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> 24dab291afSMingming Cao #include <linux/jbd2.h> 25ac27a0ecSDave Kleikamp #include <linux/slab.h> 26ac27a0ecSDave Kleikamp #include <linux/init.h> 27ac27a0ecSDave Kleikamp #include <linux/blkdev.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> 379f6200bbSTheodore Ts'o #include <linux/proc_fs.h> 383197ebdbSTheodore Ts'o #include <linux/ctype.h> 391330593eSVignesh Babu #include <linux/log2.h> 40717d50e4SAndreas Dilger #include <linux/crc16.h> 417abc52c2SDan Magenheimer #include <linux/cleancache.h> 42ac27a0ecSDave Kleikamp #include <asm/uaccess.h> 43ac27a0ecSDave Kleikamp 44bfff6873SLukas Czerner #include <linux/kthread.h> 45bfff6873SLukas Czerner #include <linux/freezer.h> 46bfff6873SLukas Czerner 473dcf5451SChristoph Hellwig #include "ext4.h" 484a092d73STheodore Ts'o #include "ext4_extents.h" /* Needed for trace points definition */ 493dcf5451SChristoph Hellwig #include "ext4_jbd2.h" 50ac27a0ecSDave Kleikamp #include "xattr.h" 51ac27a0ecSDave Kleikamp #include "acl.h" 523661d286STheodore Ts'o #include "mballoc.h" 53ac27a0ecSDave Kleikamp 549bffad1eSTheodore Ts'o #define CREATE_TRACE_POINTS 559bffad1eSTheodore Ts'o #include <trace/events/ext4.h> 569bffad1eSTheodore Ts'o 571f109d5aSTheodore Ts'o static struct proc_dir_entry *ext4_proc_root; 583197ebdbSTheodore Ts'o static struct kset *ext4_kset; 590b75a840SLukas Czerner static struct ext4_lazy_init *ext4_li_info; 600b75a840SLukas Czerner static struct mutex ext4_li_mtx; 610b75a840SLukas Czerner static struct ext4_features *ext4_feat; 629f6200bbSTheodore Ts'o 63617ba13bSMingming Cao static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 64ac27a0ecSDave Kleikamp unsigned long journal_devnum); 652adf6da8STheodore Ts'o static int ext4_show_options(struct seq_file *seq, struct dentry *root); 66e2d67052STheodore Ts'o static int ext4_commit_super(struct super_block *sb, int sync); 67617ba13bSMingming Cao static void ext4_mark_recovery_complete(struct super_block *sb, 68617ba13bSMingming Cao struct ext4_super_block *es); 69617ba13bSMingming Cao static void ext4_clear_journal_err(struct super_block *sb, 70617ba13bSMingming Cao struct ext4_super_block *es); 71617ba13bSMingming Cao static int ext4_sync_fs(struct super_block *sb, int wait); 7206a407f1SDmitry Monakhov static int ext4_sync_fs_nojournal(struct super_block *sb, int wait); 73617ba13bSMingming Cao static int ext4_remount(struct super_block *sb, int *flags, char *data); 74617ba13bSMingming Cao static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 75c4be0c1dSTakashi Sato static int ext4_unfreeze(struct super_block *sb); 76c4be0c1dSTakashi Sato static int ext4_freeze(struct super_block *sb); 77152a0836SAl Viro static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 78152a0836SAl Viro const char *dev_name, void *data); 792035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb); 802035e776STheodore Ts'o static inline int ext3_feature_set_ok(struct super_block *sb); 81d39195c3SAmir Goldstein static int ext4_feature_set_ok(struct super_block *sb, int readonly); 82bfff6873SLukas Czerner static void ext4_destroy_lazyinit_thread(void); 83bfff6873SLukas Czerner static void ext4_unregister_li_request(struct super_block *sb); 848f1f7453SEric Sandeen static void ext4_clear_request_list(void); 8527dd4385SLukas Czerner static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t); 86ac27a0ecSDave Kleikamp 872035e776STheodore Ts'o #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 882035e776STheodore Ts'o static struct file_system_type ext2_fs_type = { 892035e776STheodore Ts'o .owner = THIS_MODULE, 902035e776STheodore Ts'o .name = "ext2", 912035e776STheodore Ts'o .mount = ext4_mount, 922035e776STheodore Ts'o .kill_sb = kill_block_super, 932035e776STheodore Ts'o .fs_flags = FS_REQUIRES_DEV, 942035e776STheodore Ts'o }; 957f78e035SEric W. Biederman MODULE_ALIAS_FS("ext2"); 96fa7614ddSEric W. Biederman MODULE_ALIAS("ext2"); 972035e776STheodore Ts'o #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 982035e776STheodore Ts'o #else 992035e776STheodore Ts'o #define IS_EXT2_SB(sb) (0) 1002035e776STheodore Ts'o #endif 1012035e776STheodore Ts'o 1022035e776STheodore Ts'o 103ba69f9abSJan Kara #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 104ba69f9abSJan Kara static struct file_system_type ext3_fs_type = { 105ba69f9abSJan Kara .owner = THIS_MODULE, 106ba69f9abSJan Kara .name = "ext3", 107152a0836SAl Viro .mount = ext4_mount, 108ba69f9abSJan Kara .kill_sb = kill_block_super, 109ba69f9abSJan Kara .fs_flags = FS_REQUIRES_DEV, 110ba69f9abSJan Kara }; 1117f78e035SEric W. Biederman MODULE_ALIAS_FS("ext3"); 112fa7614ddSEric W. Biederman MODULE_ALIAS("ext3"); 113ba69f9abSJan Kara #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 114ba69f9abSJan Kara #else 115ba69f9abSJan Kara #define IS_EXT3_SB(sb) (0) 116ba69f9abSJan Kara #endif 117bd81d8eeSLaurent Vivier 118d25425f8SDarrick J. Wong static int ext4_verify_csum_type(struct super_block *sb, 119d25425f8SDarrick J. Wong struct ext4_super_block *es) 120d25425f8SDarrick J. Wong { 121d25425f8SDarrick J. Wong if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, 122d25425f8SDarrick J. Wong EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 123d25425f8SDarrick J. Wong return 1; 124d25425f8SDarrick J. Wong 125d25425f8SDarrick J. Wong return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 126d25425f8SDarrick J. Wong } 127d25425f8SDarrick J. Wong 128a9c47317SDarrick J. Wong static __le32 ext4_superblock_csum(struct super_block *sb, 129a9c47317SDarrick J. Wong struct ext4_super_block *es) 130a9c47317SDarrick J. Wong { 131a9c47317SDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 132a9c47317SDarrick J. Wong int offset = offsetof(struct ext4_super_block, s_checksum); 133a9c47317SDarrick J. Wong __u32 csum; 134a9c47317SDarrick J. Wong 135a9c47317SDarrick J. Wong csum = ext4_chksum(sbi, ~0, (char *)es, offset); 136a9c47317SDarrick J. Wong 137a9c47317SDarrick J. Wong return cpu_to_le32(csum); 138a9c47317SDarrick J. Wong } 139a9c47317SDarrick J. Wong 140a9c47317SDarrick J. Wong int ext4_superblock_csum_verify(struct super_block *sb, 141a9c47317SDarrick J. Wong struct ext4_super_block *es) 142a9c47317SDarrick J. Wong { 143a9c47317SDarrick J. Wong if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, 144a9c47317SDarrick J. Wong EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 145a9c47317SDarrick J. Wong return 1; 146a9c47317SDarrick J. Wong 147a9c47317SDarrick J. Wong return es->s_checksum == ext4_superblock_csum(sb, es); 148a9c47317SDarrick J. Wong } 149a9c47317SDarrick J. Wong 15006db49e6STheodore Ts'o void ext4_superblock_csum_set(struct super_block *sb) 151a9c47317SDarrick J. Wong { 15206db49e6STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 15306db49e6STheodore Ts'o 154a9c47317SDarrick J. Wong if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, 155a9c47317SDarrick J. Wong EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 156a9c47317SDarrick J. Wong return; 157a9c47317SDarrick J. Wong 158a9c47317SDarrick J. Wong es->s_checksum = ext4_superblock_csum(sb, es); 159a9c47317SDarrick J. Wong } 160a9c47317SDarrick J. Wong 1619933fc0aSTheodore Ts'o void *ext4_kvmalloc(size_t size, gfp_t flags) 1629933fc0aSTheodore Ts'o { 1639933fc0aSTheodore Ts'o void *ret; 