum_driver = NULL; 32460441984aSDarrick J. Wong goto failed_mount; 32470441984aSDarrick J. Wong } 32480441984aSDarrick J. Wong } 32490441984aSDarrick J. Wong 3250a9c47317SDarrick J. Wong /* Check superblock checksum */ 3251a9c47317SDarrick J. Wong if (!ext4_superblock_csum_verify(sb, es)) { 3252a9c47317SDarrick J. Wong ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3253a9c47317SDarrick J. Wong "invalid superblock checksum. Run e2fsck?"); 3254a9c47317SDarrick J. Wong silent = 1; 32556a797d27SDarrick J. Wong ret = -EFSBADCRC; 3256a9c47317SDarrick J. Wong goto cantfind_ext4; 3257a9c47317SDarrick J. Wong } 3258a9c47317SDarrick J. Wong 3259a9c47317SDarrick J. Wong /* Precompute checksum seed for all metadata */ 3260e2b911c5SDarrick J. Wong if (ext4_has_feature_csum_seed(sb)) 32618c81bd8fSDarrick J. Wong sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 32628c81bd8fSDarrick J. Wong else if (ext4_has_metadata_csum(sb)) 3263a9c47317SDarrick J. Wong sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 3264a9c47317SDarrick J. Wong sizeof(es->s_uuid)); 3265a9c47317SDarrick J. Wong 3266ac27a0ecSDave Kleikamp /* Set defaults before we parse the mount options */ 3267ac27a0ecSDave Kleikamp def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3268fd8c37ecSTheodore Ts'o set_opt(sb, INIT_INODE_TABLE); 3269617ba13bSMingming Cao if (def_mount_opts & EXT4_DEFM_DEBUG) 3270fd8c37ecSTheodore Ts'o set_opt(sb, DEBUG); 327187f26807STheodore Ts'o if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 3272fd8c37ecSTheodore Ts'o set_opt(sb, GRPID); 3273617ba13bSMingming Cao if (def_mount_opts & EXT4_DEFM_UID16) 3274fd8c37ecSTheodore Ts'o set_opt(sb, NO_UID32); 3275ea663336SEric Sandeen /* xattr user namespace & acls are now defaulted on */ 3276fd8c37ecSTheodore Ts'o set_opt(sb, XATTR_USER); 327703010a33STheodore Ts'o #ifdef CONFIG_EXT4_FS_POSIX_ACL 3278fd8c37ecSTheodore Ts'o set_opt(sb, POSIX_ACL); 32792e7842b8SHugh Dickins #endif 328098c1a759SDarrick J. Wong /* don't forget to enable journal_csum when metadata_csum is enabled. */ 328198c1a759SDarrick J. Wong if (ext4_has_metadata_csum(sb)) 328298c1a759SDarrick J. Wong set_opt(sb, JOURNAL_CHECKSUM); 328398c1a759SDarrick J. Wong 3284617ba13bSMingming Cao if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 3285fd8c37ecSTheodore Ts'o set_opt(sb, JOURNAL_DATA); 3286617ba13bSMingming Cao else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 3287fd8c37ecSTheodore Ts'o set_opt(sb, ORDERED_DATA); 3288617ba13bSMingming Cao else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 3289fd8c37ecSTheodore Ts'o set_opt(sb, WRITEBACK_DATA); 3290ac27a0ecSDave Kleikamp 3291617ba13bSMingming Cao if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 3292fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_PANIC); 3293bb4f397aSAneesh Kumar K.V else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 3294fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_CONT); 3295bb4f397aSAneesh Kumar K.V else 3296fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_RO); 329745f1a9c3SDarrick J. Wong /* block_validity enabled by default; disable with noblock_validity */ 3298fd8c37ecSTheodore Ts'o set_opt(sb, BLOCK_VALIDITY); 32998b67f04aSTheodore Ts'o if (def_mount_opts & EXT4_DEFM_DISCARD) 3300fd8c37ecSTheodore Ts'o set_opt(sb, DISCARD); 3301ac27a0ecSDave Kleikamp 330208cefc7aSEric W. Biederman sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 330308cefc7aSEric W. Biederman sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 330430773840STheodore Ts'o sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 330530773840STheodore Ts'o sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 330630773840STheodore Ts'o sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3307ac27a0ecSDave Kleikamp 33088b67f04aSTheodore Ts'o if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 3309fd8c37ecSTheodore Ts'o set_opt(sb, BARRIER); 3310ac27a0ecSDave Kleikamp 33111e2462f9SMingming Cao /* 3312dd919b98SAneesh Kumar K.V * enable delayed allocation by default 3313dd919b98SAneesh Kumar K.V * Use -o nodelalloc to turn it off 3314dd919b98SAneesh Kumar K.V */ 3315bc0b75f7STheodore Ts'o if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 33168b67f04aSTheodore Ts'o ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 3317fd8c37ecSTheodore Ts'o set_opt(sb, DELALLOC); 3318dd919b98SAneesh Kumar K.V 331951ce6511SLukas Czerner /* 332051ce6511SLukas Czerner * set default s_li_wait_mult for lazyinit, for the case there is 332151ce6511SLukas Czerner * no mount option specified. 332251ce6511SLukas Czerner */ 332351ce6511SLukas Czerner sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 332451ce6511SLukas Czerner 33258b67f04aSTheodore Ts'o if (!parse_options((char *) sbi->s_es->s_mount_opts, sb, 3326661aa520SEric Sandeen &journal_devnum, &journal_ioprio, 0)) { 33278b67f04aSTheodore Ts'o ext4_msg(sb, KERN_WARNING, 33288b67f04aSTheodore Ts'o "failed to parse options in superblock: %s", 33298b67f04aSTheodore Ts'o sbi->s_es->s_mount_opts); 33308b67f04aSTheodore Ts'o } 33315a916be1STheodore Ts'o sbi->s_def_mount_opt = sbi->s_mount_opt; 3332b3881f74STheodore Ts'o if (!parse_options((char *) data, sb, &journal_devnum, 3333661aa520SEric Sandeen &journal_ioprio, 0)) 3334ac27a0ecSDave Kleikamp goto failed_mount; 3335ac27a0ecSDave Kleikamp 333656889787STheodore Ts'o if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 333756889787STheodore Ts'o printk_once(KERN_WARNING "EXT4-fs: Warning: mounting " 333856889787STheodore Ts'o "with data=journal disables delayed " 333956889787STheodore Ts'o "allocation and O_DIRECT support!\n"); 334056889787STheodore Ts'o if (test_opt2(sb, EXPLICIT_DELALLOC)) { 334156889787STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't mount with " 334256889787STheodore Ts'o "both data=journal and delalloc"); 334356889787STheodore Ts'o goto failed_mount; 334456889787STheodore Ts'o } 334556889787STheodore Ts'o if (test_opt(sb, DIOREAD_NOLOCK)) { 334656889787STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't mount with " 33476ae6514bSPiotr Sarna "both data=journal and dioread_nolock"); 334856889787STheodore Ts'o goto failed_mount; 334956889787STheodore Ts'o } 3350923ae0ffSRoss Zwisler if (test_opt(sb, DAX)) { 3351923ae0ffSRoss Zwisler ext4_msg(sb, KERN_ERR, "can't mount with " 3352923ae0ffSRoss Zwisler "both data=journal and dax"); 3353923ae0ffSRoss Zwisler goto failed_mount; 3354923ae0ffSRoss Zwisler } 335556889787STheodore Ts'o if (test_opt(sb, DELALLOC)) 335656889787STheodore Ts'o clear_opt(sb, DELALLOC); 3357001e4a87STejun Heo } else { 3358001e4a87STejun Heo sb->s_iflags |= SB_I_CGROUPWB; 335956889787STheodore Ts'o } 336056889787STheodore Ts'o 3361ac27a0ecSDave Kleikamp sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 3362482a7425SDmitry Monakhov (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 3363ac27a0ecSDave Kleikamp 3364617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 3365e2b911c5SDarrick J. Wong (ext4_has_compat_features(sb) || 3366e2b911c5SDarrick J. Wong ext4_has_ro_compat_features(sb) || 3367e2b911c5SDarrick J. Wong ext4_has_incompat_features(sb))) 3368b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 3369b31e1552SEric Sandeen "feature flags set on rev 0 fs, " 3370b31e1552SEric Sandeen "running e2fsck is recommended"); 3371469108ffSTheodore Tso 3372ed3654ebSTheodore Ts'o if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 3373ed3654ebSTheodore Ts'o set_opt2(sb, HURD_COMPAT); 3374e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb)) { 3375ed3654ebSTheodore Ts'o ext4_msg(sb, KERN_ERR, 3376ed3654ebSTheodore Ts'o "The Hurd can't support 64-bit file systems"); 3377ed3654ebSTheodore Ts'o goto failed_mount; 3378ed3654ebSTheodore Ts'o } 3379ed3654ebSTheodore Ts'o } 3380ed3654ebSTheodore Ts'o 33812035e776STheodore Ts'o if (IS_EXT2_SB(sb)) { 33822035e776STheodore Ts'o if (ext2_feature_set_ok(sb)) 33832035e776STheodore Ts'o ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 33842035e776STheodore Ts'o "using the ext4 subsystem"); 33852035e776STheodore Ts'o else { 33862035e776STheodore Ts'o ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 33872035e776STheodore Ts'o "to feature incompatibilities"); 33882035e776STheodore Ts'o goto failed_mount; 33892035e776STheodore Ts'o } 33902035e776STheodore Ts'o } 33912035e776STheodore Ts'o 33922035e776STheodore Ts'o if (IS_EXT3_SB(sb)) { 33932035e776STheodore Ts'o if (ext3_feature_set_ok(sb)) 33942035e776STheodore Ts'o ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 33952035e776STheodore Ts'o "using the ext4 subsystem"); 33962035e776STheodore Ts'o else { 33972035e776STheodore Ts'o ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 33982035e776STheodore Ts'o "to feature incompatibilities"); 33992035e776STheodore Ts'o goto failed_mount; 34002035e776STheodore Ts'o } 34012035e776STheodore Ts'o } 34022035e776STheodore Ts'o 3403469108ffSTheodore Tso /* 3404ac27a0ecSDave Kleikamp * Check feature flags regardless of the revision level, since we 3405ac27a0ecSDave Kleikamp * previously didn't change the revision level when setting the flags, 3406ac27a0ecSDave Kleikamp * so there is a chance incompat flags are set on a rev 0 filesystem. 3407ac27a0ecSDave Kleikamp */ 3408a13fb1a4SEric Sandeen if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY))) 3409ac27a0ecSDave Kleikamp goto failed_mount; 3410a13fb1a4SEric Sandeen 3411261cb20cSJan Kara blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 3412617ba13bSMingming Cao if (blocksize < EXT4_MIN_BLOCK_SIZE || 3413617ba13bSMingming Cao blocksize > EXT4_MAX_BLOCK_SIZE) { 3414b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3415b31e1552SEric Sandeen "Unsupported filesystem blocksize %d", blocksize); 3416ac27a0ecSDave Kleikamp goto failed_mount; 3417ac27a0ecSDave Kleikamp } 3418ac27a0ecSDave Kleikamp 3419923ae0ffSRoss Zwisler if (sbi->s_mount_opt & EXT4_MOUNT_DAX) { 342087eefeb4SToshi Kani err = bdev_dax_supported(sb, blocksize); 342187eefeb4SToshi Kani if (err) 3422923ae0ffSRoss Zwisler goto failed_mount; 3423923ae0ffSRoss Zwisler } 3424923ae0ffSRoss Zwisler 3425e2b911c5SDarrick J. Wong if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 34266ddb2447STheodore Ts'o ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 34276ddb2447STheodore Ts'o es->s_encryption_level); 34286ddb2447STheodore Ts'o goto failed_mount; 34296ddb2447STheodore Ts'o } 34306ddb2447STheodore Ts'o 3431ac27a0ecSDave Kleikamp if (sb->s_blocksize != blocksize) { 3432ce40733cSAneesh Kumar K.V /* Validate the filesystem blocksize */ 3433ce40733cSAneesh Kumar K.V if (!sb_set_blocksize(sb, blocksize)) { 3434b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "bad block size %d", 3435ce40733cSAneesh Kumar K.V blocksize); 3436ac27a0ecSDave Kleikamp goto failed_mount; 3437ac27a0ecSDave Kleikamp } 3438ac27a0ecSDave Kleikamp 3439ac27a0ecSDave Kleikamp brelse(bh); 344070bbb3e0SAndrew Morton logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 344170bbb3e0SAndrew Morton offset = do_div(logical_sb_block, blocksize); 3442a8ac900bSGioh Kim bh = sb_bread_unmovable(sb, logical_sb_block); 3443ac27a0ecSDave Kleikamp if (!bh) { 3444b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3445b31e1552SEric Sandeen "Can't read superblock on 2nd try"); 3446ac27a0ecSDave Kleikamp goto failed_mount; 3447ac27a0ecSDave Kleikamp } 34482716b802STheodore Ts'o es = (struct ext4_super_block *)(bh->b_data + offset); 3449ac27a0ecSDave Kleikamp sbi->s_es = es; 3450617ba13bSMingming Cao if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 3451b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3452b31e1552SEric Sandeen "Magic mismatch, very weird!"); 3453ac27a0ecSDave Kleikamp goto failed_mount; 3454ac27a0ecSDave Kleikamp } 3455ac27a0ecSDave Kleikamp } 3456ac27a0ecSDave Kleikamp 3457e2b911c5SDarrick J. Wong has_huge_files = ext4_has_feature_huge_file(sb); 3458f287a1a5STheodore Ts'o sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 3459f287a1a5STheodore Ts'o has_huge_files); 3460f287a1a5STheodore Ts'o sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 3461ac27a0ecSDave Kleikamp 3462617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 3463617ba13bSMingming Cao sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 3464617ba13bSMingming Cao sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 3465ac27a0ecSDave Kleikamp } else { 3466ac27a0ecSDave Kleikamp sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 3467ac27a0ecSDave Kleikamp sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 3468617ba13bSMingming Cao if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 34691330593eSVignesh Babu (!is_power_of_2(sbi->s_inode_size)) || 3470ac27a0ecSDave Kleikamp (sbi->s_inode_size > blocksize)) { 3471b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3472b31e1552SEric Sandeen "unsupported inode size: %d", 3473ac27a0ecSDave Kleikamp sbi->s_inode_size); 3474ac27a0ecSDave Kleikamp goto failed_mount; 3475ac27a0ecSDave Kleikamp } 3476ef7f3835SKalpak Shah if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) 3477ef7f3835SKalpak Shah sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); 3478ac27a0ecSDave Kleikamp } 34790b8e58a1SAndreas Dilger 34800d1ee42fSAlexandre Ratchov sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 3481e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb)) { 34828fadc143SAlexandre Ratchov if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 34830d1ee42fSAlexandre Ratchov sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 3484d8ea6cf8Svignesh babu !is_power_of_2(sbi->s_desc_size)) { 3485b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3486b31e1552SEric Sandeen "unsupported descriptor size %lu", 34870d1ee42fSAlexandre Ratchov sbi->s_desc_size); 34880d1ee42fSAlexandre Ratchov goto failed_mount; 34890d1ee42fSAlexandre Ratchov } 34900d1ee42fSAlexandre Ratchov } else 34910d1ee42fSAlexandre Ratchov sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 34920b8e58a1SAndreas Dilger 3493ac27a0ecSDave Kleikamp sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 3494ac27a0ecSDave Kleikamp sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 3495b47b6f38SAndries E. Brouwer if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0) 3496617ba13bSMingming Cao goto cantfind_ext4; 34970b8e58a1SAndreas Dilger 3498617ba13bSMingming Cao sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 3499ac27a0ecSDave Kleikamp if (sbi->s_inodes_per_block == 0) 3500617ba13bSMingming Cao goto cantfind_ext4; 3501ac27a0ecSDave Kleikamp sbi->s_itb_per_group = sbi->s_inodes_per_group / 3502ac27a0ecSDave Kleikamp sbi->s_inodes_per_block; 35030d1ee42fSAlexandre Ratchov sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 3504ac27a0ecSDave Kleikamp sbi->s_sbh = bh; 3505ac27a0ecSDave Kleikamp sbi->s_mount_state = le16_to_cpu(es->s_state); 3506e57aa839SFengguang Wu sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 3507e57aa839SFengguang Wu sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 35080b8e58a1SAndreas Dilger 3509ac27a0ecSDave Kleikamp for (i = 0; i < 4; i++) 3510ac27a0ecSDave Kleikamp sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 3511ac27a0ecSDave Kleikamp sbi->s_def_hash_version = es->s_def_hash_version; 3512e2b911c5SDarrick J. Wong if (ext4_has_feature_dir_index(sb)) { 3513f99b2589STheodore Ts'o i = le32_to_cpu(es->s_flags); 3514f99b2589STheodore Ts'o if (i & EXT2_FLAGS_UNSIGNED_HASH) 3515f99b2589STheodore Ts'o sbi->s_hash_unsigned = 3; 3516f99b2589STheodore Ts'o else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 3517f99b2589STheodore Ts'o #ifdef __CHAR_UNSIGNED__ 351823301410STheodore Ts'o if (!