1649933fc0aSTheodore Ts'o 1659933fc0aSTheodore Ts'o ret = kmalloc(size, flags); 1669933fc0aSTheodore Ts'o if (!ret) 1679933fc0aSTheodore Ts'o ret = __vmalloc(size, flags, PAGE_KERNEL); 1689933fc0aSTheodore Ts'o return ret; 1699933fc0aSTheodore Ts'o } 1709933fc0aSTheodore Ts'o 1719933fc0aSTheodore Ts'o void *ext4_kvzalloc(size_t size, gfp_t flags) 1729933fc0aSTheodore Ts'o { 1739933fc0aSTheodore Ts'o void *ret; 1749933fc0aSTheodore Ts'o 175db9481c0SMathias Krause ret = kzalloc(size, flags); 1769933fc0aSTheodore Ts'o if (!ret) 1779933fc0aSTheodore Ts'o ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL); 1789933fc0aSTheodore Ts'o return ret; 1799933fc0aSTheodore Ts'o } 1809933fc0aSTheodore Ts'o 1819933fc0aSTheodore Ts'o void ext4_kvfree(void *ptr) 1829933fc0aSTheodore Ts'o { 1839933fc0aSTheodore Ts'o if (is_vmalloc_addr(ptr)) 1849933fc0aSTheodore Ts'o vfree(ptr); 1859933fc0aSTheodore Ts'o else 1869933fc0aSTheodore Ts'o kfree(ptr); 1879933fc0aSTheodore Ts'o 1889933fc0aSTheodore Ts'o } 1899933fc0aSTheodore Ts'o 1908fadc143SAlexandre Ratchov ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 1918fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 192bd81d8eeSLaurent Vivier { 1933a14589cSAneesh Kumar K.V return le32_to_cpu(bg->bg_block_bitmap_lo) | 1948fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 1958fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 196bd81d8eeSLaurent Vivier } 197bd81d8eeSLaurent Vivier 1988fadc143SAlexandre Ratchov ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 1998fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 200bd81d8eeSLaurent Vivier { 2015272f837SAneesh Kumar K.V return le32_to_cpu(bg->bg_inode_bitmap_lo) | 2028fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 2038fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 204bd81d8eeSLaurent Vivier } 205bd81d8eeSLaurent Vivier 2068fadc143SAlexandre Ratchov ext4_fsblk_t ext4_inode_table(struct super_block *sb, 2078fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 208bd81d8eeSLaurent Vivier { 2095272f837SAneesh Kumar K.V return le32_to_cpu(bg->bg_inode_table_lo) | 2108fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 2118fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 212bd81d8eeSLaurent Vivier } 213bd81d8eeSLaurent Vivier 214021b65bbSTheodore Ts'o __u32 ext4_free_group_clusters(struct super_block *sb, 215560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 216560671a0SAneesh Kumar K.V { 217560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_free_blocks_count_lo) | 218560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 219560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 220560671a0SAneesh Kumar K.V } 221560671a0SAneesh Kumar K.V 222560671a0SAneesh Kumar K.V __u32 ext4_free_inodes_count(struct super_block *sb, 223560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 224560671a0SAneesh Kumar K.V { 225560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_free_inodes_count_lo) | 226560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 227560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 228560671a0SAneesh Kumar K.V } 229560671a0SAneesh Kumar K.V 230560671a0SAneesh Kumar K.V __u32 ext4_used_dirs_count(struct super_block *sb, 231560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 232560671a0SAneesh Kumar K.V { 233560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_used_dirs_count_lo) | 234560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 235560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 236560671a0SAneesh Kumar K.V } 237560671a0SAneesh Kumar K.V 238560671a0SAneesh Kumar K.V __u32 ext4_itable_unused_count(struct super_block *sb, 239560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 240560671a0SAneesh Kumar K.V { 241560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_itable_unused_lo) | 242560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 243560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 244560671a0SAneesh Kumar K.V } 245560671a0SAneesh Kumar K.V 2468fadc143SAlexandre Ratchov void ext4_block_bitmap_set(struct super_block *sb, 2478fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 248bd81d8eeSLaurent Vivier { 2493a14589cSAneesh Kumar K.V bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 2508fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 2518fadc143SAlexandre Ratchov bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 252bd81d8eeSLaurent Vivier } 253bd81d8eeSLaurent Vivier 2548fadc143SAlexandre Ratchov void ext4_inode_bitmap_set(struct super_block *sb, 2558fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 256bd81d8eeSLaurent Vivier { 2575272f837SAneesh Kumar K.V bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 2588fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 2598fadc143SAlexandre Ratchov bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 260bd81d8eeSLaurent Vivier } 261bd81d8eeSLaurent Vivier 2628fadc143SAlexandre Ratchov void ext4_inode_table_set(struct super_block *sb, 2638fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 264bd81d8eeSLaurent Vivier { 2655272f837SAneesh Kumar K.V bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 2668fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 2678fadc143SAlexandre Ratchov bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 268bd81d8eeSLaurent Vivier } 269bd81d8eeSLaurent Vivier 270021b65bbSTheodore Ts'o void ext4_free_group_clusters_set(struct super_block *sb, 271560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 272560671a0SAneesh Kumar K.V { 273560671a0SAneesh Kumar K.V bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 274560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 275560671a0SAneesh Kumar K.V bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 276560671a0SAneesh Kumar K.V } 277560671a0SAneesh Kumar K.V 278560671a0SAneesh Kumar K.V void ext4_free_inodes_set(struct super_block *sb, 279560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 280560671a0SAneesh Kumar K.V { 281560671a0SAneesh Kumar K.V bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 282560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 283560671a0SAneesh Kumar K.V bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 284560671a0SAneesh Kumar K.V } 285560671a0SAneesh Kumar K.V 286560671a0SAneesh Kumar K.V void ext4_used_dirs_set(struct super_block *sb, 287560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 288560671a0SAneesh Kumar K.V { 289560671a0SAneesh Kumar K.V bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 290560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 291560671a0SAneesh Kumar K.V bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 292560671a0SAneesh Kumar K.V } 293560671a0SAneesh Kumar K.V 294560671a0SAneesh Kumar K.V void ext4_itable_unused_set(struct super_block *sb, 295560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 296560671a0SAneesh Kumar K.V { 297560671a0SAneesh Kumar K.V bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 298560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 299560671a0SAneesh Kumar K.V bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 300560671a0SAneesh Kumar K.V } 301560671a0SAneesh Kumar K.V 302d3d1faf6SCurt Wohlgemuth 3031c13d5c0STheodore Ts'o static void __save_error_info(struct super_block *sb, const char *func, 3041c13d5c0STheodore Ts'o unsigned int line) 3051c13d5c0STheodore Ts'o { 3061c13d5c0STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 3071c13d5c0STheodore Ts'o 3081c13d5c0STheodore Ts'o EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 3091c13d5c0STheodore Ts'o es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 3101c13d5c0STheodore Ts'o es->s_last_error_time = cpu_to_le32(get_seconds()); 3111c13d5c0STheodore Ts'o strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); 3121c13d5c0STheodore Ts'o es->s_last_error_line = cpu_to_le32(line); 3131c13d5c0STheodore Ts'o if (!es->s_first_error_time) { 3141c13d5c0STheodore Ts'o es->s_first_error_time = es->s_last_error_time; 3151c13d5c0STheodore Ts'o strncpy(es->s_first_error_func, func, 3161c13d5c0STheodore Ts'o sizeof(es->s_first_error_func)); 3171c13d5c0STheodore Ts'o es->s_first_error_line = cpu_to_le32(line); 3181c13d5c0STheodore Ts'o es->s_first_error_ino = es->s_last_error_ino; 3191c13d5c0STheodore Ts'o es->s_first_error_block = es->s_last_error_block; 3201c13d5c0STheodore Ts'o } 32166e61a9eSTheodore Ts'o /* 32266e61a9eSTheodore Ts'o * Start the daily error reporting function if it hasn't been 32366e61a9eSTheodore Ts'o * started already 32466e61a9eSTheodore Ts'o */ 32566e61a9eSTheodore Ts'o if (!es->s_error_count) 32666e61a9eSTheodore Ts'o mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); 327ba39ebb6SWei Yongjun le32_add_cpu(&es->s_error_count, 1); 3281c13d5c0STheodore Ts'o } 3291c13d5c0STheodore Ts'o 3301c13d5c0STheodore Ts'o static void save_error_info(struct super_block *sb, const char *func, 3311c13d5c0STheodore Ts'o unsigned int line) 3321c13d5c0STheodore Ts'o { 3331c13d5c0STheodore Ts'o __save_error_info(sb, func, line); 3341c13d5c0STheodore Ts'o ext4_commit_super(sb, 1); 3351c13d5c0STheodore Ts'o } 3361c13d5c0STheodore Ts'o 3377c2e7087STheodore Ts'o /* 3387c2e7087STheodore Ts'o * The del_gendisk() function uninitializes the disk-specific data 3397c2e7087STheodore Ts'o * structures, including the bdi structure, without telling anyone 3407c2e7087STheodore Ts'o * else. Once this happens, any attempt to call mark_buffer_dirty() 3417c2e7087STheodore Ts'o * (for example, by ext4_commit_super), will cause a kernel OOPS. 3427c2e7087STheodore Ts'o * This is a kludge to prevent these oops until we can put in a proper 3437c2e7087STheodore Ts'o * hook in del_gendisk() to inform the VFS and file system layers. 3447c2e7087STheodore Ts'o */ 3457c2e7087STheodore Ts'o static int block_device_ejected(struct super_block *sb) 3467c2e7087STheodore Ts'o { 3477c2e7087STheodore Ts'o struct inode *bd_inode = sb->s_bdev->bd_inode; 3487c2e7087STheodore Ts'o struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info; 3497c2e7087STheodore Ts'o 3507c2e7087STheodore Ts'o return bdi->dev == NULL; 3517c2e7087STheodore Ts'o } 3527c2e7087STheodore Ts'o 35318aadd47SBobi Jam static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 35418aadd47SBobi Jam { 35518aadd47SBobi Jam struct super_block *sb = journal->j_private; 35618aadd47SBobi Jam struct ext4_sb_info *sbi = EXT4_SB(sb); 35718aadd47SBobi Jam int error = is_journal_aborted(journal); 3585d3ee208SDmitry Monakhov struct ext4_journal_cb_entry *jce; 35918aadd47SBobi Jam 3605d3ee208SDmitry Monakhov BUG_ON(txn->t_state == T_FINISHED); 36118aadd47SBobi Jam spin_lock(&sbi->s_md_lock); 3625d3ee208SDmitry Monakhov while (!list_empty(&txn->t_private_list)) { 3635d3ee208SDmitry Monakhov jce = list_entry(txn->t_private_list.next, 3645d3ee208SDmitry Monakhov struct ext4_journal_cb_entry, jce_list); 36518aadd47SBobi Jam list_del_init(&jce->jce_list); 36618aadd47SBobi Jam spin_unlock(&sbi->s_md_lock); 36718aadd47SBobi Jam jce->jce_func(sb, jce, error); 36818aadd47SBobi Jam spin_lock(&sbi->s_md_lock); 36918aadd47SBobi Jam } 37018aadd47SBobi Jam spin_unlock(&sbi->s_md_lock); 37118aadd47SBobi Jam } 3721c13d5c0STheodore Ts'o 373ac27a0ecSDave Kleikamp /* Deal with the reporting of failure conditions on a filesystem such as 374ac27a0ecSDave Kleikamp * inconsistencies detected or read IO failures. 375ac27a0ecSDave Kleikamp * 376ac27a0ecSDave Kleikamp * On ext2, we can store the error state of the filesystem in the 377617ba13bSMingming Cao * superblock. That is not possible on ext4, because we may have other 378ac27a0ecSDave Kleikamp * write ordering constraints on the superblock which prevent us from 379ac27a0ecSDave Kleikamp * writing it out straight away; and given that the journal is about to 380ac27a0ecSDave Kleikamp * be aborted, we can't rely on the current, or future, transactions to 381ac27a0ecSDave Kleikamp * write out the superblock safely. 382ac27a0ecSDave Kleikamp * 383dab291afSMingming Cao * We'll just use the jbd2_journal_abort() error code to record an error in 384d6b198bcSThadeu Lima de Souza Cascardo * the journal instead. On recovery, the journal will complain about 385ac27a0ecSDave Kleikamp * that error until we've noted it down and cleared it. 386ac27a0ecSDave Kleikamp */ 387ac27a0ecSDave Kleikamp 388617ba13bSMingming Cao static void ext4_handle_error(struct super_block *sb) 389ac27a0ecSDave Kleikamp { 390ac27a0ecSDave Kleikamp if (sb->s_flags & MS_RDONLY) 391ac27a0ecSDave Kleikamp return; 392ac27a0ecSDave Kleikamp 393ac27a0ecSDave Kleikamp if (!test_opt(sb, ERRORS_CONT)) { 394617ba13bSMingming Cao journal_t *journal = EXT4_SB(sb)->s_journal; 395ac27a0ecSDave Kleikamp 3964ab2f15bSTheodore Ts'o EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 397ac27a0ecSDave Kleikamp if (journal) 398dab291afSMingming Cao jbd2_journal_abort(journal, -EIO); 399ac27a0ecSDave Kleikamp } 400ac27a0ecSDave Kleikamp if (test_opt(sb, ERRORS_RO)) { 401b31e1552SEric Sandeen ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 4024418e141SDmitry Monakhov /* 4034418e141SDmitry Monakhov * Make sure updated value of ->s_mount_flags will be visible 4044418e141SDmitry Monakhov * before ->s_flags update 4054418e141SDmitry Monakhov */ 4064418e141SDmitry Monakhov smp_wmb(); 407ac27a0ecSDave Kleikamp sb->s_flags |= MS_RDONLY; 408ac27a0ecSDave Kleikamp } 409ac27a0ecSDave Kleikamp if (test_opt(sb, ERRORS_PANIC)) 410617ba13bSMingming Cao panic("EXT4-fs (device %s): panic forced after error\n", 411ac27a0ecSDave Kleikamp sb->s_id); 412ac27a0ecSDave Kleikamp } 413ac27a0ecSDave Kleikamp 41412062dddSEric Sandeen void __ext4_error(struct super_block *sb, const char *function, 415c398eda0STheodore Ts'o unsigned int line, const char *fmt, ...) 416ac27a0ecSDave Kleikamp { 4170ff2ea7dSJoe Perches struct va_format vaf; 418ac27a0ecSDave Kleikamp va_list args; 419ac27a0ecSDave Kleikamp 420ac27a0ecSDave Kleikamp va_start(args, fmt); 4210ff2ea7dSJoe Perches vaf.fmt = fmt; 4220ff2ea7dSJoe Perches vaf.va = &args; 4230ff2ea7dSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 4240ff2ea7dSJoe Perches sb->s_id, function, line, current->comm, &vaf); 425ac27a0ecSDave Kleikamp va_end(args); 426f3fc0210STheodore Ts'o save_error_info(sb, function, line); 427ac27a0ecSDave Kleikamp 428617ba13bSMingming Cao ext4_handle_error(sb); 429ac27a0ecSDave Kleikamp } 430ac27a0ecSDave Kleikamp 431e7c96e8eSJoe Perches void __ext4_error_inode(struct inode *inode, const char *function, 432c398eda0STheodore Ts'o unsigned int line, ext4_fsblk_t block, 433273df556SFrank Mayhar const char *fmt, ...) 434273df556SFrank Mayhar { 435273df556SFrank Mayhar va_list args; 436f7c21177STheodore Ts'o struct va_format vaf; 4371c13d5c0STheodore Ts'o struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es; 438273df556SFrank Mayhar 4391c13d5c0STheodore Ts'o es->s_last_error_ino = cpu_to_le32(inode->i_ino); 4401c13d5c0STheodore Ts'o es->s_last_error_block = cpu_to_le64(block); 4411c13d5c0STheodore Ts'o save_error_info(inode->i_sb, function, line); 442273df556SFrank Mayhar va_start(args, fmt); 443f7c21177STheodore Ts'o vaf.fmt = fmt; 444f7c21177STheodore Ts'o vaf.va = &args; 445c398eda0STheodore Ts'o if (block) 446d9ee81daSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 447d9ee81daSJoe Perches "inode #%lu: block %llu: comm %s: %pV\n", 448d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 449d9ee81daSJoe Perches block, current->comm, &vaf); 450d9ee81daSJoe Perches else 451d9ee81daSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 452d9ee81daSJoe Perches "inode #%lu: comm %s: %pV\n", 453d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 454d9ee81daSJoe Perches current->comm, &vaf); 455273df556SFrank Mayhar va_end(args); 456273df556SFrank Mayhar 457273df556SFrank Mayhar ext4_handle_error(inode->i_sb); 458273df556SFrank Mayhar } 459273df556SFrank Mayhar 460e7c96e8eSJoe Perches void __ext4_error_file(struct file *file, const char *function, 461f7c21177STheodore Ts'o unsigned int line, ext4_fsblk_t block, 462f7c21177STheodore Ts'o const char *fmt, ...) 463273df556SFrank Mayhar { 464273df556SFrank Mayhar va_list args; 465f7c21177STheodore Ts'o struct va_format vaf; 4661c13d5c0STheodore Ts'o struct ext4_super_block *es; 467496ad9aaSAl Viro struct inode *inode = file_inode(file); 468273df556SFrank Mayhar char pathname[80], *path; 469273df556SFrank Mayhar 4701c13d5c0STheodore Ts'o es = EXT4_SB(inode->i_sb)->s_es; 4711c13d5c0STheodore Ts'o es->s_last_error_ino = cpu_to_le32(inode->i_ino); 4721c13d5c0STheodore Ts'o save_error_info(inode->i_sb, function, line); 473273df556SFrank Mayhar path = d_path(&(file->f_path), pathname, sizeof(pathname)); 474f9a62d09SDan Carpenter if (IS_ERR(path)) 475273df556SFrank Mayhar path = "(unknown)"; 476f7c21177STheodore Ts'o va_start(args, fmt); 477f7c21177STheodore Ts'o vaf.fmt = fmt; 478f7c21177STheodore Ts'o vaf.va = &args; 479d9ee81daSJoe Perches if (block) 480d9ee81daSJoe Perches printk(KERN_CRIT 481d9ee81daSJoe Perches "EXT4-fs error (device %s): %s:%d: inode #%lu: " 482d9ee81daSJoe Perches "block %llu: comm %s: path %s: %pV\n", 483d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 484d9ee81daSJoe Perches block, current->comm, path, &vaf); 485d9ee81daSJoe Perches else 486d9ee81daSJoe Perches printk(KERN_CRIT 487d9ee81daSJoe Perches "EXT4-fs error (device %s): %s:%d: inode #%lu: " 488d9ee81daSJoe Perches "comm %s: path %s: %pV\n", 489d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 490d9ee81daSJoe Perches current->comm, path, &vaf); 491273df556SFrank Mayhar va_end(args); 492273df556SFrank Mayhar 493273df556SFrank Mayhar ext4_handle_error(inode->i_sb); 494273df556SFrank Mayhar } 495273df556SFrank Mayhar 496722887ddSTheodore Ts'o const char *ext4_decode_error(struct super_block *sb, int errno, 497ac27a0ecSDave Kleikamp char nbuf[16]) 498ac27a0ecSDave Kleikamp { 499ac27a0ecSDave Kleikamp char *errstr = NULL; 500ac27a0ecSDave Kleikamp 501ac27a0ecSDave Kleikamp switch (errno) { 502ac27a0ecSDave Kleikamp case -EIO: 503ac27a0ecSDave Kleikamp errstr = "IO failure"; 504ac27a0ecSDave Kleikamp break; 505ac27a0ecSDave Kleikamp case -ENOMEM: 506ac27a0ecSDave Kleikamp errstr = "Out of memory"; 507ac27a0ecSDave Kleikamp break; 508ac27a0ecSDave Kleikamp case -EROFS: 50978f1ddbbSTheodore Ts'o if (!sb || (EXT4_SB(sb)->s_journal && 51078f1ddbbSTheodore Ts'o EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 511ac27a0ecSDave Kleikamp errstr = "Journal has aborted"; 512ac27a0ecSDave Kleikamp else 513ac27a0ecSDave Kleikamp errstr = "Readonly filesystem"; 514ac27a0ecSDave Kleikamp break; 515ac27a0ecSDave Kleikamp default: 516ac27a0ecSDave Kleikamp /* If the caller passed in an extra buffer for unknown 517ac27a0ecSDave Kleikamp * errors, textualise them now. Else we just return 518ac27a0ecSDave Kleikamp * NULL. */ 519ac27a0ecSDave Kleikamp if (nbuf) { 520ac27a0ecSDave Kleikamp /* Check for truncated error codes... */ 521ac27a0ecSDave Kleikamp if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 522ac27a0ecSDave Kleikamp errstr = nbuf; 523ac27a0ecSDave Kleikamp } 524ac27a0ecSDave Kleikamp break; 525ac27a0ecSDave Kleikamp } 526ac27a0ecSDave Kleikamp 527ac27a0ecSDave Kleikamp return errstr; 528ac27a0ecSDave Kleikamp } 529ac27a0ecSDave Kleikamp 530617ba13bSMingming Cao /* __ext4_std_error decodes expected errors from journaling functions 531ac27a0ecSDave Kleikamp * automatically and invokes the appropriate error response. */ 532ac27a0ecSDave Kleikamp 533c398eda0STheodore Ts'o void __ext4_std_error(struct super_block *sb, const char *function, 534c398eda0STheodore Ts'o unsigned int line, int errno) 535ac27a0ecSDave Kleikamp { 536ac27a0ecSDave Kleikamp char nbuf[16]; 537ac27a0ecSDave Kleikamp const char *errstr; 538ac27a0ecSDave Kleikamp 539ac27a0ecSDave Kleikamp /* Special case: if the error is EROFS, and we're not already 540ac27a0ecSDave Kleikamp * inside a transaction, then there's really no point in logging 541ac27a0ecSDave Kleikamp * an error. */ 542ac27a0ecSDave Kleikamp if (errno == -EROFS && journal_current_handle() == NULL && 543ac27a0ecSDave Kleikamp (sb->s_flags & MS_RDONLY)) 544ac27a0ecSDave Kleikamp return; 545ac27a0ecSDave Kleikamp 546617ba13bSMingming Cao errstr = ext4_decode_error(sb, errno, nbuf); 547c398eda0STheodore Ts'o printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 548c398eda0STheodore Ts'o sb->s_id, function, line, errstr); 5491c13d5c0STheodore Ts'o save_error_info(sb, function, line); 550ac27a0ecSDave Kleikamp 551617ba13bSMingming Cao ext4_handle_error(sb); 552ac27a0ecSDave Kleikamp } 553ac27a0ecSDave Kleikamp 554ac27a0ecSDave Kleikamp /* 555617ba13bSMingming Cao * ext4_abort is a much stronger failure handler than ext4_error. The 556ac27a0ecSDave Kleikamp * abort function may be used to deal with unrecoverable failures such 557ac27a0ecSDave Kleikamp * as journal IO errors or ENOMEM at a critical moment in log management. 558ac27a0ecSDave Kleikamp * 559ac27a0ecSDave Kleikamp * We unconditionally force the filesystem into an ABORT|READONLY state, 560ac27a0ecSDave Kleikamp * unless the error response on the fs has been set to panic in which 561ac27a0ecSDave Kleikamp * case we take the easy way out and panic immediately. 562ac27a0ecSDave Kleikamp */ 563ac27a0ecSDave Kleikamp 564c67d859eSTheodore Ts'o void __ext4_abort(struct super_block *sb, const char *function, 565c398eda0STheodore Ts'o unsigned int line, const char *fmt, ...) 