(sb->s_flags & MS_RDONLY)) 351923301410STheodore Ts'o es->s_flags |= 352023301410STheodore Ts'o cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 3521f99b2589STheodore Ts'o sbi->s_hash_unsigned = 3; 3522f99b2589STheodore Ts'o #else 352323301410STheodore Ts'o if (!(sb->s_flags & MS_RDONLY)) 352423301410STheodore Ts'o es->s_flags |= 352523301410STheodore Ts'o cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 3526f99b2589STheodore Ts'o #endif 3527f99b2589STheodore Ts'o } 352823301410STheodore Ts'o } 3529ac27a0ecSDave Kleikamp 3530281b5995STheodore Ts'o /* Handle clustersize */ 3531281b5995STheodore Ts'o clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 3532e2b911c5SDarrick J. Wong has_bigalloc = ext4_has_feature_bigalloc(sb); 3533281b5995STheodore Ts'o if (has_bigalloc) { 3534281b5995STheodore Ts'o if (clustersize < blocksize) { 3535281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, 3536281b5995STheodore Ts'o "cluster size (%d) smaller than " 3537281b5995STheodore Ts'o "block size (%d)", clustersize, blocksize); 3538281b5995STheodore Ts'o goto failed_mount; 3539281b5995STheodore Ts'o } 3540281b5995STheodore Ts'o sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 3541281b5995STheodore Ts'o le32_to_cpu(es->s_log_block_size); 3542281b5995STheodore Ts'o sbi->s_clusters_per_group = 3543281b5995STheodore Ts'o le32_to_cpu(es->s_clusters_per_group); 3544281b5995STheodore Ts'o if (sbi->s_clusters_per_group > blocksize * 8) { 3545281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, 3546281b5995STheodore Ts'o "#clusters per group too big: %lu", 3547281b5995STheodore Ts'o sbi->s_clusters_per_group); 3548281b5995STheodore Ts'o goto failed_mount; 3549281b5995STheodore Ts'o } 3550281b5995STheodore Ts'o if (sbi->s_blocks_per_group != 3551281b5995STheodore Ts'o (sbi->s_clusters_per_group * (clustersize / blocksize))) { 3552281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 3553281b5995STheodore Ts'o "clusters per group (%lu) inconsistent", 3554281b5995STheodore Ts'o sbi->s_blocks_per_group, 3555281b5995STheodore Ts'o sbi->s_clusters_per_group); 3556281b5995STheodore Ts'o goto failed_mount; 3557281b5995STheodore Ts'o } 3558281b5995STheodore Ts'o } else { 3559281b5995STheodore Ts'o if (clustersize != blocksize) { 3560281b5995STheodore Ts'o ext4_warning(sb, "fragment/cluster size (%d) != " 3561281b5995STheodore Ts'o "block size (%d)", clustersize, 3562281b5995STheodore Ts'o blocksize); 3563281b5995STheodore Ts'o clustersize = blocksize; 3564281b5995STheodore Ts'o } 3565ac27a0ecSDave Kleikamp if (sbi->s_blocks_per_group > blocksize * 8) { 3566b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3567b31e1552SEric Sandeen "#blocks per group too big: %lu", 3568ac27a0ecSDave Kleikamp sbi->s_blocks_per_group); 3569ac27a0ecSDave Kleikamp goto failed_mount; 3570ac27a0ecSDave Kleikamp } 3571281b5995STheodore Ts'o sbi->s_clusters_per_group = sbi->s_blocks_per_group; 3572281b5995STheodore Ts'o sbi->s_cluster_bits = 0; 3573281b5995STheodore Ts'o } 3574281b5995STheodore Ts'o sbi->s_cluster_ratio = clustersize / blocksize; 3575281b5995STheodore Ts'o 3576ac27a0ecSDave Kleikamp if (sbi->s_inodes_per_group > blocksize * 8) { 3577b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3578b31e1552SEric Sandeen "#inodes per group too big: %lu", 3579ac27a0ecSDave Kleikamp sbi->s_inodes_per_group); 3580ac27a0ecSDave Kleikamp goto failed_mount; 3581ac27a0ecSDave Kleikamp } 3582ac27a0ecSDave Kleikamp 3583960fd856STheodore Ts'o /* Do we have standard group size of clustersize * 8 blocks ? */ 3584960fd856STheodore Ts'o if (sbi->s_blocks_per_group == clustersize << 3) 3585960fd856STheodore Ts'o set_opt2(sb, STD_GROUP_SIZE); 3586960fd856STheodore Ts'o 3587bf43d84bSEric Sandeen /* 3588bf43d84bSEric Sandeen * Test whether we have more sectors than will fit in sector_t, 3589bf43d84bSEric Sandeen * and whether the max offset is addressable by the page cache. 3590bf43d84bSEric Sandeen */ 35915a9ae68aSDarrick J. Wong err = generic_check_addressable(sb->s_blocksize_bits, 359230ca22c7SPatrick J. LoPresti ext4_blocks_count(es)); 35935a9ae68aSDarrick J. Wong if (err) { 3594b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "filesystem" 3595bf43d84bSEric Sandeen " too large to mount safely on this system"); 3596ac27a0ecSDave Kleikamp if (sizeof(sector_t) < 8) 359790c699a9SBartlomiej Zolnierkiewicz ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); 3598ac27a0ecSDave Kleikamp goto failed_mount; 3599ac27a0ecSDave Kleikamp } 3600ac27a0ecSDave Kleikamp 3601617ba13bSMingming Cao if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 3602617ba13bSMingming Cao goto cantfind_ext4; 3603e7c95593SEric Sandeen 36040f2ddca6SFrom: Thiemo Nagel /* check blocks count against device size */ 36050f2ddca6SFrom: Thiemo Nagel blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 36060f2ddca6SFrom: Thiemo Nagel if (blocks_count && ext4_blocks_count(es) > blocks_count) { 3607b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 3608b31e1552SEric Sandeen "exceeds size of device (%llu blocks)", 36090f2ddca6SFrom: Thiemo Nagel ext4_blocks_count(es), blocks_count); 36100f2ddca6SFrom: Thiemo Nagel goto failed_mount; 36110f2ddca6SFrom: Thiemo Nagel } 36120f2ddca6SFrom: Thiemo Nagel 36134ec11028STheodore Ts'o /* 36144ec11028STheodore Ts'o * It makes no sense for the first data block to be beyond the end 36154ec11028STheodore Ts'o * of the filesystem. 36164ec11028STheodore Ts'o */ 36174ec11028STheodore Ts'o if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 3618b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 3619b31e1552SEric Sandeen "block %u is beyond end of filesystem (%llu)", 3620e7c95593SEric Sandeen le32_to_cpu(es->s_first_data_block), 36214ec11028STheodore Ts'o ext4_blocks_count(es)); 3622e7c95593SEric Sandeen goto failed_mount; 3623e7c95593SEric Sandeen } 3624bd81d8eeSLaurent Vivier blocks_count = (ext4_blocks_count(es) - 3625bd81d8eeSLaurent Vivier le32_to_cpu(es->s_first_data_block) + 3626bd81d8eeSLaurent Vivier EXT4_BLOCKS_PER_GROUP(sb) - 1); 3627bd81d8eeSLaurent Vivier do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 36284ec11028STheodore Ts'o if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 3629b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "groups count too large: %u " 36304ec11028STheodore Ts'o "(block count %llu, first data block %u, " 3631b31e1552SEric Sandeen "blocks per group %lu)", sbi->s_groups_count, 36324ec11028STheodore Ts'o ext4_blocks_count(es), 36334ec11028STheodore Ts'o le32_to_cpu(es->s_first_data_block), 36344ec11028STheodore Ts'o EXT4_BLOCKS_PER_GROUP(sb)); 36354ec11028STheodore Ts'o goto failed_mount; 36364ec11028STheodore Ts'o } 3637bd81d8eeSLaurent Vivier sbi->s_groups_count = blocks_count; 3638fb0a387dSEric Sandeen sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 3639fb0a387dSEric Sandeen (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 3640617ba13bSMingming Cao db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 3641617ba13bSMingming Cao EXT4_DESC_PER_BLOCK(sb); 3642f18a5f21STheodore Ts'o sbi->s_group_desc = ext4_kvmalloc(db_count * 3643f18a5f21STheodore Ts'o sizeof(struct buffer_head *), 3644ac27a0ecSDave Kleikamp GFP_KERNEL); 3645ac27a0ecSDave Kleikamp if (sbi->s_group_desc == NULL) { 3646b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "not enough memory"); 36472cde417dSTheodore Ts'o ret = -ENOMEM; 3648ac27a0ecSDave Kleikamp goto failed_mount; 3649ac27a0ecSDave Kleikamp } 3650ac27a0ecSDave Kleikamp 3651705895b6SPekka Enberg bgl_lock_init(sbi->s_blockgroup_lock); 3652ac27a0ecSDave Kleikamp 3653ac27a0ecSDave Kleikamp for (i = 0; i < db_count; i++) { 365470bbb3e0SAndrew Morton block = descriptor_loc(sb, logical_sb_block, i); 3655a8ac900bSGioh Kim sbi->s_group_desc[i] = sb_bread_unmovable(sb, block); 3656ac27a0ecSDave Kleikamp if (!sbi->s_group_desc[i]) { 3657b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3658b31e1552SEric Sandeen "can't read group descriptor %d", i); 3659ac27a0ecSDave Kleikamp db_count = i; 3660ac27a0ecSDave Kleikamp goto failed_mount2; 3661ac27a0ecSDave Kleikamp } 3662ac27a0ecSDave Kleikamp } 3663bfff6873SLukas Czerner if (!ext4_check_descriptors(sb, &first_not_zeroed)) { 3664b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 36656a797d27SDarrick J. Wong ret = -EFSCORRUPTED; 3666f9ae9cf5STheodore Ts'o goto failed_mount2; 3667ac27a0ecSDave Kleikamp } 3668772cb7c8SJose R. Santos 3669f9ae9cf5STheodore Ts'o sbi->s_gdb_count = db_count; 3670ac27a0ecSDave Kleikamp get_random_bytes(&sbi->s_next_generation, sizeof(u32)); 3671ac27a0ecSDave Kleikamp spin_lock_init(&sbi->s_next_gen_lock); 3672ac27a0ecSDave Kleikamp 367304ecddb7SJan Mrazek setup_timer(&sbi->s_err_report, print_daily_error_info, 367404ecddb7SJan Mrazek (unsigned long) sb); 367504496411STao Ma 3676a75ae78fSDmitry Monakhov /* Register extent status tree shrinker */ 3677eb68d0e2SZheng Liu if (ext4_es_register_shrinker(sbi)) 3678ce7e010aSTheodore Ts'o goto failed_mount3; 3679ce7e010aSTheodore Ts'o 3680c9de560dSAlex Tomas sbi->s_stripe = ext4_get_stripe_size(sbi); 368167a5da56SZheng Liu sbi->s_extent_max_zeroout_kb = 32; 3682c9de560dSAlex Tomas 3683f9ae9cf5STheodore Ts'o /* 3684f9ae9cf5STheodore Ts'o * set up enough so that it can read an inode 3685f9ae9cf5STheodore Ts'o */ 3686f9ae9cf5STheodore Ts'o sb->s_op = &ext4_sops; 3687617ba13bSMingming Cao sb->s_export_op = &ext4_export_ops; 3688617ba13bSMingming Cao sb->s_xattr = ext4_xattr_handlers; 3689ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 3690617ba13bSMingming Cao sb->dq_op = &ext4_quota_operations; 3691e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) 36921fa5efe3SJan Kara sb->s_qcop = &dquot_quotactl_sysfile_ops; 3693262b4662SJan Kara else 3694262b4662SJan Kara sb->s_qcop = &ext4_qctl_operations; 3695689c958cSLi Xi sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 3696ac27a0ecSDave Kleikamp #endif 3697f2fa2ffcSAneesh Kumar K.V memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 3698f2fa2ffcSAneesh Kumar K.V 3699ac27a0ecSDave Kleikamp INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 37003b9d4ed2STheodore Ts'o mutex_init(&sbi->s_orphan_lock); 3701ac27a0ecSDave Kleikamp 3702ac27a0ecSDave Kleikamp sb->s_root = NULL; 3703ac27a0ecSDave Kleikamp 3704ac27a0ecSDave Kleikamp needs_recovery = (es->s_last_orphan != 0 || 3705e2b911c5SDarrick J. Wong ext4_has_feature_journal_needs_recovery(sb)); 3706ac27a0ecSDave Kleikamp 3707e2b911c5SDarrick J. Wong if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY)) 3708c5e06d10SJohann Lombardi if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 370950460fe8SDarrick J. Wong goto failed_mount3a; 3710c5e06d10SJohann Lombardi 3711ac27a0ecSDave Kleikamp /* 3712ac27a0ecSDave Kleikamp * The first inode we look at is the journal inode. Don't try 3713ac27a0ecSDave Kleikamp * root first: it may be modified in the journal! 3714ac27a0ecSDave Kleikamp */ 3715e2b911c5SDarrick J. Wong if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 3716617ba13bSMingming Cao if (ext4_load_journal(sb, es, journal_devnum)) 371750460fe8SDarrick J. Wong goto failed_mount3a; 37180390131bSFrank Mayhar } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) && 3719e2b911c5SDarrick J. Wong ext4_has_feature_journal_needs_recovery(sb)) { 3720b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "required journal recovery " 3721b31e1552SEric Sandeen "suppressed and not mounted read-only"); 3722744692dcSJiaying Zhang goto failed_mount_wq; 3723ac27a0ecSDave Kleikamp } else { 37241e381f60SDmitry Monakhov /* Nojournal mode, all journal mount options are illegal */ 37251e381f60SDmitry Monakhov if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 37261e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 37271e381f60SDmitry Monakhov "journal_checksum, fs mounted w/o journal"); 37281e381f60SDmitry Monakhov goto failed_mount_wq; 37291e381f60SDmitry Monakhov } 37301e381f60SDmitry Monakhov if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 37311e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 37321e381f60SDmitry Monakhov "journal_async_commit, fs mounted w/o journal"); 37331e381f60SDmitry Monakhov goto failed_mount_wq; 37341e381f60SDmitry Monakhov } 37351e381f60SDmitry Monakhov if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 37361e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 37371e381f60SDmitry Monakhov "commit=%lu, fs mounted w/o journal", 37381e381f60SDmitry Monakhov sbi->s_commit_interval / HZ); 37391e381f60SDmitry Monakhov goto failed_mount_wq; 37401e381f60SDmitry Monakhov } 37411e381f60SDmitry Monakhov if (EXT4_MOUNT_DATA_FLAGS & 37421e381f60SDmitry Monakhov (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 37431e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 37441e381f60SDmitry Monakhov "data=, fs mounted w/o journal"); 37451e381f60SDmitry Monakhov goto failed_mount_wq; 37461e381f60SDmitry Monakhov } 37471e381f60SDmitry Monakhov sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM; 37481e381f60SDmitry Monakhov clear_opt(sb, JOURNAL_CHECKSUM); 3749fd8c37ecSTheodore Ts'o clear_opt(sb, DATA_FLAGS); 37500390131bSFrank Mayhar sbi->s_journal = NULL; 37510390131bSFrank Mayhar needs_recovery = 0; 37520390131bSFrank Mayhar goto no_journal; 3753ac27a0ecSDave Kleikamp } 3754ac27a0ecSDave Kleikamp 3755e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb) && 3756eb40a09cSJose R. Santos !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 3757eb40a09cSJose R. Santos JBD2_FEATURE_INCOMPAT_64BIT)) { 3758b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 3759744692dcSJiaying Zhang goto failed_mount_wq; 3760eb40a09cSJose R. Santos } 3761eb40a09cSJose R. Santos 376225ed6e8aSDarrick J. Wong if (!set_journal_csum_feature_set(sb)) { 376325ed6e8aSDarrick J. Wong ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 376425ed6e8aSDarrick J. Wong "feature set"); 376525ed6e8aSDarrick J. Wong goto failed_mount_wq; 3766d4da6c9cSLinus Torvalds } 3767818d276cSGirish Shilamkar 3768ac27a0ecSDave Kleikamp /* We have now updated the journal if required, so we can 3769ac27a0ecSDave Kleikamp * validate the data journaling mode. */ 3770ac27a0ecSDave Kleikamp switch (test_opt(sb, DATA_FLAGS)) { 3771ac27a0ecSDave Kleikamp case 0: 3772ac27a0ecSDave Kleikamp /* No mode set, assume a default based on the journal 377363f57933SAndrew Morton * capabilities: ORDERED_DATA if the journal can 377463f57933SAndrew Morton * cope, else JOURNAL_DATA 377563f57933SAndrew Morton */ 3776dab291afSMingming Cao if (jbd2_journal_check_available_features 3777dab291afSMingming Cao (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) 3778fd8c37ecSTheodore Ts'o set_opt(sb, ORDERED_DATA); 3779ac27a0ecSDave Kleikamp else 3780fd8c37ecSTheodore Ts'o set_opt(sb, JOURNAL_DATA); 3781ac27a0ecSDave Kleikamp break; 3782ac27a0ecSDave Kleikamp 3783617ba13bSMingming Cao case EXT4_MOUNT_ORDERED_DATA: 3784617ba13bSMingming Cao case EXT4_MOUNT_WRITEBACK_DATA: 3785dab291afSMingming Cao if (!jbd2_journal_check_available_features 3786dab291afSMingming Cao (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 3787b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Journal does not support " 3788b31e1552SEric Sandeen "requested data journaling mode"); 3789744692dcSJiaying Zhang goto failed_mount_wq; 3790ac27a0ecSDave Kleikamp } 3791ac27a0ecSDave Kleikamp default: 3792ac27a0ecSDave Kleikamp break; 3793ac27a0ecSDave Kleikamp } 3794b3881f74STheodore Ts'o set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 3795ac27a0ecSDave Kleikamp 379618aadd47SBobi Jam sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; 379718aadd47SBobi Jam 3798ce7e010aSTheodore Ts'o no_journal: 379982939d79SJan Kara sbi->s_mb_cache = ext4_xattr_create_cache(); 38009c191f70ST Makphaibulchoke if (!sbi->s_mb_cache) { 38019c191f70ST Makphaibulchoke ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache"); 38029c191f70ST Makphaibulchoke goto failed_mount_wq; 38039c191f70ST Makphaibulchoke } 38049c191f70ST Makphaibulchoke 3805e2b911c5SDarrick J. Wong if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) && 380609cbfeafSKirill A. Shutemov (blocksize != PAGE_SIZE)) { 38071cb767cdSTheodore Ts'o ext4_msg(sb, KERN_ERR, 38081cb767cdSTheodore Ts'o "Unsupported blocksize for fs encryption"); 38091cb767cdSTheodore Ts'o goto failed_mount_wq; 38101cb767cdSTheodore Ts'o } 38111cb767cdSTheodore Ts'o 3812e2b911c5SDarrick J. Wong if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) && 3813e2b911c5SDarrick J. Wong !ext4_has_feature_encrypt(sb)) { 3814e2b911c5SDarrick J. Wong ext4_set_feature_encrypt(sb); 38156ddb2447STheodore Ts'o ext4_commit_super(sb, 1); 38166ddb2447STheodore Ts'o } 38176ddb2447STheodore Ts'o 3818fd89d5f2STejun Heo /* 3819952fc18eSTheodore Ts'o * Get the # of file system overhead blocks from the 3820952fc18eSTheodore Ts'o * superblock if present. 3821952fc18eSTheodore Ts'o */ 3822952fc18eSTheodore Ts'o if (es->s_overhead_clusters) 3823952fc18eSTheodore Ts'o sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 3824952fc18eSTheodore Ts'o else { 382507aa2ea1SLukas Czerner err = ext4_calculate_overhead(sb); 382607aa2ea1SLukas Czerner if (err) 3827952fc18eSTheodore Ts'o goto failed_mount_wq; 3828952fc18eSTheodore Ts'o } 3829952fc18eSTheodore Ts'o 3830952fc18eSTheodore Ts'o /* 3831fd89d5f2STejun Heo * The maximum number of concurrent works can be high and 3832fd89d5f2STejun Heo * concurrency isn't really necessary. Limit it to 1. 3833fd89d5f2STejun Heo */ 38342e8fa54eSJan Kara EXT4_SB(sb)->rsv_conversion_wq = 38352e8fa54eSJan Kara alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 38362e8fa54eSJan Kara if (!EXT4_SB(sb)->rsv_conversion_wq) { 38372e8fa54eSJan Kara printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 383807aa2ea1SLukas Czerner ret = -ENOMEM; 38392e8fa54eSJan Kara goto failed_mount4; 38402e8fa54eSJan Kara } 38412e8fa54eSJan Kara 3842ac27a0ecSDave Kleikamp /* 3843dab291afSMingming Cao * The jbd2_journal_load will have done any necessary log recovery, 3844ac27a0ecSDave Kleikamp * so we can safely mount the rest of the filesystem now. 3845ac27a0ecSDave Kleikamp */ 3846ac27a0ecSDave Kleikamp 38471d1fe1eeSDavid Howells root = ext4_iget(sb, EXT4_ROOT_INO); 38481d1fe1eeSDavid Howells if (IS_ERR(root)) { 3849b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "get root inode failed"); 38501d1fe1eeSDavid Howells ret = PTR_ERR(root); 385132a9bb57SManish Katiyar root = NULL; 3852ac27a0ecSDave Kleikamp goto failed_mount4; 3853ac27a0ecSDave Kleikamp } 3854ac27a0ecSDave Kleikamp if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 3855b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 385694bf608aSAl Viro iput(root); 3857ac27a0ecSDave Kleikamp goto failed_mount4; 3858ac27a0ecSDave Kleikamp } 385948fde701SAl Viro sb->s_root = d_make_root(root); 38601d1fe1eeSDavid Howells if (!sb->s_root) { 3861b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "get root dentry failed"); 38621d1fe1eeSDavid Howells ret = -ENOMEM; 38631d1fe1eeSDavid Howells goto failed_mount4; 38641d1fe1eeSDavid Howells } 3865ac27a0ecSDave Kleikamp 38667e84b621SEric Sandeen if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY)) 38677e84b621SEric Sandeen sb->s_flags |= MS_RDONLY; 3868ef7f3835SKalpak Shah 3869ef7f3835SKalpak Shah /* determine the minimum size of new large inodes, if present */ 3870ef7f3835SKalpak Shah if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 3871ef7f3835SKalpak Shah sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 3872ef7f3835SKalpak Shah EXT4_GOOD_OLD_INODE_SIZE; 3873e2b911c5SDarrick J. Wong if (ext4_has_feature_extra_isize(sb)) { 3874ef7f3835SKalpak Shah if (sbi->s_want_extra_isize < 3875ef7f3835SKalpak Shah le16_to_cpu(es->s_want_extra_isize)) 3876ef7f3835SKalpak Shah sbi->s_want_extra_isize = 3877ef7f3835SKalpak Shah le16_to_cpu(es->s_want_extra_isize); 3878ef7f3835SKalpak Shah if (sbi->s_want_extra_isize < 3879ef7f3835SKalpak Shah le16_to_cpu(es->s_min_extra_isize)) 3880ef7f3835SKalpak Shah sbi->s_want_extra_isize = 3881ef7f3835SKalpak Shah le16_to_cpu(es->s_min_extra_isize); 3882ef7f3835SKalpak Shah } 3883ef7f3835SKalpak Shah } 3884ef7f3835SKalpak Shah /* Check if enough inode space is available */ 3885ef7f3835SKalpak Shah if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > 3886ef7f3835SKalpak Shah sbi->s_inode_size) { 3887ef7f3835SKalpak Shah sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 3888ef7f3835SKalpak Shah EXT4_GOOD_OLD_INODE_SIZE; 3889b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "required extra inode space not" 3890b31e1552SEric Sandeen "available"); 3891ef7f3835SKalpak Shah } 3892ef7f3835SKalpak Shah 3893b5799018STheodore Ts'o ext4_set_resv_clusters(sb); 389427dd4385SLukas Czerner 38956fd058f7STheodore Ts'o err = ext4_setup_system_zone(sb); 38966fd058f7STheodore Ts'o if (err) { 3897b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "failed to initialize system " 3898fbe845ddSCurt Wohlgemuth "zone (%d)", err); 3899f9ae9cf5STheodore Ts'o goto failed_mount4a; 3900f9ae9cf5STheodore Ts'o } 3901f9ae9cf5STheodore Ts'o 3902f9ae9cf5STheodore Ts'o ext4_ext_init(sb); 3903f9ae9cf5STheodore Ts'o err = ext4_mb_init(sb); 3904f9ae9cf5STheodore Ts'o if (err) { 3905f9ae9cf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 3906f9ae9cf5STheodore Ts'o err); 3907dcf2d804STao Ma goto failed_mount5; 3908c2774d84SAneesh Kumar K.V } 3909c2774d84SAneesh Kumar K.V 3910d5e03cbbSTheodore Ts'o block = ext4_count_free_clusters(sb); 3911d5e03cbbSTheodore Ts'o ext4_free_blocks_count_set(sbi->s_es, 3912d5e03cbbSTheodore Ts'o EXT4_C2B(sbi, block)); 3913908c7f19STejun Heo err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 3914908c7f19STejun Heo GFP_KERNEL); 3915d5e03cbbSTheodore Ts'o if (!err) { 3916d5e03cbbSTheodore Ts'o unsigned long freei = ext4_count_free_inodes(sb); 3917d5e03cbbSTheodore Ts'o sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 3918908c7f19STejun Heo err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 3919908c7f19STejun Heo GFP_KERNEL); 3920d5e03cbbSTheodore Ts'o } 3921d5e03cbbSTheodore Ts'o if (!err) 3922d5e03cbbSTheodore Ts'o err = percpu_counter_init(&sbi->s_dirs_counter, 3923908c7f19STejun Heo ext4_count_dirs(sb), GFP_KERNEL); 3924d5e03cbbSTheodore Ts'o if (!err) 3925908c7f19STejun Heo err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 3926908c7f19STejun Heo GFP_KERNEL); 3927c8585c6fSDaeho Jeong if (!err) 3928c8585c6fSDaeho Jeong err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem); 3929c8585c6fSDaeho Jeong 3930d5e03cbbSTheodore Ts'o if (err) { 3931d5e03cbbSTheodore Ts'o ext4_msg(sb, KERN_ERR, "insufficient memory"); 3932d5e03cbbSTheodore Ts'o goto failed_mount6; 3933d5e03cbbSTheodore Ts'o } 3934d5e03cbbSTheodore Ts'o 3935e2b911c5SDarrick J. Wong if (ext4_has_feature_flex_bg(sb)) 3936d5e03cbbSTheodore Ts'o if (!ext4_fill_flex_info(sb)) { 3937d5e03cbbSTheodore Ts'o ext4_msg(sb, KERN_ERR, 3938d5e03cbbSTheodore Ts'o "unable to initialize " 3939d5e03cbbSTheodore Ts'o "flex_bg meta info!"); 3940d5e03cbbSTheodore Ts'o goto failed_mount6; 3941d5e03cbbSTheodore Ts'o } 3942d5e03cbbSTheodore Ts'o 3943bfff6873SLukas Czerner err = ext4_register_li_request(sb, first_not_zeroed); 3944bfff6873SLukas Czerner if (err) 3945dcf2d804STao Ma goto failed_mount6; 3946bfff6873SLukas Czerner 3947b5799018STheodore Ts'o err = ext4_register_sysfs(sb); 3948dcf2d804STao Ma if (err) 3949dcf2d804STao Ma goto failed_mount7; 39503197ebdbSTheodore Ts'o 39519b2ff357SJan Kara #ifdef CONFIG_QUOTA 39529b2ff357SJan Kara /* Enable quota usage during mount. */ 3953e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) { 39549b2ff357SJan Kara err = ext4_enable_quotas(sb); 39559b2ff357SJan Kara if (err) 39569b2ff357SJan Kara goto failed_mount8; 39579b2ff357SJan Kara } 39589b2ff357SJan Kara #endif /* CONFIG_QUOTA */ 39599b2ff357SJan Kara 3960617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 3961617ba13bSMingming Cao ext4_orphan_cleanup(sb, es); 3962617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 39630390131bSFrank Mayhar if (needs_recovery) { 3964b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "recovery complete"); 3965617ba13bSMingming Cao ext4_mark_recovery_complete(sb, es); 39660390131bSFrank Mayhar } 39670390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal) { 39680390131bSFrank Mayhar if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 39690390131bSFrank Mayhar descr = " journalled data mode"; 39700390131bSFrank Mayhar else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 39710390131bSFrank Mayhar descr = " ordered data mode"; 39720390131bSFrank Mayhar else 39730390131bSFrank Mayhar descr = " writeback data mode"; 39740390131bSFrank Mayhar } else 39750390131bSFrank Mayhar descr = "out journal"; 39760390131bSFrank Mayhar 397779add3a3SLukas Czerner if (test_opt(sb, DISCARD)) { 397879add3a3SLukas Czerner struct request_queue *q = bdev_get_queue(sb->s_bdev); 397979add3a3SLukas Czerner if (!blk_queue_discard(q)) 398079add3a3SLukas Czerner ext4_msg(sb, KERN_WARNING, 398179add3a3SLukas Czerner "mounting with \"discard\" option, but " 398279add3a3SLukas Czerner "the device does not support discard"); 398379add3a3SLukas Czerner } 398479add3a3SLukas Czerner 3985e294a537STheodore Ts'o if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 3986d4c402d9SCurt Wohlgemuth ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 39878b67f04aSTheodore Ts'o "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts, 39888b67f04aSTheodore Ts'o *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 3989ac27a0ecSDave Kleikamp 399066e61a9eSTheodore Ts'o if (es->s_error_count) 399166e61a9eSTheodore Ts'o mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 3992ac27a0ecSDave Kleikamp 3993efbed4dcSTheodore Ts'o /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 3994efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 3995efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 3996efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 3997efbed4dcSTheodore Ts'o 3998d4c402d9SCurt Wohlgemuth kfree(orig_data); 3999ac27a0ecSDave Kleikamp return 0; 4000ac27a0ecSDave Kleikamp 4001617ba13bSMingming Cao cantfind_ext4: 4002ac27a0ecSDave Kleikamp if (!silent) 4003b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 4004ac27a0ecSDave Kleikamp goto failed_mount; 4005ac27a0ecSDave Kleikamp 400672ba7450STheodore Ts'o #ifdef CONFIG_QUOTA 400772ba7450STheodore Ts'o failed_mount8: 4008ebd173beSTheodore Ts'o ext4_unregister_sysfs(sb); 400972ba7450STheodore Ts'o #endif 4010dcf2d804STao Ma failed_mount7: 4011dcf2d804STao Ma ext4_unregister_li_request(sb); 4012dcf2d804STao Ma failed_mount6: 4013f9ae9cf5STheodore Ts'o ext4_mb_release(sb); 4014d5e03cbbSTheodore Ts'o if (sbi->s_flex_groups) 4015b93b41d4SAl Viro kvfree(sbi->s_flex_groups); 4016d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_freeclusters_counter); 4017d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_freeinodes_counter); 4018d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_dirs_counter); 4019d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 402000764937SAzat Khuzhin failed_mount5: 4021f9ae9cf5STheodore Ts'o ext4_ext_release(sb); 4022f9ae9cf5STheodore Ts'o ext4_release_system_zone(sb); 4023f9ae9cf5STheodore Ts'o failed_mount4a: 402494bf608aSAl Viro dput(sb->s_root); 402532a9bb57SManish Katiyar sb->s_root = NULL; 402694bf608aSAl Viro failed_mount4: 4027b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "mount failed"); 40282e8fa54eSJan Kara if (EXT4_SB(sb)->rsv_conversion_wq) 40292e8fa54eSJan Kara destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 40304c0425ffSMingming Cao failed_mount_wq: 403182939d79SJan Kara if (sbi->s_mb_cache) { 403282939d79SJan Kara ext4_xattr_destroy_cache(sbi->s_mb_cache); 403382939d79SJan Kara sbi->s_mb_cache = NULL; 403482939d79SJan Kara } 40350390131bSFrank Mayhar if (sbi->s_journal) { 4036dab291afSMingming Cao jbd2_journal_destroy(sbi->s_journal); 403747b4a50bSJan Kara sbi->s_journal = NULL; 40380390131bSFrank Mayhar } 403950460fe8SDarrick J. Wong failed_mount3a: 4040d3922a77SZheng Liu ext4_es_unregister_shrinker(sbi); 4041eb68d0e2SZheng Liu failed_mount3: 40429105bb14SAl Viro del_timer_sync(&sbi->s_err_report); 4043c5e06d10SJohann Lombardi if (sbi->s_mmp_tsk) 4044c5e06d10SJohann Lombardi kthread_stop(sbi->s_mmp_tsk); 4045ac27a0ecSDave Kleikamp failed_mount2: 4046ac27a0ecSDave Kleikamp for (i = 0; i < db_count; i++) 4047ac27a0ecSDave Kleikamp brelse(sbi->s_group_desc[i]); 4048b93b41d4SAl Viro kvfree(sbi->s_group_desc); 4049ac27a0ecSDave Kleikamp failed_mount: 40500441984aSDarrick J. Wong if (sbi->s_chksum_driver) 40510441984aSDarrick J. Wong crypto_free_shash(sbi->s_chksum_driver); 4052ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4053a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 4054ac27a0ecSDave Kleikamp kfree(sbi->s_qf_names[i]); 4055ac27a0ecSDave Kleikamp #endif 4056617ba13bSMingming Cao ext4_blkdev_remove(sbi); 4057ac27a0ecSDave Kleikamp brelse(bh); 4058ac27a0ecSDave Kleikamp out_fail: 4059ac27a0ecSDave Kleikamp sb->s_fs_info = NULL; 4060f6830165SManish Katiyar kfree(sbi->s_blockgroup_lock); 4061ac27a0ecSDave Kleikamp kfree(sbi); 4062dcc7dae3SCyrill Gorcunov out_free_orig: 4063d4c402d9SCurt Wohlgemuth kfree(orig_data); 406407aa2ea1SLukas Czerner return err ? err : ret; 4065ac27a0ecSDave Kleikamp } 4066ac27a0ecSDave Kleikamp 4067ac27a0ecSDave Kleikamp /* 4068ac27a0ecSDave Kleikamp * Setup any per-fs journal parameters now. We'll do this both on 4069ac27a0ecSDave Kleikamp * initial mount, once the journal has been initialised but before we've 4070ac27a0ecSDave Kleikamp * done any recovery; and again on any subsequent remount. 