566ac27a0ecSDave Kleikamp { 567ac27a0ecSDave Kleikamp va_list args; 568ac27a0ecSDave Kleikamp 5691c13d5c0STheodore Ts'o save_error_info(sb, function, line); 570ac27a0ecSDave Kleikamp va_start(args, fmt); 571c398eda0STheodore Ts'o printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id, 572c398eda0STheodore Ts'o function, line); 573ac27a0ecSDave Kleikamp vprintk(fmt, args); 574ac27a0ecSDave Kleikamp printk("\n"); 575ac27a0ecSDave Kleikamp va_end(args); 576ac27a0ecSDave Kleikamp 5771c13d5c0STheodore Ts'o if ((sb->s_flags & MS_RDONLY) == 0) { 578b31e1552SEric Sandeen ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 5794ab2f15bSTheodore Ts'o EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 5804418e141SDmitry Monakhov /* 5814418e141SDmitry Monakhov * Make sure updated value of ->s_mount_flags will be visible 5824418e141SDmitry Monakhov * before ->s_flags update 5834418e141SDmitry Monakhov */ 5844418e141SDmitry Monakhov smp_wmb(); 5854418e141SDmitry Monakhov sb->s_flags |= MS_RDONLY; 586ef2cabf7SHidehiro Kawai if (EXT4_SB(sb)->s_journal) 587dab291afSMingming Cao jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); 5881c13d5c0STheodore Ts'o save_error_info(sb, function, line); 5891c13d5c0STheodore Ts'o } 5901c13d5c0STheodore Ts'o if (test_opt(sb, ERRORS_PANIC)) 5911c13d5c0STheodore Ts'o panic("EXT4-fs panic from previous error\n"); 592ac27a0ecSDave Kleikamp } 593ac27a0ecSDave Kleikamp 594e7c96e8eSJoe Perches void __ext4_msg(struct super_block *sb, 595e7c96e8eSJoe Perches const char *prefix, const char *fmt, ...) 596b31e1552SEric Sandeen { 5970ff2ea7dSJoe Perches struct va_format vaf; 598b31e1552SEric Sandeen va_list args; 599b31e1552SEric Sandeen 600b31e1552SEric Sandeen va_start(args, fmt); 6010ff2ea7dSJoe Perches vaf.fmt = fmt; 6020ff2ea7dSJoe Perches vaf.va = &args; 6030ff2ea7dSJoe Perches printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 604b31e1552SEric Sandeen va_end(args); 605b31e1552SEric Sandeen } 606b31e1552SEric Sandeen 60712062dddSEric Sandeen void __ext4_warning(struct super_block *sb, const char *function, 608c398eda0STheodore Ts'o unsigned int line, const char *fmt, ...) 609ac27a0ecSDave Kleikamp { 6100ff2ea7dSJoe Perches struct va_format vaf; 611ac27a0ecSDave Kleikamp va_list args; 612ac27a0ecSDave Kleikamp 613ac27a0ecSDave Kleikamp va_start(args, fmt); 6140ff2ea7dSJoe Perches vaf.fmt = fmt; 6150ff2ea7dSJoe Perches vaf.va = &args; 6160ff2ea7dSJoe Perches printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 6170ff2ea7dSJoe Perches sb->s_id, function, line, &vaf); 618ac27a0ecSDave Kleikamp va_end(args); 619ac27a0ecSDave Kleikamp } 620ac27a0ecSDave Kleikamp 621e29136f8STheodore Ts'o void __ext4_grp_locked_error(const char *function, unsigned int line, 622e29136f8STheodore Ts'o struct super_block *sb, ext4_group_t grp, 623e29136f8STheodore Ts'o unsigned long ino, ext4_fsblk_t block, 624e29136f8STheodore Ts'o const char *fmt, ...) 6255d1b1b3fSAneesh Kumar K.V __releases(bitlock) 6265d1b1b3fSAneesh Kumar K.V __acquires(bitlock) 6275d1b1b3fSAneesh Kumar K.V { 6280ff2ea7dSJoe Perches struct va_format vaf; 6295d1b1b3fSAneesh Kumar K.V va_list args; 6305d1b1b3fSAneesh Kumar K.V struct ext4_super_block *es = EXT4_SB(sb)->s_es; 6315d1b1b3fSAneesh Kumar K.V 6321c13d5c0STheodore Ts'o es->s_last_error_ino = cpu_to_le32(ino); 6331c13d5c0STheodore Ts'o es->s_last_error_block = cpu_to_le64(block); 6341c13d5c0STheodore Ts'o __save_error_info(sb, function, line); 6350ff2ea7dSJoe Perches 6365d1b1b3fSAneesh Kumar K.V va_start(args, fmt); 6370ff2ea7dSJoe Perches 6380ff2ea7dSJoe Perches vaf.fmt = fmt; 6390ff2ea7dSJoe Perches vaf.va = &args; 64021149d61SRobin Dong printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 641e29136f8STheodore Ts'o sb->s_id, function, line, grp); 642e29136f8STheodore Ts'o if (ino) 6430ff2ea7dSJoe Perches printk(KERN_CONT "inode %lu: ", ino); 644e29136f8STheodore Ts'o if (block) 6450ff2ea7dSJoe Perches printk(KERN_CONT "block %llu:", (unsigned long long) block); 6460ff2ea7dSJoe Perches printk(KERN_CONT "%pV\n", &vaf); 6475d1b1b3fSAneesh Kumar K.V va_end(args); 6485d1b1b3fSAneesh Kumar K.V 6495d1b1b3fSAneesh Kumar K.V if (test_opt(sb, ERRORS_CONT)) { 650e2d67052STheodore Ts'o ext4_commit_super(sb, 0); 6515d1b1b3fSAneesh Kumar K.V return; 6525d1b1b3fSAneesh Kumar K.V } 6531c13d5c0STheodore Ts'o 6545d1b1b3fSAneesh Kumar K.V ext4_unlock_group(sb, grp); 6555d1b1b3fSAneesh Kumar K.V ext4_handle_error(sb); 6565d1b1b3fSAneesh Kumar K.V /* 6575d1b1b3fSAneesh Kumar K.V * We only get here in the ERRORS_RO case; relocking the group 6585d1b1b3fSAneesh Kumar K.V * may be dangerous, but nothing bad will happen since the 6595d1b1b3fSAneesh Kumar K.V * filesystem will have already been marked read/only and the 6605d1b1b3fSAneesh Kumar K.V * journal has been aborted. We return 1 as a hint to callers 6615d1b1b3fSAneesh Kumar K.V * who might what to use the return value from 66225985edcSLucas De Marchi * ext4_grp_locked_error() to distinguish between the 6635d1b1b3fSAneesh Kumar K.V * ERRORS_CONT and ERRORS_RO case, and perhaps return more 6645d1b1b3fSAneesh Kumar K.V * aggressively from the ext4 function in question, with a 6655d1b1b3fSAneesh Kumar K.V * more appropriate error code. 6665d1b1b3fSAneesh Kumar K.V */ 6675d1b1b3fSAneesh Kumar K.V ext4_lock_group(sb, grp); 6685d1b1b3fSAneesh Kumar K.V return; 6695d1b1b3fSAneesh Kumar K.V } 6705d1b1b3fSAneesh Kumar K.V 671617ba13bSMingming Cao void ext4_update_dynamic_rev(struct super_block *sb) 672ac27a0ecSDave Kleikamp { 673617ba13bSMingming Cao struct ext4_super_block *es = EXT4_SB(sb)->s_es; 674ac27a0ecSDave Kleikamp 675617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 676ac27a0ecSDave Kleikamp return; 677ac27a0ecSDave Kleikamp 67812062dddSEric Sandeen ext4_warning(sb, 679ac27a0ecSDave Kleikamp "updating to rev %d because of new feature flag, " 680ac27a0ecSDave Kleikamp "running e2fsck is recommended", 681617ba13bSMingming Cao EXT4_DYNAMIC_REV); 682ac27a0ecSDave Kleikamp 683617ba13bSMingming Cao es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 684617ba13bSMingming Cao es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 685617ba13bSMingming Cao es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 686ac27a0ecSDave Kleikamp /* leave es->s_feature_*compat flags alone */ 687ac27a0ecSDave Kleikamp /* es->s_uuid will be set by e2fsck if empty */ 688ac27a0ecSDave Kleikamp 689ac27a0ecSDave Kleikamp /* 690ac27a0ecSDave Kleikamp * The rest of the superblock fields should be zero, and if not it 691ac27a0ecSDave Kleikamp * means they are likely already in use, so leave them alone. We 692ac27a0ecSDave Kleikamp * can leave it up to e2fsck to clean up any inconsistencies there. 693ac27a0ecSDave Kleikamp */ 694ac27a0ecSDave Kleikamp } 695ac27a0ecSDave Kleikamp 696ac27a0ecSDave Kleikamp /* 697ac27a0ecSDave Kleikamp * Open the external journal device 698ac27a0ecSDave Kleikamp */ 699b31e1552SEric Sandeen static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 700ac27a0ecSDave Kleikamp { 701ac27a0ecSDave Kleikamp struct block_device *bdev; 702ac27a0ecSDave Kleikamp char b[BDEVNAME_SIZE]; 703ac27a0ecSDave Kleikamp 704d4d77629STejun Heo bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 705ac27a0ecSDave Kleikamp if (IS_ERR(bdev)) 706ac27a0ecSDave Kleikamp goto fail; 707ac27a0ecSDave Kleikamp return bdev; 708ac27a0ecSDave Kleikamp 709ac27a0ecSDave Kleikamp fail: 710b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld", 711ac27a0ecSDave Kleikamp __bdevname(dev, b), PTR_ERR(bdev)); 712ac27a0ecSDave Kleikamp return NULL; 713ac27a0ecSDave Kleikamp } 714ac27a0ecSDave Kleikamp 715ac27a0ecSDave Kleikamp /* 716ac27a0ecSDave Kleikamp * Release the journal device 717ac27a0ecSDave Kleikamp */ 7184385bab1SAl Viro static void ext4_blkdev_put(struct block_device *bdev) 719ac27a0ecSDave Kleikamp { 7204385bab1SAl Viro blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 721ac27a0ecSDave Kleikamp } 722ac27a0ecSDave Kleikamp 7234385bab1SAl Viro static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 724ac27a0ecSDave Kleikamp { 725ac27a0ecSDave Kleikamp struct block_device *bdev; 726ac27a0ecSDave Kleikamp bdev = sbi->journal_bdev; 727ac27a0ecSDave Kleikamp if (bdev) { 7284385bab1SAl Viro ext4_blkdev_put(bdev); 729ac27a0ecSDave Kleikamp sbi->journal_bdev = NULL; 730ac27a0ecSDave Kleikamp } 731ac27a0ecSDave Kleikamp } 732ac27a0ecSDave Kleikamp 733ac27a0ecSDave Kleikamp static inline struct inode *orphan_list_entry(struct list_head *l) 734ac27a0ecSDave Kleikamp { 735617ba13bSMingming Cao return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 736ac27a0ecSDave Kleikamp } 737ac27a0ecSDave Kleikamp 738617ba13bSMingming Cao static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 739ac27a0ecSDave Kleikamp { 740ac27a0ecSDave Kleikamp struct list_head *l; 741ac27a0ecSDave Kleikamp 742b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 743ac27a0ecSDave Kleikamp le32_to_cpu(sbi->s_es->s_last_orphan)); 744ac27a0ecSDave Kleikamp 745ac27a0ecSDave Kleikamp printk(KERN_ERR "sb_info orphan list:\n"); 746ac27a0ecSDave Kleikamp list_for_each(l, &sbi->s_orphan) { 747ac27a0ecSDave Kleikamp struct inode *inode = orphan_list_entry(l); 748ac27a0ecSDave Kleikamp printk(KERN_ERR " " 749ac27a0ecSDave Kleikamp "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 750ac27a0ecSDave Kleikamp inode->i_sb->s_id, inode->i_ino, inode, 751ac27a0ecSDave Kleikamp inode->i_mode, inode->i_nlink, 752ac27a0ecSDave Kleikamp NEXT_ORPHAN(inode)); 753ac27a0ecSDave Kleikamp } 754ac27a0ecSDave Kleikamp } 755ac27a0ecSDave Kleikamp 756617ba13bSMingming Cao static void ext4_put_super(struct super_block *sb) 757ac27a0ecSDave Kleikamp { 758617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 759617ba13bSMingming Cao struct ext4_super_block *es = sbi->s_es; 760ef2cabf7SHidehiro Kawai int i, err; 761ac27a0ecSDave Kleikamp 762857ac889SLukas Czerner ext4_unregister_li_request(sb); 763e0ccfd95SChristoph Hellwig dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED); 764e0ccfd95SChristoph Hellwig 7652e8fa54eSJan Kara flush_workqueue(sbi->unrsv_conversion_wq); 7662e8fa54eSJan Kara flush_workqueue(sbi->rsv_conversion_wq); 7672e8fa54eSJan Kara destroy_workqueue(sbi->unrsv_conversion_wq); 7682e8fa54eSJan Kara destroy_workqueue(sbi->rsv_conversion_wq); 7694c0425ffSMingming Cao 7700390131bSFrank Mayhar if (sbi->s_journal) { 771ef2cabf7SHidehiro Kawai err = jbd2_journal_destroy(sbi->s_journal); 77247b4a50bSJan Kara sbi->s_journal = NULL; 773ef2cabf7SHidehiro Kawai if (err < 0) 774c67d859eSTheodore Ts'o ext4_abort(sb, "Couldn't clean up the journal"); 7750390131bSFrank Mayhar } 776d4edac31SJosef Bacik 777d3922a77SZheng Liu ext4_es_unregister_shrinker(sbi); 778a1c6c569SSergey Senozhatsky del_timer(&sbi->s_err_report); 779d4edac31SJosef Bacik ext4_release_system_zone(sb); 780d4edac31SJosef Bacik ext4_mb_release(sb); 781d4edac31SJosef Bacik ext4_ext_release(sb); 782d4edac31SJosef Bacik ext4_xattr_put_super(sb); 783d4edac31SJosef Bacik 784ac27a0ecSDave Kleikamp if (!(sb->s_flags & MS_RDONLY)) { 785617ba13bSMingming Cao EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 786ac27a0ecSDave Kleikamp es->s_state = cpu_to_le16(sbi->s_mount_state); 787ac27a0ecSDave Kleikamp } 78858c5873aSArtem Bityutskiy if (!(sb->s_flags & MS_RDONLY)) 789a8e25a83SArtem Bityutskiy ext4_commit_super(sb, 1); 790a8e25a83SArtem Bityutskiy 791240799cdSTheodore Ts'o if (sbi->s_proc) { 79266acdcf4STheodore Ts'o remove_proc_entry("options", sbi->s_proc); 7939f6200bbSTheodore Ts'o remove_proc_entry(sb->s_id, ext4_proc_root); 794240799cdSTheodore Ts'o } 7953197ebdbSTheodore Ts'o kobject_del(&sbi->s_kobj); 796ac27a0ecSDave Kleikamp 797ac27a0ecSDave Kleikamp for (i = 0; i < sbi->s_gdb_count; i++) 798ac27a0ecSDave Kleikamp brelse(sbi->s_group_desc[i]); 799f18a5f21STheodore Ts'o ext4_kvfree(sbi->s_group_desc); 8009933fc0aSTheodore Ts'o ext4_kvfree(sbi->s_flex_groups); 80157042651STheodore Ts'o percpu_counter_destroy(&sbi->s_freeclusters_counter); 802ac27a0ecSDave Kleikamp percpu_counter_destroy(&sbi->s_freeinodes_counter); 803ac27a0ecSDave Kleikamp percpu_counter_destroy(&sbi->s_dirs_counter); 80457042651STheodore Ts'o percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 8051ac6466fSTheodore Ts'o percpu_counter_destroy(&sbi->s_extent_cache_cnt); 806ac27a0ecSDave Kleikamp brelse(sbi->s_sbh); 807ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 808ac27a0ecSDave Kleikamp for (i = 0; i < MAXQUOTAS; i++) 809ac27a0ecSDave Kleikamp kfree(sbi->s_qf_names[i]); 810ac27a0ecSDave Kleikamp #endif 811ac27a0ecSDave Kleikamp 812ac27a0ecSDave Kleikamp /* Debugging code just in case the in-memory inode orphan list 813ac27a0ecSDave Kleikamp * isn't empty. The on-disk one can be non-empty if we've 814ac27a0ecSDave Kleikamp * detected an error and taken the fs readonly, but the 815ac27a0ecSDave Kleikamp * in-memory list had better be clean by this point. */ 816ac27a0ecSDave Kleikamp if (!list_empty(&sbi->s_orphan)) 817ac27a0ecSDave Kleikamp dump_orphan_list(sb, sbi); 818ac27a0ecSDave Kleikamp J_ASSERT(list_empty(&sbi->s_orphan)); 819ac27a0ecSDave Kleikamp 820f98393a6SPeter Zijlstra invalidate_bdev(sb->s_bdev); 821ac27a0ecSDave Kleikamp if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) { 822ac27a0ecSDave Kleikamp /* 823ac27a0ecSDave Kleikamp * Invalidate the journal device's buffers. We don't want them 824ac27a0ecSDave Kleikamp * floating about in memory - the physical journal device may 825ac27a0ecSDave Kleikamp * hotswapped, and it breaks the `ro-after' testing code. 826ac27a0ecSDave Kleikamp */ 827ac27a0ecSDave Kleikamp sync_blockdev(sbi->journal_bdev); 828f98393a6SPeter Zijlstra invalidate_bdev(sbi->journal_bdev); 829617ba13bSMingming Cao ext4_blkdev_remove(sbi); 830ac27a0ecSDave Kleikamp } 831c5e06d10SJohann Lombardi if (sbi->s_mmp_tsk) 832c5e06d10SJohann Lombardi kthread_stop(sbi->s_mmp_tsk); 833ac27a0ecSDave Kleikamp sb->s_fs_info = NULL; 8343197ebdbSTheodore Ts'o /* 8353197ebdbSTheodore Ts'o * Now that we are completely done shutting down the 8363197ebdbSTheodore Ts'o * superblock, we need to actually destroy the kobject. 8373197ebdbSTheodore Ts'o */ 8383197ebdbSTheodore Ts'o kobject_put(&sbi->s_kobj); 8393197ebdbSTheodore Ts'o wait_for_completion(&sbi->s_kobj_unregister); 8400441984aSDarrick J. Wong if (sbi->s_chksum_driver) 8410441984aSDarrick J. Wong crypto_free_shash(sbi->s_chksum_driver); 842705895b6SPekka Enberg kfree(sbi->s_blockgroup_lock); 843ac27a0ecSDave Kleikamp kfree(sbi); 844ac27a0ecSDave Kleikamp } 845ac27a0ecSDave Kleikamp 846e18b890bSChristoph Lameter static struct kmem_cache *ext4_inode_cachep; 847ac27a0ecSDave Kleikamp 848ac27a0ecSDave Kleikamp /* 849ac27a0ecSDave Kleikamp * Called inside transaction, so use GFP_NOFS 850ac27a0ecSDave Kleikamp */ 851617ba13bSMingming Cao static struct inode *ext4_alloc_inode(struct super_block *sb) 852ac27a0ecSDave Kleikamp { 853617ba13bSMingming Cao struct ext4_inode_info *ei; 854ac27a0ecSDave Kleikamp 855e6b4f8daSChristoph Lameter ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 856ac27a0ecSDave Kleikamp if (!ei) 857ac27a0ecSDave Kleikamp return NULL; 8580b8e58a1SAndreas Dilger 859ac27a0ecSDave Kleikamp ei->vfs_inode.i_version = 1; 860c9de560dSAlex