4071ac27a0ecSDave Kleikamp */ 4072617ba13bSMingming Cao static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 4073ac27a0ecSDave Kleikamp { 4074617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 4075ac27a0ecSDave Kleikamp 4076ac27a0ecSDave Kleikamp journal->j_commit_interval = sbi->s_commit_interval; 407730773840STheodore Ts'o journal->j_min_batch_time = sbi->s_min_batch_time; 407830773840STheodore Ts'o journal->j_max_batch_time = sbi->s_max_batch_time; 4079ac27a0ecSDave Kleikamp 4080a931da6aSTheodore Ts'o write_lock(&journal->j_state_lock); 4081ac27a0ecSDave Kleikamp if (test_opt(sb, BARRIER)) 4082dab291afSMingming Cao journal->j_flags |= JBD2_BARRIER; 4083ac27a0ecSDave Kleikamp else 4084dab291afSMingming Cao journal->j_flags &= ~JBD2_BARRIER; 40855bf5683aSHidehiro Kawai if (test_opt(sb, DATA_ERR_ABORT)) 40865bf5683aSHidehiro Kawai journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 40875bf5683aSHidehiro Kawai else 40885bf5683aSHidehiro Kawai journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 4089a931da6aSTheodore Ts'o write_unlock(&journal->j_state_lock); 4090ac27a0ecSDave Kleikamp } 4091ac27a0ecSDave Kleikamp 4092617ba13bSMingming Cao static journal_t *ext4_get_journal(struct super_block *sb, 4093ac27a0ecSDave Kleikamp unsigned int journal_inum) 4094ac27a0ecSDave Kleikamp { 4095ac27a0ecSDave Kleikamp struct inode *journal_inode; 4096ac27a0ecSDave Kleikamp journal_t *journal; 4097ac27a0ecSDave Kleikamp 4098e2b911c5SDarrick J. Wong BUG_ON(!ext4_has_feature_journal(sb)); 40990390131bSFrank Mayhar 4100ac27a0ecSDave Kleikamp /* First, test for the existence of a valid inode on disk. Bad 4101ac27a0ecSDave Kleikamp * things happen if we iget() an unused inode, as the subsequent 4102ac27a0ecSDave Kleikamp * iput() will try to delete it. */ 4103ac27a0ecSDave Kleikamp 41041d1fe1eeSDavid Howells journal_inode = ext4_iget(sb, journal_inum); 41051d1fe1eeSDavid Howells if (IS_ERR(journal_inode)) { 4106b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "no journal found"); 4107ac27a0ecSDave Kleikamp return NULL; 4108ac27a0ecSDave Kleikamp } 4109ac27a0ecSDave Kleikamp if (!journal_inode->i_nlink) { 4110ac27a0ecSDave Kleikamp make_bad_inode(journal_inode); 4111ac27a0ecSDave Kleikamp iput(journal_inode); 4112b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 4113ac27a0ecSDave Kleikamp return NULL; 4114ac27a0ecSDave Kleikamp } 4115ac27a0ecSDave Kleikamp 4116e5f8eab8STheodore Ts'o jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 4117ac27a0ecSDave Kleikamp journal_inode, journal_inode->i_size); 41181d1fe1eeSDavid Howells if (!S_ISREG(journal_inode->i_mode)) { 4119b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "invalid journal inode"); 4120ac27a0ecSDave Kleikamp iput(journal_inode); 4121ac27a0ecSDave Kleikamp return NULL; 4122ac27a0ecSDave Kleikamp } 4123ac27a0ecSDave Kleikamp 4124dab291afSMingming Cao journal = jbd2_journal_init_inode(journal_inode); 4125ac27a0ecSDave Kleikamp if (!journal) { 4126b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 4127ac27a0ecSDave Kleikamp iput(journal_inode); 4128ac27a0ecSDave Kleikamp return NULL; 4129ac27a0ecSDave Kleikamp } 4130ac27a0ecSDave Kleikamp journal->j_private = sb; 4131617ba13bSMingming Cao ext4_init_journal_params(sb, journal); 4132ac27a0ecSDave Kleikamp return journal; 4133ac27a0ecSDave Kleikamp } 4134ac27a0ecSDave Kleikamp 4135617ba13bSMingming Cao static journal_t *ext4_get_dev_journal(struct super_block *sb, 4136ac27a0ecSDave Kleikamp dev_t j_dev) 4137ac27a0ecSDave Kleikamp { 4138ac27a0ecSDave Kleikamp struct buffer_head *bh; 4139ac27a0ecSDave Kleikamp journal_t *journal; 4140617ba13bSMingming Cao ext4_fsblk_t start; 4141617ba13bSMingming Cao ext4_fsblk_t len; 4142ac27a0ecSDave Kleikamp int hblock, blocksize; 4143617ba13bSMingming Cao ext4_fsblk_t sb_block; 4144ac27a0ecSDave Kleikamp unsigned long offset; 4145617ba13bSMingming Cao struct ext4_super_block *es; 4146ac27a0ecSDave Kleikamp struct block_device *bdev; 4147ac27a0ecSDave Kleikamp 4148e2b911c5SDarrick J. Wong BUG_ON(!ext4_has_feature_journal(sb)); 41490390131bSFrank Mayhar 4150b31e1552SEric Sandeen bdev = ext4_blkdev_get(j_dev, sb); 4151ac27a0ecSDave Kleikamp if (bdev == NULL) 4152ac27a0ecSDave Kleikamp return NULL; 4153ac27a0ecSDave Kleikamp 4154ac27a0ecSDave Kleikamp blocksize = sb->s_blocksize; 4155e1defc4fSMartin K. Petersen hblock = bdev_logical_block_size(bdev); 4156ac27a0ecSDave Kleikamp if (blocksize < hblock) { 4157b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4158b31e1552SEric Sandeen "blocksize too small for journal device"); 4159ac27a0ecSDave Kleikamp goto out_bdev; 4160ac27a0ecSDave Kleikamp } 4161ac27a0ecSDave Kleikamp 4162617ba13bSMingming Cao sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 4163617ba13bSMingming Cao offset = EXT4_MIN_BLOCK_SIZE % blocksize; 4164ac27a0ecSDave Kleikamp set_blocksize(bdev, blocksize); 4165ac27a0ecSDave Kleikamp if (!(bh = __bread(bdev, sb_block, blocksize))) { 4166b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 4167b31e1552SEric Sandeen "external journal"); 4168ac27a0ecSDave Kleikamp goto out_bdev; 4169ac27a0ecSDave Kleikamp } 4170ac27a0ecSDave Kleikamp 41712716b802STheodore Ts'o es = (struct ext4_super_block *) (bh->b_data + offset); 4172617ba13bSMingming Cao if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 4173ac27a0ecSDave Kleikamp !(le32_to_cpu(es->s_feature_incompat) & 4174617ba13bSMingming Cao EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 4175b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "external journal has " 4176b31e1552SEric Sandeen "bad superblock"); 4177ac27a0ecSDave Kleikamp brelse(bh); 4178ac27a0ecSDave Kleikamp goto out_bdev; 4179ac27a0ecSDave Kleikamp } 4180ac27a0ecSDave Kleikamp 4181df4763beSDarrick J. Wong if ((le32_to_cpu(es->s_feature_ro_compat) & 4182df4763beSDarrick J. Wong EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 4183df4763beSDarrick J. Wong es->s_checksum != ext4_superblock_csum(sb, es)) { 4184df4763beSDarrick J. Wong ext4_msg(sb, KERN_ERR, "external journal has " 4185df4763beSDarrick J. Wong "corrupt superblock"); 4186df4763beSDarrick J. Wong brelse(bh); 4187df4763beSDarrick J. Wong goto out_bdev; 4188df4763beSDarrick J. Wong } 4189df4763beSDarrick J. Wong 4190617ba13bSMingming Cao if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 4191b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 4192ac27a0ecSDave Kleikamp brelse(bh); 4193ac27a0ecSDave Kleikamp goto out_bdev; 4194ac27a0ecSDave Kleikamp } 4195ac27a0ecSDave Kleikamp 4196bd81d8eeSLaurent Vivier len = ext4_blocks_count(es); 4197ac27a0ecSDave Kleikamp start = sb_block + 1; 4198ac27a0ecSDave Kleikamp brelse(bh); /* we're done with the superblock */ 4199ac27a0ecSDave Kleikamp 4200dab291afSMingming Cao journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 4201ac27a0ecSDave Kleikamp start, len, blocksize); 4202ac27a0ecSDave Kleikamp if (!journal) { 4203b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "failed to create device journal"); 4204ac27a0ecSDave Kleikamp goto out_bdev; 4205ac27a0ecSDave Kleikamp } 4206ac27a0ecSDave Kleikamp journal->j_private = sb; 42079f203507STheodore Ts'o ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer); 4208ac27a0ecSDave Kleikamp wait_on_buffer(journal->j_sb_buffer); 4209ac27a0ecSDave Kleikamp if (!buffer_uptodate(journal->j_sb_buffer)) { 4210b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 4211ac27a0ecSDave Kleikamp goto out_journal; 4212ac27a0ecSDave Kleikamp } 4213ac27a0ecSDave Kleikamp if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 4214b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "External journal has more than one " 4215b31e1552SEric Sandeen "user (unsupported) - %d", 4216ac27a0ecSDave Kleikamp be32_to_cpu(journal->j_superblock->s_nr_users)); 4217ac27a0ecSDave Kleikamp goto out_journal; 4218ac27a0ecSDave Kleikamp } 4219617ba13bSMingming Cao EXT4_SB(sb)->journal_bdev = bdev; 4220617ba13bSMingming Cao ext4_init_journal_params(sb, journal); 4221ac27a0ecSDave Kleikamp return journal; 42220b8e58a1SAndreas Dilger 4223ac27a0ecSDave Kleikamp out_journal: 4224dab291afSMingming Cao jbd2_journal_destroy(journal); 4225ac27a0ecSDave Kleikamp out_bdev: 4226617ba13bSMingming Cao ext4_blkdev_put(bdev); 4227ac27a0ecSDave Kleikamp return NULL; 4228ac27a0ecSDave Kleikamp } 4229ac27a0ecSDave Kleikamp 4230617ba13bSMingming Cao static int ext4_load_journal(struct super_block *sb, 4231617ba13bSMingming Cao struct ext4_super_block *es, 4232ac27a0ecSDave Kleikamp unsigned long journal_devnum) 4233ac27a0ecSDave Kleikamp { 4234ac27a0ecSDave Kleikamp journal_t *journal; 4235ac27a0ecSDave Kleikamp unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 4236ac27a0ecSDave Kleikamp dev_t journal_dev; 4237ac27a0ecSDave Kleikamp int err = 0; 4238ac27a0ecSDave Kleikamp int really_read_only; 4239ac27a0ecSDave Kleikamp 4240e2b911c5SDarrick J. Wong BUG_ON(!ext4_has_feature_journal(sb)); 42410390131bSFrank Mayhar 4242ac27a0ecSDave Kleikamp if (journal_devnum && 4243ac27a0ecSDave Kleikamp journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4244b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "external journal device major/minor " 4245b31e1552SEric Sandeen "numbers have changed"); 4246ac27a0ecSDave Kleikamp journal_dev = new_decode_dev(journal_devnum); 4247ac27a0ecSDave Kleikamp } else 4248ac27a0ecSDave Kleikamp journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 4249ac27a0ecSDave Kleikamp 4250ac27a0ecSDave Kleikamp really_read_only = bdev_read_only(sb->s_bdev); 4251ac27a0ecSDave Kleikamp 4252ac27a0ecSDave Kleikamp /* 4253ac27a0ecSDave Kleikamp * Are we loading a blank journal or performing recovery after a 4254ac27a0ecSDave Kleikamp * crash? For recovery, we need to check in advance whether we 4255ac27a0ecSDave Kleikamp * can get read-write access to the device. 4256ac27a0ecSDave Kleikamp */ 4257e2b911c5SDarrick J. Wong if (ext4_has_feature_journal_needs_recovery(sb)) { 4258ac27a0ecSDave Kleikamp if (sb->s_flags & MS_RDONLY) { 4259b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "INFO: recovery " 4260b31e1552SEric Sandeen "required on readonly filesystem"); 4261ac27a0ecSDave Kleikamp if (really_read_only) { 4262b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "write access " 4263b31e1552SEric Sandeen "unavailable, cannot proceed"); 4264ac27a0ecSDave Kleikamp return -EROFS; 4265ac27a0ecSDave Kleikamp } 4266b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "write access will " 4267b31e1552SEric Sandeen "be enabled during recovery"); 4268ac27a0ecSDave Kleikamp } 4269ac27a0ecSDave Kleikamp } 4270ac27a0ecSDave Kleikamp 4271ac27a0ecSDave Kleikamp if (journal_inum && journal_dev) { 4272b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "filesystem has both journal " 4273b31e1552SEric Sandeen "and inode journals!"); 4274ac27a0ecSDave Kleikamp return -EINVAL; 4275ac27a0ecSDave Kleikamp } 4276ac27a0ecSDave Kleikamp 4277ac27a0ecSDave Kleikamp if (journal_inum) { 4278617ba13bSMingming Cao if (!(journal = ext4_get_journal(sb, journal_inum))) 4279ac27a0ecSDave Kleikamp return -EINVAL; 4280ac27a0ecSDave Kleikamp } else { 4281617ba13bSMingming Cao if (!(journal = ext4_get_dev_journal(sb, journal_dev))) 4282ac27a0ecSDave Kleikamp return -EINVAL; 4283ac27a0ecSDave Kleikamp } 4284ac27a0ecSDave Kleikamp 428590576c0bSTheodore Ts'o if (!(journal->j_flags & JBD2_BARRIER)) 4286b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "barriers disabled"); 42874776004fSTheodore Ts'o 4288e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal_needs_recovery(sb)) 4289dab291afSMingming Cao err = jbd2_journal_wipe(journal, !really_read_only); 42901c13d5c0STheodore Ts'o if (!err) { 42911c13d5c0STheodore Ts'o char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 42921c13d5c0STheodore Ts'o if (save) 42931c13d5c0STheodore Ts'o memcpy(save, ((char *) es) + 42941c13d5c0STheodore Ts'o EXT4_S_ERR_START, EXT4_S_ERR_LEN); 4295dab291afSMingming Cao err = jbd2_journal_load(journal); 42961c13d5c0STheodore Ts'o if (save) 42971c13d5c0STheodore Ts'o memcpy(((char *) es) + EXT4_S_ERR_START, 42981c13d5c0STheodore Ts'o save, EXT4_S_ERR_LEN); 42991c13d5c0STheodore Ts'o kfree(save); 43001c13d5c0STheodore Ts'o } 4301ac27a0ecSDave Kleikamp 4302ac27a0ecSDave Kleikamp if (err) { 4303b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "error loading journal"); 4304dab291afSMingming Cao jbd2_journal_destroy(journal); 4305ac27a0ecSDave Kleikamp return err; 4306ac27a0ecSDave Kleikamp } 4307ac27a0ecSDave Kleikamp 4308617ba13bSMingming Cao EXT4_SB(sb)->s_journal = journal; 4309617ba13bSMingming Cao ext4_clear_journal_err(sb, es); 4310ac27a0ecSDave Kleikamp 4311c41303ceSMaciej Żenczykowski if (!really_read_only && journal_devnum && 4312ac27a0ecSDave Kleikamp journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4313ac27a0ecSDave Kleikamp es->s_journal_dev = cpu_to_le32(journal_devnum); 4314ac27a0ecSDave Kleikamp 4315ac27a0ecSDave Kleikamp /* Make sure we flush the recovery flag to disk. */ 4316e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 4317ac27a0ecSDave Kleikamp } 4318ac27a0ecSDave Kleikamp 4319ac27a0ecSDave Kleikamp return 0; 4320ac27a0ecSDave Kleikamp } 4321ac27a0ecSDave Kleikamp 4322e2d67052STheodore Ts'o static int ext4_commit_super(struct super_block *sb, int sync) 4323ac27a0ecSDave Kleikamp { 4324e2d67052STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 4325617ba13bSMingming Cao struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 4326c4be0c1dSTakashi Sato int error = 0; 4327ac27a0ecSDave Kleikamp 4328bdfe0cbdSTheodore Ts'o if (!sbh || block_device_ejected(sb)) 4329c4be0c1dSTakashi Sato return error; 4330914258bfSTheodore Ts'o if (buffer_write_io_error(sbh)) { 4331914258bfSTheodore Ts'o /* 4332914258bfSTheodore Ts'o * Oh, dear. A previous attempt to write the 4333914258bfSTheodore Ts'o * superblock failed. This could happen because the 4334914258bfSTheodore Ts'o * USB device was yanked out. Or it could happen to 4335914258bfSTheodore Ts'o * be a transient write error and maybe the block will 4336914258bfSTheodore Ts'o * be remapped. Nothing we can do but to retry the 4337914258bfSTheodore Ts'o * write and hope for the best. 4338914258bfSTheodore Ts'o */ 4339b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "previous I/O error to " 4340b31e1552SEric Sandeen "superblock detected"); 4341914258bfSTheodore Ts'o clear_buffer_write_io_error(sbh); 4342914258bfSTheodore Ts'o set_buffer_uptodate(sbh); 4343914258bfSTheodore Ts'o } 434471290b36STheodore Ts'o /* 434571290b36STheodore Ts'o * If the file system is mounted read-only, don't update the 434671290b36STheodore Ts'o * superblock write time. This avoids updating the superblock 434771290b36STheodore Ts'o * write time when we are mounting the root file system 434871290b36STheodore Ts'o * read/only but we need to replay the journal; at that point, 434971290b36STheodore Ts'o * for people who are east of GMT and who make their clock 435071290b36STheodore Ts'o * tick in localtime for Windows bug-for-bug compatibility, 435171290b36STheodore Ts'o * the clock is set in the future, and this will cause e2fsck 435271290b36STheodore Ts'o * to complain and force a full file system check. 435371290b36STheodore Ts'o */ 435471290b36STheodore Ts'o if (!(sb->s_flags & MS_RDONLY)) 4355ac27a0ecSDave Kleikamp es->s_wtime = cpu_to_le32(get_seconds()); 4356f613dfcbSTheodore Ts'o if (sb->s_bdev->bd_part) 4357afc32f7eSTheodore Ts'o es->s_kbytes_written = 4358afc32f7eSTheodore Ts'o cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 4359afc32f7eSTheodore Ts'o ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 4360afc32f7eSTheodore Ts'o EXT4_SB(sb)->s_sectors_written_start) >> 1)); 4361f613dfcbSTheodore Ts'o else 4362f613dfcbSTheodore Ts'o es->s_kbytes_written = 4363f613dfcbSTheodore Ts'o cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 4364d5e03cbbSTheodore Ts'o if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter)) 436557042651STheodore Ts'o ext4_free_blocks_count_set(es, 436657042651STheodore Ts'o EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 436757042651STheodore Ts'o &EXT4_SB(sb)->s_freeclusters_counter))); 4368d5e03cbbSTheodore Ts'o if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter)) 43697f93cff9STheodore Ts'o es->s_free_inodes_count = 43707f93cff9STheodore Ts'o cpu_to_le32(percpu_counter_sum_positive( 43715d1b1b3fSAneesh Kumar K.V &EXT4_SB(sb)->s_freeinodes_counter)); 4372ac27a0ecSDave Kleikamp BUFFER_TRACE(sbh, "marking dirty"); 437306db49e6STheodore Ts'o ext4_superblock_csum_set(sb); 4374ac27a0ecSDave Kleikamp mark_buffer_dirty(sbh); 4375914258bfSTheodore Ts'o if (sync) { 4376564bc402SDaeho Jeong error = __sync_dirty_buffer(sbh, 4377564bc402SDaeho Jeong test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC); 4378c4be0c1dSTakashi Sato if (error) 4379c4be0c1dSTakashi Sato return error; 4380c4be0c1dSTakashi Sato 4381c4be0c1dSTakashi Sato error = buffer_write_io_error(sbh); 4382c4be0c1dSTakashi Sato if (error) { 4383b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "I/O error while writing " 4384b31e1552SEric Sandeen "superblock"); 4385914258bfSTheodore Ts'o clear_buffer_write_io_error(sbh); 4386914258bfSTheodore Ts'o set_buffer_uptodate(sbh); 4387914258bfSTheodore Ts'o } 4388914258bfSTheodore Ts'o } 4389c4be0c1dSTakashi Sato return error; 4390ac27a0ecSDave Kleikamp } 4391ac27a0ecSDave Kleikamp 4392ac27a0ecSDave Kleikamp /* 4393ac27a0ecSDave Kleikamp * Have we just finished recovery? If so, and if we are mounting (or 4394ac27a0ecSDave Kleikamp * remounting) the filesystem readonly, then we will end up with a 4395ac27a0ecSDave Kleikamp * consistent fs on disk. Record that fact. 4396ac27a0ecSDave Kleikamp */ 4397617ba13bSMingming Cao static void ext4_mark_recovery_complete(struct super_block *sb, 4398617ba13bSMingming Cao struct ext4_super_block *es) 4399ac27a0ecSDave Kleikamp { 4400617ba13bSMingming Cao journal_t *journal = EXT4_SB(sb)->s_journal; 4401ac27a0ecSDave Kleikamp 4402e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal(sb)) { 44030390131bSFrank Mayhar BUG_ON(journal != NULL); 44040390131bSFrank Mayhar return; 44050390131bSFrank Mayhar } 4406dab291afSMingming Cao jbd2_journal_lock_updates(journal); 44077ffe1ea8SHidehiro Kawai if (jbd2_journal_flush(journal) < 0) 44087ffe1ea8SHidehiro Kawai goto out; 44097ffe1ea8SHidehiro Kawai 4410e2b911c5SDarrick J. Wong if (ext4_has_feature_journal_needs_recovery(sb) && 4411ac27a0ecSDave Kleikamp sb->s_flags & MS_RDONLY) { 4412e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 4413e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 4414ac27a0ecSDave Kleikamp } 44157ffe1ea8SHidehiro Kawai 44167ffe1ea8SHidehiro Kawai out: 4417dab291afSMingming Cao jbd2_journal_unlock_updates(journal); 4418ac27a0ecSDave Kleikamp } 4419ac27a0ecSDave Kleikamp 4420ac27a0ecSDave Kleikamp /* 4421ac27a0ecSDave Kleikamp * If we are mounting (or read-write remounting) a filesystem whose journal 4422ac27a0ecSDave Kleikamp * has recorded an error from a previous lifetime, move that error to the 4423ac27a0ecSDave Kleikamp * main filesystem now. 4424ac27a0ecSDave Kleikamp */ 4425617ba13bSMingming Cao static void ext4_clear_journal_err(struct super_block *sb, 4426617ba13bSMingming Cao struct ext4_super_block *es) 4427ac27a0ecSDave Kleikamp { 4428ac27a0ecSDave Kleikamp journal_t *journal; 4429ac27a0ecSDave Kleikamp int j_errno; 4430ac27a0ecSDave Kleikamp const char *errstr; 4431ac27a0ecSDave Kleikamp 4432e2b911c5SDarrick J. Wong BUG_ON(!ext4_has_feature_journal(sb)); 44330390131bSFrank Mayhar 4434617ba13bSMingming Cao journal = EXT4_SB(sb)->s_journal; 4435ac27a0ecSDave Kleikamp 4436ac27a0ecSDave Kleikamp /* 4437ac27a0ecSDave Kleikamp * Now check for any error status which may have been recorded in the 4438617ba13bSMingming Cao * journal by a prior ext4_error() or ext4_abort() 4439ac27a0ecSDave Kleikamp */ 4440ac27a0ecSDave Kleikamp 4441dab291afSMingming Cao j_errno = jbd2_journal_errno(journal); 4442ac27a0ecSDave Kleikamp if (j_errno) { 4443ac27a0ecSDave Kleikamp char nbuf[16]; 4444ac27a0ecSDave Kleikamp 4445617ba13bSMingming Cao errstr = ext4_decode_error(sb, j_errno, nbuf); 444612062dddSEric Sandeen ext4_warning(sb, "Filesystem error recorded " 4447ac27a0ecSDave Kleikamp "from previous mount: %s", errstr); 444812062dddSEric Sandeen ext4_warning(sb, "Marking fs in need of filesystem check."); 4449ac27a0ecSDave Kleikamp 4450617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 4451617ba13bSMingming Cao es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 4452e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 4453ac27a0ecSDave Kleikamp 4454dab291afSMingming Cao jbd2_journal_clear_err(journal); 4455d796c52eSTheodore Ts'o jbd2_journal_update_sb_errno(journal); 4456ac27a0ecSDave Kleikamp } 4457ac27a0ecSDave Kleikamp } 4458ac27a0ecSDave Kleikamp 4459ac27a0ecSDave Kleikamp /* 4460ac27a0ecSDave Kleikamp * Force the running and committing transactions to commit, 4461ac27a0ecSDave Kleikamp * and wait on the commit. 4462ac27a0ecSDave Kleikamp */ 4463617ba13bSMingming Cao int ext4_force_commit(struct super_block *sb) 4464ac27a0ecSDave Kleikamp { 4465ac27a0ecSDave Kleikamp journal_t *journal; 4466ac27a0ecSDave Kleikamp 4467ac27a0ecSDave Kleikamp if (sb->s_flags & MS_RDONLY) 4468ac27a0ecSDave Kleikamp return 0; 4469ac27a0ecSDave Kleikamp 4470617ba13bSMingming Cao journal = EXT4_SB(sb)->s_journal; 4471b1deefc9SGuo Chao return ext4_journal_force_commit(journal); 4472ac27a0ecSDave Kleikamp } 4473ac27a0ecSDave Kleikamp 4474617ba13bSMingming Cao static int ext4_sync_fs(struct super_block *sb, int wait) 4475ac27a0ecSDave Kleikamp { 447614ce0cb4STheodore Ts'o int ret = 0; 44779eddacf9SJan Kara tid_t target; 447806a407f1SDmitry Monakhov bool needs_barrier = false; 44798d5d02e6SMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 4480ac27a0ecSDave Kleikamp 44819bffad1eSTheodore Ts'o trace_ext4_sync_fs(sb, wait); 44822e8fa54eSJan Kara flush_workqueue(sbi->rsv_conversion_wq); 4483a1177825SJan Kara /* 4484a1177825SJan Kara * Writeback quota in non-journalled quota case - journalled quota has 4485a1177825SJan Kara * no dirty dquots 4486a1177825SJan Kara */ 4487a1177825SJan Kara dquot_writeback_dquots(sb, -1); 448806a407f1SDmitry Monakhov /* 448906a407f1SDmitry Monakhov * Data writeback is possible w/o journal transaction, so barrier must 449006a407f1SDmitry Monakhov * being sent at the end of the function. But we can skip it if 449106a407f1SDmitry Monakhov * transaction_commit will do it for us. 449206a407f1SDmitry Monakhov */ 4493bda32530STheodore Ts'o if (sbi->s_journal) { 449406a407f1SDmitry Monakhov target = jbd2_get_latest_transaction(sbi->s_journal); 449506a407f1SDmitry Monakhov if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 449606a407f1SDmitry Monakhov !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 449706a407f1SDmitry Monakhov needs_barrier = true; 449806a407f1SDmitry Monakhov 44998d5d02e6SMingming Cao if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 4500ac27a0ecSDave Kleikamp if (wait) 4501bda32530STheodore Ts'o ret = jbd2_log_wait_commit(sbi->s_journal, 4502bda32530STheodore Ts'o target); 45030390131bSFrank Mayhar } 4504bda32530STheodore Ts'o } else if (wait && test_opt(sb, BARRIER)) 4505bda32530STheodore Ts'o needs_barrier = true; 450606a407f1SDmitry Monakhov if (needs_barrier) { 450706a407f1SDmitry Monakhov int err; 450806a407f1SDmitry Monakhov err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); 450906a407f1SDmitry Monakhov if (!ret) 451006a407f1SDmitry Monakhov ret = err; 451106a407f1SDmitry Monakhov } 451206a407f1SDmitry Monakhov 451306a407f1SDmitry Monakhov return ret; 451406a407f1SDmitry Monakhov } 451506a407f1SDmitry Monakhov 4516ac27a0ecSDave Kleikamp /* 4517ac27a0ecSDave Kleikamp * LVM calls this function before a (read-only) snapshot is created. This 4518ac27a0ecSDave Kleikamp * gives us a chance to flush the journal completely and mark the fs clean. 4519be4f27d3SYongqiang Yang * 4520be4f27d3SYongqiang Yang * Note that only this function cannot bring a filesystem to be in a clean 45218e8ad8a5SJan Kara * state independently. It relies on upper layer to stop all data & metadata 45228e8ad8a5SJan Kara * modifications. 4523ac27a0ecSDave Kleikamp */ 4524c4be0c1dSTakashi Sato static int ext4_freeze(struct super_block *sb) 4525ac27a0ecSDave Kleikamp { 4526c4be0c1dSTakashi Sato int error = 0; 4527c4be0c1dSTakashi Sato journal_t *journal; 4528ac27a0ecSDave Kleikamp 45299ca92389STheodore Ts'o if (sb->s_flags & MS_RDONLY) 45309ca92389STheodore Ts'o return 0; 45319ca92389STheodore Ts'o 4532c4be0c1dSTakashi Sato journal = EXT4_SB(sb)->s_journal; 4533ac27a0ecSDave Kleikamp 4534bb044576STheodore Ts'o if (journal) { 4535ac27a0ecSDave Kleikamp /* Now we set up the journal barrier. */ 4536dab291afSMingming Cao jbd2_journal_lock_updates(journal); 45377ffe1ea8SHidehiro Kawai 45387ffe1ea8SHidehiro Kawai /* 4539bb044576STheodore Ts'o * Don't clear the needs_recovery flag if we failed to 4540bb044576STheodore Ts'o * flush the journal. 45417ffe1ea8SHidehiro Kawai */ 4542c4be0c1dSTakashi Sato error = jbd2_journal_flush(journal); 45436b0310fbSEric Sandeen if (error < 0) 45446b0310fbSEric Sandeen goto out; 4545ac27a0ecSDave Kleikamp 4546ac27a0ecSDave Kleikamp /* Journal blocked and flushed, clear needs_recovery flag. */ 4547e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 4548c642dc9eSEric Sandeen } 4549c642dc9eSEric Sandeen 4550e2d67052STheodore Ts'o error = ext4_commit_super(sb, 1); 45516b0310fbSEric Sandeen out: 4552bb044576STheodore Ts'o if (journal) 45538e8ad8a5SJan Kara /* we rely on upper layer to stop further updates */ 4554bb044576STheodore Ts'o jbd2_journal_unlock_updates(journal); 45556b0310fbSEric Sandeen return error; 4556ac27a0ecSDave Kleikamp } 4557ac27a0ecSDave Kleikamp 4558ac27a0ecSDave Kleikamp /* 4559ac27a0ecSDave Kleikamp * Called by LVM after the snapshot is done. We need to reset the RECOVER 4560ac27a0ecSDave Kleikamp * flag here, even though the filesystem is not technically dirty yet. 4561ac27a0ecSDave Kleikamp */ 4562c4be0c1dSTakashi Sato static int ext4_unfreeze(struct super_block *sb) 4563ac27a0ecSDave Kleikamp { 45649ca92389STheodore Ts'o if (sb->s_flags & MS_RDONLY) 45659ca92389STheodore Ts'o return 0; 45669ca92389STheodore Ts'o 4567c642dc9eSEric Sandeen if (EXT4_SB(sb)->s_journal) { 45689ca92389STheodore Ts'o /* Reset the needs_recovery flag before the fs is unlocked. */ 4569e2b911c5SDarrick J. Wong ext4_set_feature_journal_needs_recovery(sb); 4570c642dc9eSEric Sandeen } 4571c642dc9eSEric Sandeen 4572e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 4573c4be0c1dSTakashi Sato return 0; 4574ac27a0ecSDave Kleikamp } 4575ac27a0ecSDave Kleikamp 4576673c6100STheodore Ts'o /* 4577673c6100STheodore Ts'o * Structure to save mount options for ext4_remount's benefit 4578673c6100STheodore Ts'o */ 4579673c6100STheodore Ts'o struct ext4_mount_options { 4580673c6100STheodore Ts'o unsigned long s_mount_opt; 4581a2595b8aSTheodore Ts'o unsigned long s_mount_opt2; 458208cefc7aSEric W. Biederman kuid_t s_resuid; 458308cefc7aSEric W. Biederman kgid_t s_resgid; 4584673c6100STheodore Ts'o unsigned long s_commit_interval; 4585673c6100STheodore Ts'o u32 s_min_batch_time, s_max_batch_time; 4586673c6100STheodore Ts'o #ifdef CONFIG_QUOTA 4587673c6100STheodore Ts'o int s_jquota_fmt; 4588a2d4a646SJan Kara char *s_qf_names[EXT4_MAXQUOTAS]; 