Tomas INIT_LIST_HEAD(&ei->i_prealloc_list); 861c9de560dSAlex Tomas spin_lock_init(&ei->i_prealloc_lock); 8629a26b661SZheng Liu ext4_es_init_tree(&ei->i_es_tree); 8639a26b661SZheng Liu rwlock_init(&ei->i_es_lock); 86474cd15cdSZheng Liu INIT_LIST_HEAD(&ei->i_es_lru); 86574cd15cdSZheng Liu ei->i_es_lru_nr = 0; 866d3922a77SZheng Liu ei->i_touch_when = 0; 867d2a17637SMingming Cao ei->i_reserved_data_blocks = 0; 868d2a17637SMingming Cao ei->i_reserved_meta_blocks = 0; 869d2a17637SMingming Cao ei->i_allocated_meta_blocks = 0; 8709d0be502STheodore Ts'o ei->i_da_metadata_calc_len = 0; 8717e731bc9STheodore Ts'o ei->i_da_metadata_calc_last_lblock = 0; 872d2a17637SMingming Cao spin_lock_init(&(ei->i_block_reservation_lock)); 873a9e7f447SDmitry Monakhov #ifdef CONFIG_QUOTA 874a9e7f447SDmitry Monakhov ei->i_reserved_quota = 0; 875a9e7f447SDmitry Monakhov #endif 8768aefcd55STheodore Ts'o ei->jinode = NULL; 8772e8fa54eSJan Kara INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 8782e8fa54eSJan Kara INIT_LIST_HEAD(&ei->i_unrsv_conversion_list); 879744692dcSJiaying Zhang spin_lock_init(&ei->i_completed_io_lock); 880b436b9beSJan Kara ei->i_sync_tid = 0; 881b436b9beSJan Kara ei->i_datasync_tid = 0; 882f7ad6d2eSTheodore Ts'o atomic_set(&ei->i_ioend_count, 0); 883e27f41e1SDmitry Monakhov atomic_set(&ei->i_unwritten, 0); 8842e8fa54eSJan Kara INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 8852e8fa54eSJan Kara INIT_WORK(&ei->i_unrsv_conversion_work, ext4_end_io_unrsv_work); 8860b8e58a1SAndreas Dilger 887ac27a0ecSDave Kleikamp return &ei->vfs_inode; 888ac27a0ecSDave Kleikamp } 889ac27a0ecSDave Kleikamp 8907ff9c073STheodore Ts'o static int ext4_drop_inode(struct inode *inode) 8917ff9c073STheodore Ts'o { 8927ff9c073STheodore Ts'o int drop = generic_drop_inode(inode); 8937ff9c073STheodore Ts'o 8947ff9c073STheodore Ts'o trace_ext4_drop_inode(inode, drop); 8957ff9c073STheodore Ts'o return drop; 8967ff9c073STheodore Ts'o } 8977ff9c073STheodore Ts'o 898fa0d7e3dSNick Piggin static void ext4_i_callback(struct rcu_head *head) 899fa0d7e3dSNick Piggin { 900fa0d7e3dSNick Piggin struct inode *inode = container_of(head, struct inode, i_rcu); 901fa0d7e3dSNick Piggin kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 902fa0d7e3dSNick Piggin } 903fa0d7e3dSNick Piggin 904617ba13bSMingming Cao static void ext4_destroy_inode(struct inode *inode) 905ac27a0ecSDave Kleikamp { 9069f7dd93dSVasily Averin if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 907b31e1552SEric Sandeen ext4_msg(inode->i_sb, KERN_ERR, 908b31e1552SEric Sandeen "Inode %lu (%p): orphan list check failed!", 909b31e1552SEric Sandeen inode->i_ino, EXT4_I(inode)); 9109f7dd93dSVasily Averin print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 9119f7dd93dSVasily Averin EXT4_I(inode), sizeof(struct ext4_inode_info), 9129f7dd93dSVasily Averin true); 9139f7dd93dSVasily Averin dump_stack(); 9149f7dd93dSVasily Averin } 915fa0d7e3dSNick Piggin call_rcu(&inode->i_rcu, ext4_i_callback); 916ac27a0ecSDave Kleikamp } 917ac27a0ecSDave Kleikamp 91851cc5068SAlexey Dobriyan static void init_once(void *foo) 919ac27a0ecSDave Kleikamp { 920617ba13bSMingming Cao struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 921ac27a0ecSDave Kleikamp 922ac27a0ecSDave Kleikamp INIT_LIST_HEAD(&ei->i_orphan); 923ac27a0ecSDave Kleikamp init_rwsem(&ei->xattr_sem); 9240e855ac8SAneesh Kumar K.V init_rwsem(&ei->i_data_sem); 925ac27a0ecSDave Kleikamp inode_init_once(&ei->vfs_inode); 926ac27a0ecSDave Kleikamp } 927ac27a0ecSDave Kleikamp 928ac27a0ecSDave Kleikamp static int init_inodecache(void) 929ac27a0ecSDave Kleikamp { 930617ba13bSMingming Cao ext4_inode_cachep = kmem_cache_create("ext4_inode_cache", 931617ba13bSMingming Cao sizeof(struct ext4_inode_info), 932ac27a0ecSDave Kleikamp 0, (SLAB_RECLAIM_ACCOUNT| 933ac27a0ecSDave Kleikamp SLAB_MEM_SPREAD), 93420c2df83SPaul Mundt init_once); 935617ba13bSMingming Cao if (ext4_inode_cachep == NULL) 936ac27a0ecSDave Kleikamp return -ENOMEM; 937ac27a0ecSDave Kleikamp return 0; 938ac27a0ecSDave Kleikamp } 939ac27a0ecSDave Kleikamp 940ac27a0ecSDave Kleikamp static void destroy_inodecache(void) 941ac27a0ecSDave Kleikamp { 9428c0a8537SKirill A. Shutemov /* 9438c0a8537SKirill A. Shutemov * Make sure all delayed rcu free inodes are flushed before we 9448c0a8537SKirill A. Shutemov * destroy cache. 9458c0a8537SKirill A. Shutemov */ 9468c0a8537SKirill A. Shutemov rcu_barrier(); 947617ba13bSMingming Cao kmem_cache_destroy(ext4_inode_cachep); 948ac27a0ecSDave Kleikamp } 949ac27a0ecSDave Kleikamp 9500930fcc1SAl Viro void ext4_clear_inode(struct inode *inode) 951ac27a0ecSDave Kleikamp { 9520930fcc1SAl Viro invalidate_inode_buffers(inode); 953dbd5768fSJan Kara clear_inode(inode); 9549f754758SChristoph Hellwig dquot_drop(inode); 955c2ea3fdeSTheodore Ts'o ext4_discard_preallocations(inode); 95651865fdaSZheng Liu ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 95774cd15cdSZheng Liu ext4_es_lru_del(inode); 9588aefcd55STheodore Ts'o if (EXT4_I(inode)->jinode) { 9598aefcd55STheodore Ts'o jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 9608aefcd55STheodore Ts'o EXT4_I(inode)->jinode); 9618aefcd55STheodore Ts'o jbd2_free_inode(EXT4_I(inode)->jinode); 9628aefcd55STheodore Ts'o EXT4_I(inode)->jinode = NULL; 9638aefcd55STheodore Ts'o } 964ac27a0ecSDave Kleikamp } 965ac27a0ecSDave Kleikamp 9661b961ac0SChristoph Hellwig static struct inode *ext4_nfs_get_inode(struct super_block *sb, 9671b961ac0SChristoph Hellwig u64 ino, u32 generation) 968ac27a0ecSDave Kleikamp { 969ac27a0ecSDave Kleikamp struct inode *inode; 970ac27a0ecSDave Kleikamp 971617ba13bSMingming Cao if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO) 972ac27a0ecSDave Kleikamp return ERR_PTR(-ESTALE); 973617ba13bSMingming Cao if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count)) 974ac27a0ecSDave Kleikamp return ERR_PTR(-ESTALE); 975ac27a0ecSDave Kleikamp 976ac27a0ecSDave Kleikamp /* iget isn't really right if the inode is currently unallocated!! 977ac27a0ecSDave Kleikamp * 978617ba13bSMingming Cao * ext4_read_inode will return a bad_inode if the inode had been 979ac27a0ecSDave Kleikamp * deleted, so we should be safe. 980ac27a0ecSDave Kleikamp * 981ac27a0ecSDave Kleikamp * Currently we don't know the generation for parent directory, so 982ac27a0ecSDave Kleikamp * a generation of 0 means "accept any" 983ac27a0ecSDave Kleikamp */ 9841d1fe1eeSDavid Howells inode = ext4_iget(sb, ino); 9851d1fe1eeSDavid Howells if (IS_ERR(inode)) 9861d1fe1eeSDavid Howells return ERR_CAST(inode); 9871d1fe1eeSDavid Howells if (generation && inode->i_generation != generation) { 988ac27a0ecSDave Kleikamp iput(inode); 989ac27a0ecSDave Kleikamp return ERR_PTR(-ESTALE); 990ac27a0ecSDave Kleikamp } 9911b961ac0SChristoph Hellwig 9921b961ac0SChristoph Hellwig return inode; 993ac27a0ecSDave Kleikamp } 9941b961ac0SChristoph Hellwig 9951b961ac0SChristoph Hellwig static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 9961b961ac0SChristoph Hellwig int fh_len, int fh_type) 9971b961ac0SChristoph Hellwig { 9981b961ac0SChristoph Hellwig return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 9991b961ac0SChristoph Hellwig ext4_nfs_get_inode); 10001b961ac0SChristoph Hellwig } 10011b961ac0SChristoph Hellwig 10021b961ac0SChristoph Hellwig static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 