4589673c6100STheodore Ts'o #endif 4590673c6100STheodore Ts'o }; 4591673c6100STheodore Ts'o 4592617ba13bSMingming Cao static int ext4_remount(struct super_block *sb, int *flags, char *data) 4593ac27a0ecSDave Kleikamp { 4594617ba13bSMingming Cao struct ext4_super_block *es; 4595617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 4596ac27a0ecSDave Kleikamp unsigned long old_sb_flags; 4597617ba13bSMingming Cao struct ext4_mount_options old_opts; 4598c79d967dSChristoph Hellwig int enable_quota = 0; 45998a266467STheodore Ts'o ext4_group_t g; 4600b3881f74STheodore Ts'o unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 4601c5e06d10SJohann Lombardi int err = 0; 4602ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 460303dafb5fSChen Gang int i, j; 4604ac27a0ecSDave Kleikamp #endif 4605d4c402d9SCurt Wohlgemuth char *orig_data = kstrdup(data, GFP_KERNEL); 4606ac27a0ecSDave Kleikamp 4607ac27a0ecSDave Kleikamp /* Store the original options */ 4608ac27a0ecSDave Kleikamp old_sb_flags = sb->s_flags; 4609ac27a0ecSDave Kleikamp old_opts.s_mount_opt = sbi->s_mount_opt; 4610a2595b8aSTheodore Ts'o old_opts.s_mount_opt2 = sbi->s_mount_opt2; 4611ac27a0ecSDave Kleikamp old_opts.s_resuid = sbi->s_resuid; 4612ac27a0ecSDave Kleikamp old_opts.s_resgid = sbi->s_resgid; 4613ac27a0ecSDave Kleikamp old_opts.s_commit_interval = sbi->s_commit_interval; 461430773840STheodore Ts'o old_opts.s_min_batch_time = sbi->s_min_batch_time; 461530773840STheodore Ts'o old_opts.s_max_batch_time = sbi->s_max_batch_time; 4616ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4617ac27a0ecSDave Kleikamp old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 4618a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 461903dafb5fSChen Gang if (sbi->s_qf_names[i]) { 462003dafb5fSChen Gang old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i], 462103dafb5fSChen Gang GFP_KERNEL); 462203dafb5fSChen Gang if (!old_opts.s_qf_names[i]) { 462303dafb5fSChen Gang for (j = 0; j < i; j++) 462403dafb5fSChen Gang kfree(old_opts.s_qf_names[j]); 46253e36a163SWei Yongjun kfree(orig_data); 462603dafb5fSChen Gang return -ENOMEM; 462703dafb5fSChen Gang } 462803dafb5fSChen Gang } else 462903dafb5fSChen Gang old_opts.s_qf_names[i] = NULL; 4630ac27a0ecSDave Kleikamp #endif 4631b3881f74STheodore Ts'o if (sbi->s_journal && sbi->s_journal->j_task->io_context) 4632b3881f74STheodore Ts'o journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 4633ac27a0ecSDave Kleikamp 4634661aa520SEric Sandeen if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 4635ac27a0ecSDave Kleikamp err = -EINVAL; 4636ac27a0ecSDave Kleikamp goto restore_opts; 4637ac27a0ecSDave Kleikamp } 4638ac27a0ecSDave Kleikamp 46396b992ff2SDarrick J. Wong if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 46406b992ff2SDarrick J. Wong test_opt(sb, JOURNAL_CHECKSUM)) { 46416b992ff2SDarrick J. Wong ext4_msg(sb, KERN_ERR, "changing journal_checksum " 46422d5b86e0SEric Sandeen "during remount not supported; ignoring"); 46432d5b86e0SEric Sandeen sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 4644c6d3d56dSDarrick J. Wong } 4645c6d3d56dSDarrick J. Wong 46466ae6514bSPiotr Sarna if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 46476ae6514bSPiotr Sarna if (test_opt2(sb, EXPLICIT_DELALLOC)) { 46486ae6514bSPiotr Sarna ext4_msg(sb, KERN_ERR, "can't mount with " 46496ae6514bSPiotr Sarna "both data=journal and delalloc"); 46506ae6514bSPiotr Sarna err = -EINVAL; 46516ae6514bSPiotr Sarna goto restore_opts; 46526ae6514bSPiotr Sarna } 46536ae6514bSPiotr Sarna if (test_opt(sb, DIOREAD_NOLOCK)) { 46546ae6514bSPiotr Sarna ext4_msg(sb, KERN_ERR, "can't mount with " 46556ae6514bSPiotr Sarna "both data=journal and dioread_nolock"); 46566ae6514bSPiotr Sarna err = -EINVAL; 46576ae6514bSPiotr Sarna goto restore_opts; 46586ae6514bSPiotr Sarna } 4659923ae0ffSRoss Zwisler if (test_opt(sb, DAX)) { 4660923ae0ffSRoss Zwisler ext4_msg(sb, KERN_ERR, "can't mount with " 4661923ae0ffSRoss Zwisler "both data=journal and dax"); 4662923ae0ffSRoss Zwisler err = -EINVAL; 4663923ae0ffSRoss Zwisler goto restore_opts; 4664923ae0ffSRoss Zwisler } 4665923ae0ffSRoss Zwisler } 4666923ae0ffSRoss Zwisler 4667923ae0ffSRoss Zwisler if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) { 4668923ae0ffSRoss Zwisler ext4_msg(sb, KERN_WARNING, "warning: refusing change of " 4669923ae0ffSRoss Zwisler "dax flag with busy inodes while remounting"); 4670923ae0ffSRoss Zwisler sbi->s_mount_opt ^= EXT4_MOUNT_DAX; 46716ae6514bSPiotr Sarna } 46726ae6514bSPiotr Sarna 46734ab2f15bSTheodore Ts'o if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 4674c67d859eSTheodore Ts'o ext4_abort(sb, "Abort forced by user"); 4675ac27a0ecSDave Kleikamp 4676ac27a0ecSDave Kleikamp sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 4677482a7425SDmitry Monakhov (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 4678ac27a0ecSDave Kleikamp 4679ac27a0ecSDave Kleikamp es = sbi->s_es; 4680ac27a0ecSDave Kleikamp 4681b3881f74STheodore Ts'o if (sbi->s_journal) { 4682617ba13bSMingming Cao ext4_init_journal_params(sb, sbi->s_journal); 4683b3881f74STheodore Ts'o set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4684b3881f74STheodore Ts'o } 4685ac27a0ecSDave Kleikamp 4686a2fd66d0STheodore Ts'o if (*flags & MS_LAZYTIME) 4687a2fd66d0STheodore Ts'o sb->s_flags |= MS_LAZYTIME; 4688a2fd66d0STheodore Ts'o 4689661aa520SEric Sandeen if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) { 46904ab2f15bSTheodore Ts'o if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 4691ac27a0ecSDave Kleikamp err = -EROFS; 4692ac27a0ecSDave Kleikamp goto restore_opts; 4693ac27a0ecSDave Kleikamp } 4694ac27a0ecSDave Kleikamp 4695ac27a0ecSDave Kleikamp if (*flags & MS_RDONLY) { 469638c03b34STheodore Ts'o err = sync_filesystem(sb); 469738c03b34STheodore Ts'o if (err < 0) 469838c03b34STheodore Ts'o goto restore_opts; 46990f0dd62fSChristoph Hellwig err = dquot_suspend(sb, -1); 47000f0dd62fSChristoph Hellwig if (err < 0) 4701c79d967dSChristoph Hellwig goto restore_opts; 4702c79d967dSChristoph Hellwig 4703ac27a0ecSDave Kleikamp /* 4704ac27a0ecSDave Kleikamp * First of all, the unconditional stuff we have to do 4705ac27a0ecSDave Kleikamp * to disable replay of the journal when we next remount 4706ac27a0ecSDave Kleikamp */ 4707ac27a0ecSDave Kleikamp sb->s_flags |= MS_RDONLY; 4708ac27a0ecSDave Kleikamp 4709ac27a0ecSDave Kleikamp /* 4710ac27a0ecSDave Kleikamp * OK, test if we are remounting a valid rw partition 4711ac27a0ecSDave Kleikamp * readonly, and if so set the rdonly flag and then 4712ac27a0ecSDave Kleikamp * mark the partition as valid again. 4713ac27a0ecSDave Kleikamp */ 4714617ba13bSMingming Cao if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 4715617ba13bSMingming Cao (sbi->s_mount_state & EXT4_VALID_FS)) 4716ac27a0ecSDave Kleikamp es->s_state = cpu_to_le16(sbi->s_mount_state); 4717ac27a0ecSDave Kleikamp 4718a63c9eb2STheodore Ts'o if (sbi->s_journal) 4719617ba13bSMingming Cao ext4_mark_recovery_complete(sb, es); 4720ac27a0ecSDave Kleikamp } else { 4721a13fb1a4SEric Sandeen /* Make sure we can mount this feature set readwrite */ 4722e2b911c5SDarrick J. Wong if (ext4_has_feature_readonly(sb) || 47232cb5cc8bSDarrick J. Wong !ext4_feature_set_ok(sb, 0)) { 4724ac27a0ecSDave Kleikamp err = -EROFS; 4725ac27a0ecSDave Kleikamp goto restore_opts; 4726ac27a0ecSDave Kleikamp } 4727ead6596bSEric Sandeen /* 47288a266467STheodore Ts'o * Make sure the group descriptor checksums 47290b8e58a1SAndreas Dilger * are sane. If they aren't, refuse to remount r/w. 47308a266467STheodore Ts'o */ 47318a266467STheodore Ts'o for (g = 0; g < sbi->s_groups_count; g++) { 47328a266467STheodore Ts'o struct ext4_group_desc *gdp = 47338a266467STheodore Ts'o ext4_get_group_desc(sb, g, NULL); 47348a266467STheodore Ts'o 4735feb0ab32SDarrick J. Wong if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 4736b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4737b31e1552SEric Sandeen "ext4_remount: Checksum for group %u failed (%u!=%u)", 4738e2b911c5SDarrick J. Wong g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 47398a266467STheodore Ts'o le16_to_cpu(gdp->bg_checksum)); 47406a797d27SDarrick J. Wong err = -EFSBADCRC; 47418a266467STheodore Ts'o goto restore_opts; 47428a266467STheodore Ts'o } 47438a266467STheodore Ts'o } 47448a266467STheodore Ts'o 47458a266467STheodore Ts'o /* 4746ead6596bSEric Sandeen * If we have an unprocessed orphan list hanging 4747ead6596bSEric Sandeen * around from a previously readonly bdev mount, 4748ead6596bSEric Sandeen * require a full umount/remount for now. 4749ead6596bSEric Sandeen */ 4750ead6596bSEric Sandeen if (es->s_last_orphan) { 4751b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "Couldn't " 4752ead6596bSEric Sandeen "remount RDWR because of unprocessed " 4753ead6596bSEric Sandeen "orphan inode list. Please " 4754b31e1552SEric Sandeen "umount/remount instead"); 4755ead6596bSEric Sandeen err = -EINVAL; 4756ead6596bSEric Sandeen goto restore_opts; 4757ead6596bSEric Sandeen } 4758ead6596bSEric Sandeen 4759ac27a0ecSDave Kleikamp /* 4760ac27a0ecSDave Kleikamp * Mounting a RDONLY partition read-write, so reread 4761ac27a0ecSDave Kleikamp * and store the current valid flag. (It may have 4762ac27a0ecSDave Kleikamp * been changed by e2fsck since we originally mounted 4763ac27a0ecSDave Kleikamp * the partition.) 4764ac27a0ecSDave Kleikamp */ 47650390131bSFrank Mayhar if (sbi->s_journal) 4766617ba13bSMingming Cao ext4_clear_journal_err(sb, es); 4767ac27a0ecSDave Kleikamp sbi->s_mount_state = le16_to_cpu(es->s_state); 4768617ba13bSMingming Cao if (!ext4_setup_super(sb, es, 0)) 4769ac27a0ecSDave Kleikamp sb->s_flags &= ~MS_RDONLY; 4770e2b911c5SDarrick J. Wong if (ext4_has_feature_mmp(sb)) 4771c5e06d10SJohann Lombardi if (ext4_multi_mount_protect(sb, 4772c5e06d10SJohann Lombardi le64_to_cpu(es->s_mmp_block))) { 4773c5e06d10SJohann Lombardi err = -EROFS; 4774c5e06d10SJohann Lombardi goto restore_opts; 4775c5e06d10SJohann Lombardi } 4776c79d967dSChristoph Hellwig enable_quota = 1; 4777ac27a0ecSDave Kleikamp } 4778ac27a0ecSDave Kleikamp } 4779bfff6873SLukas Czerner 4780bfff6873SLukas Czerner /* 4781bfff6873SLukas Czerner * Reinitialize lazy itable initialization thread based on 4782bfff6873SLukas Czerner * current settings 4783bfff6873SLukas Czerner */ 4784bfff6873SLukas Czerner if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE)) 4785bfff6873SLukas Czerner ext4_unregister_li_request(sb); 4786bfff6873SLukas Czerner else { 4787bfff6873SLukas Czerner ext4_group_t first_not_zeroed; 4788bfff6873SLukas Czerner first_not_zeroed = ext4_has_uninit_itable(sb); 4789bfff6873SLukas Czerner ext4_register_li_request(sb, first_not_zeroed); 4790bfff6873SLukas Czerner } 4791bfff6873SLukas Czerner 47926fd058f7STheodore Ts'o ext4_setup_system_zone(sb); 4793d096ad0fSMichael Tokarev if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY)) 4794e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 47950390131bSFrank Mayhar 4796ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4797ac27a0ecSDave Kleikamp /* Release old quota file names */ 4798a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 4799ac27a0ecSDave Kleikamp kfree(old_opts.s_qf_names[i]); 48007c319d32SAditya Kali if (enable_quota) { 48017c319d32SAditya Kali if (sb_any_quota_suspended(sb)) 48020f0dd62fSChristoph Hellwig dquot_resume(sb, -1); 4803e2b911c5SDarrick J. Wong else if (ext4_has_feature_quota(sb)) { 48047c319d32SAditya Kali err = ext4_enable_quotas(sb); 480507724f98STheodore Ts'o if (err) 48067c319d32SAditya Kali goto restore_opts; 48077c319d32SAditya Kali } 48087c319d32SAditya Kali } 48097c319d32SAditya Kali #endif 4810d4c402d9SCurt Wohlgemuth 4811a26f4992STheodore Ts'o *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME); 4812d4c402d9SCurt Wohlgemuth ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 4813d4c402d9SCurt Wohlgemuth kfree(orig_data); 4814ac27a0ecSDave Kleikamp return 0; 48150b8e58a1SAndreas Dilger 4816ac27a0ecSDave Kleikamp restore_opts: 4817ac27a0ecSDave Kleikamp sb->s_flags = old_sb_flags; 4818ac27a0ecSDave Kleikamp sbi->s_mount_opt = old_opts.s_mount_opt; 4819a2595b8aSTheodore Ts'o sbi->s_mount_opt2 = old_opts.s_mount_opt2; 4820ac27a0ecSDave Kleikamp sbi->s_resuid = old_opts.s_resuid; 4821ac27a0ecSDave Kleikamp sbi->s_resgid = old_opts.s_resgid; 4822ac27a0ecSDave Kleikamp sbi->s_commit_interval = old_opts.s_commit_interval; 482330773840STheodore Ts'o sbi->s_min_batch_time = old_opts.s_min_batch_time; 482430773840STheodore Ts'o sbi->s_max_batch_time = old_opts.s_max_batch_time; 4825ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4826ac27a0ecSDave Kleikamp sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 4827a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 4828ac27a0ecSDave Kleikamp kfree(sbi->s_qf_names[i]); 4829ac27a0ecSDave Kleikamp sbi->s_qf_names[i] = old_opts.s_qf_names[i]; 4830ac27a0ecSDave Kleikamp } 4831ac27a0ecSDave Kleikamp #endif 4832d4c402d9SCurt Wohlgemuth kfree(orig_data); 4833ac27a0ecSDave Kleikamp return err; 4834ac27a0ecSDave Kleikamp } 4835ac27a0ecSDave Kleikamp 4836689c958cSLi Xi #ifdef CONFIG_QUOTA 4837689c958cSLi Xi static int ext4_statfs_project(struct super_block *sb, 4838689c958cSLi Xi kprojid_t projid, struct kstatfs *buf) 4839689c958cSLi Xi { 4840689c958cSLi Xi struct kqid qid; 4841689c958cSLi Xi struct dquot *dquot; 4842689c958cSLi Xi u64 limit; 4843689c958cSLi Xi u64 curblock; 4844689c958cSLi Xi 4845689c958cSLi Xi qid = make_kqid_projid(projid); 4846689c958cSLi Xi dquot = dqget(sb, qid); 4847689c958cSLi Xi if (IS_ERR(dquot)) 4848689c958cSLi Xi return PTR_ERR(dquot); 4849689c958cSLi Xi spin_lock(&dq_data_lock); 4850689c958cSLi Xi 4851689c958cSLi Xi limit = (dquot->dq_dqb.dqb_bsoftlimit ? 4852689c958cSLi Xi dquot->dq_dqb.dqb_bsoftlimit : 4853689c958cSLi Xi dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits; 4854689c958cSLi Xi if (limit && buf->f_blocks > limit) { 4855689c958cSLi Xi curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits; 4856689c958cSLi Xi buf->f_blocks = limit; 4857689c958cSLi Xi buf->f_bfree = buf->f_bavail = 4858689c958cSLi Xi (buf->f_blocks > curblock) ? 4859689c958cSLi Xi (buf->f_blocks - curblock) : 0; 4860689c958cSLi Xi } 4861689c958cSLi Xi 4862689c958cSLi Xi limit = dquot->dq_dqb.dqb_isoftlimit ? 4863689c958cSLi Xi dquot->dq_dqb.dqb_isoftlimit : 4864689c958cSLi Xi dquot->dq_dqb.dqb_ihardlimit; 4865689c958cSLi Xi if (limit && buf->f_files > limit) { 4866689c958cSLi Xi buf->f_files = limit; 4867689c958cSLi Xi buf->f_ffree = 4868689c958cSLi Xi (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 4869689c958cSLi Xi (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 4870689c958cSLi Xi } 4871689c958cSLi Xi 4872689c958cSLi Xi spin_unlock(&dq_data_lock); 4873689c958cSLi Xi dqput(dquot); 4874689c958cSLi Xi return 0; 4875689c958cSLi Xi } 4876689c958cSLi Xi #endif 4877689c958cSLi Xi 4878617ba13bSMingming Cao static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 4879ac27a0ecSDave Kleikamp { 4880ac27a0ecSDave Kleikamp struct super_block *sb = dentry->d_sb; 4881617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 4882617ba13bSMingming Cao struct ext4_super_block *es = sbi->s_es; 488327dd4385SLukas Czerner ext4_fsblk_t overhead = 0, resv_blocks; 4884960cc398SPekka Enberg u64 fsid; 4885d02a9391SKazuya Mio s64 bfree; 488627dd4385SLukas Czerner resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 4887ac27a0ecSDave Kleikamp 4888952fc18eSTheodore Ts'o if (!test_opt(sb, MINIX_DF)) 4889952fc18eSTheodore Ts'o overhead = sbi->s_overhead; 4890ac27a0ecSDave Kleikamp 4891617ba13bSMingming Cao buf->f_type = EXT4_SUPER_MAGIC; 4892ac27a0ecSDave Kleikamp buf->f_bsize = sb->s_blocksize; 4893b72f78cbSEric Sandeen buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 489457042651STheodore Ts'o bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 489557042651STheodore Ts'o percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 4896d02a9391SKazuya Mio /* prevent underflow in case that few free space is available */ 489757042651STheodore Ts'o buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 489827dd4385SLukas Czerner buf->f_bavail = buf->f_bfree - 489927dd4385SLukas Czerner (ext4_r_blocks_count(es) + resv_blocks); 490027dd4385SLukas Czerner if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 4901ac27a0ecSDave Kleikamp buf->f_bavail = 0; 4902ac27a0ecSDave Kleikamp buf->f_files = le32_to_cpu(es->s_inodes_count); 490352d9f3b4SPeter Zijlstra buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 4904617ba13bSMingming Cao buf->f_namelen = EXT4_NAME_LEN; 4905960cc398SPekka Enberg fsid = le64_to_cpup((void *)es->s_uuid) ^ 4906960cc398SPekka Enberg le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 4907960cc398SPekka Enberg buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 4908960cc398SPekka Enberg buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 49090b8e58a1SAndreas Dilger 4910689c958cSLi Xi #ifdef CONFIG_QUOTA 4911689c958cSLi Xi if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 4912689c958cSLi Xi sb_has_quota_limits_enabled(sb, PRJQUOTA)) 4913689c958cSLi Xi ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 4914689c958cSLi Xi #endif 4915ac27a0ecSDave Kleikamp return 0; 4916ac27a0ecSDave Kleikamp } 4917ac27a0ecSDave Kleikamp 49180b8e58a1SAndreas Dilger /* Helper function for writing quotas on sync - we need to start transaction 49190b8e58a1SAndreas Dilger * before quota file is locked for write. Otherwise the are possible deadlocks: 4920ac27a0ecSDave Kleikamp * Process 1 Process 2 4921617ba13bSMingming Cao * ext4_create() quota_sync() 4922dab291afSMingming Cao * jbd2_journal_start() write_dquot() 4923871a2931SChristoph Hellwig * dquot_initialize() down(dqio_mutex) 4924dab291afSMingming Cao * down(dqio_mutex) jbd2_journal_start() 4925ac27a0ecSDave Kleikamp * 4926ac27a0ecSDave Kleikamp */ 4927ac27a0ecSDave Kleikamp 4928ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4929ac27a0ecSDave Kleikamp 4930ac27a0ecSDave Kleikamp static inline struct inode *dquot_to_inode(struct dquot *dquot) 4931ac27a0ecSDave Kleikamp { 49324c376dcaSEric W. Biederman return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 4933ac27a0ecSDave Kleikamp } 4934ac27a0ecSDave Kleikamp 4935617ba13bSMingming Cao static int ext4_write_dquot(struct dquot *dquot) 4936ac27a0ecSDave Kleikamp { 4937ac27a0ecSDave Kleikamp int ret, err; 4938ac27a0ecSDave Kleikamp handle_t *handle; 4939ac27a0ecSDave Kleikamp struct inode *inode; 4940ac27a0ecSDave Kleikamp 4941ac27a0ecSDave Kleikamp inode = dquot_to_inode(dquot); 49429924a92aSTheodore Ts'o handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 4943617ba13bSMingming Cao EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 4944ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 4945ac27a0ecSDave Kleikamp return PTR_ERR(handle); 4946ac27a0ecSDave Kleikamp ret = dquot_commit(dquot); 4947617ba13bSMingming Cao err = ext4_journal_stop(handle); 4948ac27a0ecSDave Kleikamp if (!ret) 4949ac27a0ecSDave Kleikamp ret = err; 4950ac27a0ecSDave Kleikamp return ret; 4951ac27a0ecSDave Kleikamp } 4952ac27a0ecSDave Kleikamp 4953617ba13bSMingming Cao static int ext4_acquire_dquot(struct dquot *dquot) 4954ac27a0ecSDave Kleikamp { 4955ac27a0ecSDave Kleikamp int ret, err; 4956ac27a0ecSDave Kleikamp handle_t *handle; 4957ac27a0ecSDave Kleikamp 49589924a92aSTheodore Ts'o handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 4959617ba13bSMingming Cao EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 4960ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 4961ac27a0ecSDave Kleikamp return PTR_ERR(handle); 4962ac27a0ecSDave Kleikamp ret = dquot_acquire(dquot); 4963617ba13bSMingming Cao err = ext4_journal_stop(handle); 4964ac27a0ecSDave Kleikamp if (!ret) 4965ac27a0ecSDave Kleikamp ret = err; 4966ac27a0ecSDave Kleikamp return ret; 4967ac27a0ecSDave Kleikamp } 4968ac27a0ecSDave Kleikamp 4969617ba13bSMingming Cao static int ext4_release_dquot(struct dquot *dquot) 4970ac27a0ecSDave Kleikamp { 4971ac27a0ecSDave Kleikamp int ret, err; 4972ac27a0ecSDave Kleikamp handle_t *handle; 4973ac27a0ecSDave Kleikamp 49749924a92aSTheodore Ts'o handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 4975617ba13bSMingming Cao EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 49769c3013e9SJan Kara if (IS_ERR(handle)) { 49779c3013e9SJan Kara /* Release dquot anyway to avoid endless cycle in dqput() */ 49789c3013e9SJan Kara dquot_release(dquot); 4979ac27a0ecSDave Kleikamp return PTR_ERR(handle); 49809c3013e9SJan Kara } 4981ac27a0ecSDave Kleikamp ret = dquot_release(dquot); 4982617ba13bSMingming Cao err = ext4_journal_stop(handle); 4983ac27a0ecSDave Kleikamp if (!ret) 4984ac27a0ecSDave Kleikamp ret = err; 4985ac27a0ecSDave Kleikamp return ret; 4986ac27a0ecSDave Kleikamp } 4987ac27a0ecSDave Kleikamp 4988617ba13bSMingming Cao static int ext4_mark_dquot_dirty(struct dquot *dquot) 4989ac27a0ecSDave Kleikamp { 4990262b4662SJan Kara struct super_block *sb = dquot->dq_sb; 4991262b4662SJan Kara struct ext4_sb_info *sbi = EXT4_SB(sb); 4992262b4662SJan Kara 49932c8be6b2SJan Kara /* Are we journaling quotas? */ 4994e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb) || 4995262b4662SJan Kara sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 4996ac27a0ecSDave Kleikamp dquot_mark_dquot_dirty(dquot); 4997617ba13bSMingming Cao return ext4_write_dquot(dquot); 4998ac27a0ecSDave Kleikamp } else { 4999ac27a0ecSDave Kleikamp return dquot_mark_dquot_dirty(dquot); 5000ac27a0ecSDave Kleikamp } 5001ac27a0ecSDave Kleikamp } 5002ac27a0ecSDave Kleikamp 5003617ba13bSMingming Cao static int ext4_write_info(struct super_block *sb, int type) 5004ac27a0ecSDave Kleikamp { 5005ac27a0ecSDave Kleikamp int ret, err; 5006ac27a0ecSDave Kleikamp handle_t *handle; 5007ac27a0ecSDave Kleikamp 5008ac27a0ecSDave Kleikamp /* Data block + inode block */ 50092b0143b5SDavid Howells handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); 5010ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 5011ac27a0ecSDave Kleikamp return PTR_ERR(handle); 5012ac27a0ecSDave Kleikamp ret = dquot_commit_info(sb, type); 5013617ba13bSMingming Cao err = ext4_journal_stop(handle); 5014ac27a0ecSDave Kleikamp if (!ret) 5015ac27a0ecSDave Kleikamp ret = err; 5016ac27a0ecSDave Kleikamp return ret; 5017ac27a0ecSDave Kleikamp } 5018ac27a0ecSDave Kleikamp 5019ac27a0ecSDave Kleikamp /* 5020ac27a0ecSDave Kleikamp * Turn on quotas during mount time - we need to find 5021ac27a0ecSDave Kleikamp * the quota file and such... 5022ac27a0ecSDave Kleikamp */ 5023617ba13bSMingming Cao static int ext4_quota_on_mount(struct super_block *sb, int type) 5024ac27a0ecSDave Kleikamp { 5025287a8095SChristoph Hellwig return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], 5026617ba13bSMingming Cao EXT4_SB(sb)->s_jquota_fmt, type); 5027ac27a0ecSDave Kleikamp } 5028ac27a0ecSDave Kleikamp 5029daf647d2STheodore Ts'o static void lockdep_set_quota_inode(struct inode *inode, int subclass) 5030daf647d2STheodore Ts'o { 5031daf647d2STheodore Ts'o struct ext4_inode_info *ei = EXT4_I(inode); 5032daf647d2STheodore Ts'o 5033daf647d2STheodore Ts'o /* The first argument of lockdep_set_subclass has to be 5034daf647d2STheodore Ts'o * *exactly* the same as the argument to init_rwsem() --- in 5035daf647d2STheodore Ts'o * this case, in init_once() --- or lockdep gets unhappy 5036daf647d2STheodore Ts'o * because the name of the lock is set using the 5037daf647d2STheodore Ts'o * stringification of the argument to init_rwsem(). 5038daf647d2STheodore Ts'o */ 5039daf647d2STheodore Ts'o (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 5040daf647d2STheodore Ts'o lockdep_set_subclass(&ei->i_data_sem, subclass); 5041daf647d2STheodore Ts'o } 5042daf647d2STheodore Ts'o 5043ac27a0ecSDave Kleikamp /* 5044ac27a0ecSDave Kleikamp * Standard function to be called on quota_on 5045ac27a0ecSDave Kleikamp */ 5046617ba13bSMingming Cao static int ext4_quota_on(struct super_block *sb, int type, int format_id, 5047f00c9e44SJan Kara struct path *path) 5048ac27a0ecSDave Kleikamp { 5049ac27a0ecSDave Kleikamp int err; 5050ac27a0ecSDave Kleikamp 5051ac27a0ecSDave Kleikamp if (!test_opt(sb, QUOTA)) 5052ac27a0ecSDave Kleikamp return -EINVAL; 50530623543bSJan Kara 5054ac27a0ecSDave Kleikamp /* Quotafile not on the same filesystem? */ 5055d8c9584eSAl Viro if (path->dentry->d_sb != sb) 5056ac27a0ecSDave Kleikamp return -EXDEV; 50570623543bSJan Kara /* Journaling quota? */ 50580623543bSJan Kara if (EXT4_SB(sb)->s_qf_names[type]) { 50592b2d6d01STheodore Ts'o /* Quotafile not in fs root? */ 5060f00c9e44SJan Kara if (path->dentry->d_parent != sb->s_root) 5061b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 5062b31e1552SEric Sandeen "Quota file not on filesystem root. " 5063b31e1552SEric Sandeen "Journaled quota will not work"); 50640623543bSJan Kara } 50650623543bSJan Kara 50660623543bSJan Kara /* 50670623543bSJan Kara * When we journal data on quota file, we have to flush journal to see 50680623543bSJan Kara * all updates to the file when we bypass pagecache... 50690623543bSJan Kara */ 50700390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal && 50712b0143b5SDavid Howells ext4_should_journal_data(d_inode(path->dentry))) { 50720623543bSJan Kara /* 50730623543bSJan Kara * We don't need to lock updates but journal_flush() could 50740623543bSJan Kara * otherwise be livelocked... 50750623543bSJan Kara */ 50760623543bSJan Kara jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 50777ffe1ea8SHidehiro Kawai err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 50780623543bSJan Kara jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 5079f00c9e44SJan Kara if (err) 50807ffe1ea8SHidehiro Kawai return err; 50817ffe1ea8SHidehiro Kawai } 5082daf647d2STheodore Ts'o lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 5083daf647d2STheodore Ts'o err = dquot_quota_on(sb, type, format_id, path); 5084daf647d2STheodore Ts'o if (err) 5085daf647d2STheodore Ts'o lockdep_set_quota_inode(path->dentry->d_inode, 5086daf647d2STheodore Ts'o I_DATA_SEM_NORMAL); 5087daf647d2STheodore Ts'o return err; 5088ac27a0ecSDave Kleikamp } 5089ac27a0ecSDave Kleikamp 50907c319d32SAditya Kali static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 50917c319d32SAditya Kali unsigned int flags) 50927c319d32SAditya Kali { 50937c319d32SAditya Kali int err; 50947c319d32SAditya Kali struct inode *qf_inode; 5095a2d4a646SJan Kara unsigned long qf_inums[EXT4_MAXQUOTAS] = { 50967c319d32SAditya Kali le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5097689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5098689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 50997c319d32SAditya Kali }; 51007c319d32SAditya Kali 5101e2b911c5SDarrick J. Wong BUG_ON(!ext4_has_feature_quota(sb)); 51027c319d32SAditya Kali 51037c319d32SAditya Kali if (!qf_inums[type]) 51047c319d32SAditya Kali return -EPERM; 51057c319d32SAditya Kali 51067c319d32SAditya Kali qf_inode = ext4_iget(sb, qf_inums[type]); 51077c319d32SAditya Kali if (IS_ERR(qf_inode)) { 51087c319d32SAditya Kali ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 51097c319d32SAditya Kali return PTR_ERR(qf_inode); 51107c319d32SAditya Kali } 51117c319d32SAditya Kali 5112bcb13850SJan Kara /* Don't account quota for quota files to avoid recursion */ 5113bcb13850SJan Kara qf_inode->i_flags |= S_NOQUOTA; 5114daf647d2STheodore Ts'o lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 51157c319d32SAditya Kali err = dquot_enable(qf_inode, type, format_id, flags); 51167c319d32SAditya Kali iput(qf_inode); 5117daf647d2STheodore Ts'o if (err) 5118daf647d2STheodore Ts'o lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 51197c319d32SAditya Kali 51207c319d32SAditya Kali return err; 51217c319d32SAditya Kali } 51227c319d32SAditya Kali 51237c319d32SAditya Kali /* Enable usage tracking for all quota types. */ 51247c319d32SAditya Kali static int ext4_enable_quotas(struct super_block *sb) 51257c319d32SAditya Kali { 51267c319d32SAditya Kali int type, err = 0; 5127a2d4a646SJan Kara unsigned long qf_inums[EXT4_MAXQUOTAS] = { 51287c319d32SAditya Kali le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 5129689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 5130689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 51317c319d32SAditya Kali }; 51327c319d32SAditya Kali 51337c319d32SAditya Kali sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 5134a2d4a646SJan Kara for (type = 0; type < EXT4_MAXQUOTAS; type++) { 51357c319d32SAditya Kali if (qf_inums[type]) { 51367c319d32SAditya Kali err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 51377c319d32SAditya Kali DQUOT_USAGE_ENABLED); 51387c319d32SAditya Kali if (err) { 51397c319d32SAditya Kali ext4_warning(sb, 514072ba7450STheodore Ts'o "Failed to enable quota tracking " 514172ba7450STheodore Ts'o "(type=%d, err=%d). Please run " 514272ba7450STheodore Ts'o "e2fsck to fix.", type, err); 51437c319d32SAditya Kali return err; 51447c319d32SAditya Kali } 51457c319d32SAditya Kali } 51467c319d32SAditya Kali } 51477c319d32SAditya Kali return 0; 51487c319d32SAditya Kali } 51497c319d32SAditya Kali 5150ca0e05e4SDmitry Monakhov static int ext4_quota_off(struct super_block *sb, int type) 5151ca0e05e4SDmitry Monakhov { 515221f97697SJan Kara struct inode *inode = sb_dqopt(sb)->files[type]; 515321f97697SJan Kara handle_t *handle; 515421f97697SJan Kara 515587009d86SDmitry Monakhov /* Force all delayed allocation blocks to be allocated. 515687009d86SDmitry Monakhov * Caller already holds s_umount sem */ 515787009d86SDmitry Monakhov if (test_opt(sb, DELALLOC)) 5158ca0e05e4SDmitry Monakhov sync_filesystem(sb); 5159ca0e05e4SDmitry Monakhov 51600b268590SAmir Goldstein if (!inode) 51610b268590SAmir Goldstein goto out; 51620b268590SAmir Goldstein 516321f97697SJan Kara /* Update modification times of quota files when userspace can 516421f97697SJan Kara * start looking at them */ 51659924a92aSTheodore Ts'o handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 516621f97697SJan Kara if (IS_ERR(handle)) 516721f97697SJan Kara goto out; 516821f97697SJan Kara inode->i_mtime = inode->i_ctime = CURRENT_TIME; 516921f97697SJan Kara ext4_mark_inode_dirty(handle, inode); 517021f97697SJan Kara ext4_journal_stop(handle); 517121f97697SJan Kara 517221f97697SJan Kara out: 5173ca0e05e4SDmitry Monakhov return dquot_quota_off(sb, type); 5174ca0e05e4SDmitry Monakhov } 5175ca0e05e4SDmitry Monakhov 5176ac27a0ecSDave Kleikamp /* Read data from quotafile - avoid pagecache and such because we cannot afford 5177ac27a0ecSDave Kleikamp * acquiring the locks... As quota files are never truncated and quota code 5178ac27a0ecSDave Kleikamp * itself serializes the operations (and no one else should touch the files) 5179ac27a0ecSDave Kleikamp * we don't have to be afraid of races */ 5180617ba13bSMingming Cao static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 5181ac27a0ecSDave Kleikamp size_t len, loff_t off) 5182ac27a0ecSDave Kleikamp { 5183ac27a0ecSDave Kleikamp struct inode *inode = sb_dqopt(sb)->files[type]; 5184725d26d3SAneesh Kumar K.V ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5185ac27a0ecSDave Kleikamp int offset = off & (sb->s_blocksize - 1); 5186ac27a0ecSDave Kleikamp int tocopy; 5187ac27a0ecSDave Kleikamp size_t toread; 5188ac27a0ecSDave Kleikamp struct buffer_head *bh; 5189ac27a0ecSDave Kleikamp loff_t i_size = i_size_read(inode); 5190ac27a0ecSDave Kleikamp 5191ac27a0ecSDave Kleikamp if (off > i_size) 5192ac27a0ecSDave Kleikamp return 0; 5193ac27a0ecSDave Kleikamp if (off+len > i_size) 5194ac27a0ecSDave Kleikamp len = i_size-off; 5195ac27a0ecSDave Kleikamp toread = len; 5196ac27a0ecSDave Kleikamp while (toread > 0) { 5197ac27a0ecSDave Kleikamp tocopy = sb->s_blocksize - offset < toread ? 5198ac27a0ecSDave Kleikamp sb->s_blocksize - offset : toread; 51991c215028STheodore Ts'o bh = ext4_bread(NULL, inode, blk, 0); 52001c215028STheodore Ts'o if (IS_ERR(bh)) 52011c215028STheodore Ts'o return PTR_ERR(bh); 5202ac27a0ecSDave Kleikamp if (!bh) /* A hole? */ 5203ac27a0ecSDave Kleikamp memset(data, 0, tocopy); 5204ac27a0ecSDave Kleikamp else 5205ac27a0ecSDave Kleikamp memcpy(data, bh->b_data+offset, tocopy); 5206ac27a0ecSDave Kleikamp brelse(bh); 5207ac27a0ecSDave Kleikamp offset = 0; 5208ac27a0ecSDave Kleikamp toread -= tocopy; 5209ac27a0ecSDave Kleikamp data += tocopy; 5210ac27a0ecSDave Kleikamp blk++; 5211ac27a0ecSDave Kleikamp } 5212ac27a0ecSDave Kleikamp return len; 5213ac27a0ecSDave Kleikamp } 5214ac27a0ecSDave Kleikamp 5215ac27a0ecSDave Kleikamp /* Write to quotafile (we know the transaction is already started and has 5216ac27a0ecSDave Kleikamp * enough credits) */ 5217617ba13bSMingming Cao static ssize_t ext4_quota_write(struct super_block *sb, int type, 5218ac27a0ecSDave Kleikamp const char *data, size_t len, loff_t off) 5219ac27a0ecSDave Kleikamp { 5220ac27a0ecSDave Kleikamp struct inode *inode = sb_dqopt(sb)->files[type]; 5221725d26d3SAneesh Kumar K.V ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 52221c215028STheodore Ts'o int err, offset = off & (sb->s_blocksize - 1); 5223c5e298aeSTheodore Ts'o int retries = 0; 5224ac27a0ecSDave Kleikamp struct buffer_head *bh; 5225ac27a0ecSDave Kleikamp handle_t *handle = journal_current_handle(); 5226ac27a0ecSDave Kleikamp 52270390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal && !handle) { 5228b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5229b31e1552SEric Sandeen " cancelled because transaction is not started", 52309c3013e9SJan Kara (unsigned long long)off, (unsigned long long)len); 52319c3013e9SJan Kara return -EIO; 52329c3013e9SJan Kara } 523367eeb568SDmitry Monakhov /* 523467eeb568SDmitry Monakhov * Since we account only one data block in transaction credits, 523567eeb568SDmitry Monakhov * then it is impossible to cross a block boundary. 523667eeb568SDmitry Monakhov */ 523767eeb568SDmitry Monakhov if (sb->s_blocksize - offset < len) { 523867eeb568SDmitry Monakhov ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 523967eeb568SDmitry Monakhov " cancelled because not block aligned", 524067eeb568SDmitry Monakhov (unsigned long long)off, (unsigned long long)len); 524167eeb568SDmitry Monakhov return -EIO; 524267eeb568SDmitry Monakhov } 524367eeb568SDmitry Monakhov 5244c5e298aeSTheodore Ts'o do { 5245c5e298aeSTheodore Ts'o bh = ext4_bread(handle, inode, blk, 5246c5e298aeSTheodore Ts'o EXT4_GET_BLOCKS_CREATE | 5247c5e298aeSTheodore Ts'o EXT4_GET_BLOCKS_METADATA_NOFAIL); 5248c5e298aeSTheodore Ts'o } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) && 5249c5e298aeSTheodore Ts'o ext4_should_retry_alloc(inode->i_sb, &retries)); 52501c215028STheodore Ts'o if (IS_ERR(bh)) 52511c215028STheodore Ts'o return PTR_ERR(bh); 5252ac27a0ecSDave Kleikamp if (!bh) 5253ac27a0ecSDave Kleikamp goto out; 52545d601255Sliang xie BUFFER_TRACE(bh, "get write access"); 5255617ba13bSMingming Cao err = ext4_journal_get_write_access(handle, bh); 5256ac27a0ecSDave Kleikamp if (err) { 5257ac27a0ecSDave Kleikamp brelse(bh); 52581c215028STheodore Ts'o return err; 5259ac27a0ecSDave Kleikamp } 5260ac27a0ecSDave Kleikamp lock_buffer(bh); 526167eeb568SDmitry Monakhov memcpy(bh->b_data+offset, data, len); 5262ac27a0ecSDave Kleikamp flush_dcache_page(bh->b_page); 5263ac27a0ecSDave Kleikamp unlock_buffer(bh); 52640390131bSFrank Mayhar err = ext4_handle_dirty_metadata(handle, NULL, bh); 5265ac27a0ecSDave Kleikamp brelse(bh); 5266ac27a0ecSDave Kleikamp out: 526767eeb568SDmitry Monakhov if (inode->i_size < off + len) { 526867eeb568SDmitry Monakhov i_size_write(inode, off + len); 5269617ba13bSMingming Cao EXT4_I(inode)->i_disksize = inode->i_size; 5270617ba13bSMingming Cao ext4_mark_inode_dirty(handle, inode); 527121f97697SJan Kara } 527267eeb568SDmitry Monakhov return len; 5273ac27a0ecSDave Kleikamp } 5274ac27a0ecSDave Kleikamp 52758f0e8746STheodore Ts'o static int ext4_get_next_id(struct super_block *sb, struct kqid *qid) 52768f0e8746STheodore Ts'o { 52778f0e8746STheodore Ts'o const struct quota_format_ops *ops; 52788f0e8746STheodore Ts'o 52798f0e8746STheodore Ts'o if (!sb_has_quota_loaded(sb, qid->type)) 52808f0e8746STheodore Ts'o return -ESRCH; 52818f0e8746STheodore Ts'o ops = sb_dqopt(sb)->ops[qid->type]; 52828f0e8746STheodore Ts'o if (!ops || !ops->get_next_id) 52838f0e8746STheodore Ts'o return -ENOSYS; 52848f0e8746STheodore Ts'o return dquot_get_next_id(sb, qid); 52858f0e8746STheodore Ts'o } 5286ac27a0ecSDave Kleikamp #endif 5287ac27a0ecSDave Kleikamp 5288152a0836SAl Viro static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 5289152a0836SAl Viro const char *dev_name, void *data) 5290ac27a0ecSDave Kleikamp { 5291152a0836SAl Viro return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 5292ac27a0ecSDave Kleikamp } 5293ac27a0ecSDave Kleikamp 5294c290ea01SJan Kara #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 529524b58424STheodore Ts'o static inline void register_as_ext2(void) 529624b58424STheodore Ts'o { 529724b58424STheodore Ts'o int err = register_filesystem(&ext2_fs_type); 529824b58424STheodore Ts'o if (err) 529924b58424STheodore Ts'o printk(KERN_WARNING 530024b58424STheodore Ts'o "EXT4-fs: Unable to register as ext2 (%d)\n", err); 530124b58424STheodore Ts'o } 530224b58424STheodore Ts'o 530324b58424STheodore Ts'o static inline void unregister_as_ext2(void) 530424b58424STheodore Ts'o { 530524b58424STheodore Ts'o unregister_filesystem(&ext2_fs_type); 530624b58424STheodore Ts'o } 53072035e776STheodore Ts'o 53082035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb) 53092035e776STheodore Ts'o { 5310e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext2_incompat_features(sb)) 53112035e776STheodore Ts'o return 0; 53122035e776STheodore Ts'o if (sb->s_flags & MS_RDONLY) 53132035e776STheodore Ts'o return 1; 5314e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext2_ro_compat_features(sb)) 53152035e776STheodore Ts'o return 0; 53162035e776STheodore Ts'o return 1; 53172035e776STheodore Ts'o } 531824b58424STheodore Ts'o #else 531924b58424STheodore Ts'o static inline void register_as_ext2(void) { } 532024b58424STheodore Ts'o static inline void unregister_as_ext2(void) { } 53212035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 532224b58424STheodore Ts'o #endif 532324b58424STheodore Ts'o 532424b58424STheodore Ts'o static inline void register_as_ext3(void) 532524b58424STheodore Ts'o { 532624b58424STheodore Ts'o int err = register_filesystem(&ext3_fs_type); 532724b58424STheodore Ts'o if (err) 532824b58424STheodore Ts'o printk(KERN_WARNING 532924b58424STheodore Ts'o "EXT4-fs: Unable to register as ext3 (%d)\n", err); 533024b58424STheodore Ts'o } 533124b58424STheodore Ts'o 533224b58424STheodore Ts'o static inline void unregister_as_ext3(void) 533324b58424STheodore Ts'o { 533424b58424STheodore Ts'o unregister_filesystem(&ext3_fs_type); 533524b58424STheodore Ts'o } 53362035e776STheodore Ts'o 53372035e776STheodore Ts'o static inline int ext3_feature_set_ok(struct super_block *sb) 53382035e776STheodore Ts'o { 5339e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext3_incompat_features(sb)) 53402035e776STheodore Ts'o return 0; 5341e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal(sb)) 53422035e776STheodore Ts'o return 0; 53432035e776STheodore Ts'o if (sb->s_flags & MS_RDONLY) 53442035e776STheodore Ts'o return 1; 5345e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext3_ro_compat_features(sb)) 53462035e776STheodore Ts'o return 0; 53472035e776STheodore Ts'o return 1; 53482035e776STheodore Ts'o } 534924b58424STheodore Ts'o 535003010a33STheodore Ts'o static struct file_system_type ext4_fs_type = { 5351ac27a0ecSDave Kleikamp .owner = THIS_MODULE, 535203010a33STheodore Ts'o .name = "ext4", 5353152a0836SAl Viro .mount = ext4_mount, 5354ac27a0ecSDave Kleikamp .kill_sb = kill_block_super, 5355ac27a0ecSDave Kleikamp .fs_flags = FS_REQUIRES_DEV, 5356ac27a0ecSDave Kleikamp }; 53577f78e035SEric W. Biederman MODULE_ALIAS_FS("ext4"); 5358ac27a0ecSDave Kleikamp 5359e9e3bcecSEric Sandeen /* Shared across all ext4 file systems */ 5360e9e3bcecSEric Sandeen wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 5361e9e3bcecSEric Sandeen 53625dabfc78STheodore Ts'o static int __init ext4_init_fs(void) 5363ac27a0ecSDave Kleikamp { 5364e9e3bcecSEric Sandeen int i, err; 5365c9de560dSAlex Tomas 5366e294a537STheodore Ts'o ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 536707c0c5d8SAl Viro ext4_li_info = NULL; 536807c0c5d8SAl Viro mutex_init(&ext4_li_mtx); 536907c0c5d8SAl Viro 53709a4c8019SCarlos Maiolino /* Build-time check for flags consistency */ 537112e9b892SDmitry Monakhov ext4_check_flag_values(); 5372e9e3bcecSEric Sandeen 5373e142d052SJan Kara for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 5374e9e3bcecSEric Sandeen init_waitqueue_head(&ext4__ioend_wq[i]); 5375e9e3bcecSEric Sandeen 537651865fdaSZheng Liu err = ext4_init_es(); 53776fd058f7STheodore Ts'o if (err) 53786fd058f7STheodore Ts'o return err; 537951865fdaSZheng Liu 538051865fdaSZheng Liu err = ext4_init_pageio(); 538151865fdaSZheng Liu if (err) 5382b5799018STheodore Ts'o goto out5; 538351865fdaSZheng Liu 53845dabfc78STheodore Ts'o err = ext4_init_system_zone(); 5385bd2d0210STheodore Ts'o if (err) 5386b5799018STheodore Ts'o goto out4; 5387857ac889SLukas Czerner 5388b5799018STheodore Ts'o err = ext4_init_sysfs(); 5389dd68314cSTheodore Ts'o if (err) 5390b5799018STheodore Ts'o goto out3; 5391857ac889SLukas Czerner 53925dabfc78STheodore Ts'o err = ext4_init_mballoc(); 5393ac27a0ecSDave Kleikamp if (err) 5394c9de560dSAlex Tomas goto out2; 5395ac27a0ecSDave Kleikamp err = init_inodecache(); 5396ac27a0ecSDave Kleikamp if (err) 5397ac27a0ecSDave Kleikamp goto out1; 539824b58424STheodore Ts'o register_as_ext3(); 53992035e776STheodore Ts'o register_as_ext2(); 540003010a33STheodore Ts'o err = register_filesystem(&ext4_fs_type); 5401ac27a0ecSDave Kleikamp if (err) 5402ac27a0ecSDave Kleikamp goto out; 5403bfff6873SLukas Czerner 5404ac27a0ecSDave Kleikamp return 0; 5405ac27a0ecSDave Kleikamp out: 540624b58424STheodore Ts'o unregister_as_ext2(); 540724b58424STheodore Ts'o unregister_as_ext3(); 5408ac27a0ecSDave Kleikamp destroy_inodecache(); 5409ac27a0ecSDave Kleikamp out1: 54105dabfc78STheodore Ts'o ext4_exit_mballoc(); 54119c191f70ST Makphaibulchoke out2: 5412b5799018STheodore Ts'o ext4_exit_sysfs(); 5413b5799018STheodore Ts'o out3: 5414dd68314cSTheodore Ts'o ext4_exit_system_zone(); 5415b5799018STheodore Ts'o out4: 54165dabfc78STheodore Ts'o ext4_exit_pageio(); 5417b5799018STheodore Ts'o out5: 541851865fdaSZheng Liu ext4_exit_es(); 541951865fdaSZheng Liu 5420ac27a0ecSDave Kleikamp return err; 5421ac27a0ecSDave Kleikamp } 5422ac27a0ecSDave Kleikamp 54235dabfc78STheodore Ts'o static void __exit ext4_exit_fs(void) 5424ac27a0ecSDave Kleikamp { 5425e298e73bSChao Yu ext4_exit_crypto(); 5426bfff6873SLukas Czerner ext4_destroy_lazyinit_thread(); 542724b58424STheodore Ts'o unregister_as_ext2(); 542824b58424STheodore Ts'o unregister_as_ext3(); 542903010a33STheodore Ts'o unregister_filesystem(&ext4_fs_type); 5430ac27a0ecSDave Kleikamp destroy_inodecache(); 54315dabfc78STheodore Ts'o ext4_exit_mballoc(); 5432b5799018STheodore Ts'o ext4_exit_sysfs(); 54335dabfc78STheodore Ts'o ext4_exit_system_zone(); 54345dabfc78STheodore Ts'o ext4_exit_pageio(); 5435dd12ed14SEric Sandeen ext4_exit_es(); 5436ac27a0ecSDave Kleikamp } 5437ac27a0ecSDave Kleikamp 5438ac27a0ecSDave Kleikamp MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 543983982b6fSTheodore Ts'o MODULE_DESCRIPTION("Fourth Extended Filesystem"); 5440ac27a0ecSDave Kleikamp MODULE_LICENSE("GPL"); 54415dabfc78STheodore Ts'o module_init(ext4_init_fs) 54425dabfc78STheodore Ts'o module_exit(ext4_exit_fs) 5443