10031b961ac0SChristoph Hellwig int fh_len, int fh_type) 10041b961ac0SChristoph Hellwig { 10051b961ac0SChristoph Hellwig return generic_fh_to_parent(sb, fid, fh_len, fh_type, 10061b961ac0SChristoph Hellwig ext4_nfs_get_inode); 1007ac27a0ecSDave Kleikamp } 1008ac27a0ecSDave Kleikamp 1009c39a7f84SToshiyuki Okajima /* 1010c39a7f84SToshiyuki Okajima * Try to release metadata pages (indirect blocks, directories) which are 1011c39a7f84SToshiyuki Okajima * mapped via the block device. Since these pages could have journal heads 1012c39a7f84SToshiyuki Okajima * which would prevent try_to_free_buffers() from freeing them, we must use 1013c39a7f84SToshiyuki Okajima * jbd2 layer's try_to_free_buffers() function to release them. 1014c39a7f84SToshiyuki Okajima */ 10150b8e58a1SAndreas Dilger static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 10160b8e58a1SAndreas Dilger gfp_t wait) 1017c39a7f84SToshiyuki Okajima { 1018c39a7f84SToshiyuki Okajima journal_t *journal = EXT4_SB(sb)->s_journal; 1019c39a7f84SToshiyuki Okajima 1020c39a7f84SToshiyuki Okajima WARN_ON(PageChecked(page)); 1021c39a7f84SToshiyuki Okajima if (!page_has_buffers(page)) 1022c39a7f84SToshiyuki Okajima return 0; 1023c39a7f84SToshiyuki Okajima if (journal) 1024c39a7f84SToshiyuki Okajima return jbd2_journal_try_to_free_buffers(journal, page, 1025c39a7f84SToshiyuki Okajima wait & ~__GFP_WAIT); 1026c39a7f84SToshiyuki Okajima return try_to_free_buffers(page); 1027c39a7f84SToshiyuki Okajima } 1028c39a7f84SToshiyuki Okajima 1029ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1030ac27a0ecSDave Kleikamp #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group") 1031ac27a0ecSDave Kleikamp #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA)) 1032ac27a0ecSDave Kleikamp 1033617ba13bSMingming Cao static int ext4_write_dquot(struct dquot *dquot); 1034617ba13bSMingming Cao static int ext4_acquire_dquot(struct dquot *dquot); 1035617ba13bSMingming Cao static int ext4_release_dquot(struct dquot *dquot); 1036617ba13bSMingming Cao static int ext4_mark_dquot_dirty(struct dquot *dquot); 1037617ba13bSMingming Cao static int ext4_write_info(struct super_block *sb, int type); 10386f28e087SJan Kara static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1039f00c9e44SJan Kara struct path *path); 10407c319d32SAditya Kali static int ext4_quota_on_sysfile(struct super_block *sb, int type, 10417c319d32SAditya Kali int format_id); 1042ca0e05e4SDmitry Monakhov static int ext4_quota_off(struct super_block *sb, int type); 10437c319d32SAditya Kali static int ext4_quota_off_sysfile(struct super_block *sb, int type); 1044617ba13bSMingming Cao static int ext4_quota_on_mount(struct super_block *sb, int type); 1045617ba13bSMingming Cao static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1046ac27a0ecSDave Kleikamp size_t len, loff_t off); 1047617ba13bSMingming Cao static ssize_t ext4_quota_write(struct super_block *sb, int type, 1048ac27a0ecSDave Kleikamp const char *data, size_t len, loff_t off); 10497c319d32SAditya Kali static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 10507c319d32SAditya Kali unsigned int flags); 10517c319d32SAditya Kali static int ext4_enable_quotas(struct super_block *sb); 1052ac27a0ecSDave Kleikamp 105361e225dcSAlexey Dobriyan static const struct dquot_operations ext4_quota_operations = { 105460e58e0fSMingming Cao .get_reserved_space = ext4_get_reserved_space, 1055617ba13bSMingming Cao .write_dquot = ext4_write_dquot, 1056617ba13bSMingming Cao .acquire_dquot = ext4_acquire_dquot, 1057617ba13bSMingming Cao .release_dquot = ext4_release_dquot, 1058617ba13bSMingming Cao .mark_dirty = ext4_mark_dquot_dirty, 1059a5b5ee32SJan Kara .write_info = ext4_write_info, 1060a5b5ee32SJan Kara .alloc_dquot = dquot_alloc, 1061a5b5ee32SJan Kara .destroy_dquot = dquot_destroy, 1062ac27a0ecSDave Kleikamp }; 1063ac27a0ecSDave Kleikamp 10640d54b217SAlexey Dobriyan static const struct quotactl_ops ext4_qctl_operations = { 1065617ba13bSMingming Cao .quota_on = ext4_quota_on, 1066ca0e05e4SDmitry Monakhov .quota_off = ext4_quota_off, 1067287a8095SChristoph Hellwig .quota_sync = dquot_quota_sync, 1068287a8095SChristoph Hellwig .get_info = dquot_get_dqinfo, 1069287a8095SChristoph Hellwig .set_info = dquot_set_dqinfo, 1070287a8095SChristoph Hellwig .get_dqblk = dquot_get_dqblk, 1071287a8095SChristoph Hellwig .set_dqblk = dquot_set_dqblk 1072ac27a0ecSDave Kleikamp }; 10737c319d32SAditya Kali 10747c319d32SAditya Kali static const struct quotactl_ops ext4_qctl_sysfile_operations = { 10757c319d32SAditya Kali .quota_on_meta = ext4_quota_on_sysfile, 10767c319d32SAditya Kali .quota_off = ext4_quota_off_sysfile, 10777c319d32SAditya Kali .quota_sync = dquot_quota_sync, 10787c319d32SAditya Kali .get_info = dquot_get_dqinfo, 10797c319d32SAditya Kali .set_info = dquot_set_dqinfo, 10807c319d32SAditya Kali .get_dqblk = dquot_get_dqblk, 10817c319d32SAditya Kali .set_dqblk = dquot_set_dqblk 10827c319d32SAditya Kali }; 1083ac27a0ecSDave Kleikamp #endif 1084ac27a0ecSDave Kleikamp 1085ee9b6d61SJosef 'Jeff' Sipek static const struct super_operations ext4_sops = { 1086617ba13bSMingming Cao .alloc_inode = ext4_alloc_inode, 1087617ba13bSMingming Cao .destroy_inode = ext4_destroy_inode, 1088617ba13bSMingming Cao .write_inode = ext4_write_inode, 1089617ba13bSMingming Cao .dirty_inode = ext4_dirty_inode, 10907ff9c073STheodore Ts'o .drop_inode = ext4_drop_inode, 10910930fcc1SAl Viro .evict_inode = ext4_evict_inode, 1092617ba13bSMingming Cao .put_super = ext4_put_super, 1093617ba13bSMingming Cao .sync_fs = ext4_sync_fs, 1094c4be0c1dSTakashi Sato .freeze_fs = ext4_freeze, 1095c4be0c1dSTakashi Sato .unfreeze_fs = ext4_unfreeze, 1096617ba13bSMingming Cao .statfs = ext4_statfs, 1097617ba13bSMingming Cao .remount_fs = ext4_remount, 1098617ba13bSMingming Cao .show_options = ext4_show_options, 1099ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1100617ba13bSMingming Cao .quota_read = ext4_quota_read, 1101617ba13bSMingming Cao .quota_write = ext4_quota_write, 1102ac27a0ecSDave Kleikamp #endif 1103c39a7f84SToshiyuki Okajima .bdev_try_to_free_page = bdev_try_to_free_page, 1104ac27a0ecSDave Kleikamp }; 1105ac27a0ecSDave Kleikamp 11069ca92389STheodore Ts'o static const struct super_operations ext4_nojournal_sops = { 11079ca92389STheodore Ts'o .alloc_inode = ext4_alloc_inode, 11089ca92389STheodore Ts'o .destroy_inode = ext4_destroy_inode, 11099ca92389STheodore Ts'o .write_inode = ext4_write_inode, 11109ca92389STheodore Ts'o .dirty_inode = ext4_dirty_inode, 11117ff9c073STheodore Ts'o .drop_inode = ext4_drop_inode, 11120930fcc1SAl Viro .evict_inode = ext4_evict_inode, 111306a407f1SDmitry Monakhov .sync_fs = ext4_sync_fs_nojournal, 11149ca92389STheodore Ts'o .put_super = ext4_put_super, 11159ca92389STheodore Ts'o .statfs = ext4_statfs, 11169ca92389STheodore Ts'o .remount_fs = ext4_remount, 11179ca92389STheodore Ts'o .show_options = ext4_show_options, 11189ca92389STheodore Ts'o #ifdef CONFIG_QUOTA 11199ca92389STheodore Ts'o .quota_read = ext4_quota_read, 11209ca92389STheodore Ts'o .quota_write = ext4_quota_write, 11219ca92389STheodore Ts'o #endif 11229ca92389STheodore Ts'o .bdev_try_to_free_page = bdev_try_to_free_page, 11239ca92389STheodore Ts'o }; 11249ca92389STheodore Ts'o 112539655164SChristoph Hellwig static const struct export_operations ext4_export_ops = { 11261b961ac0SChristoph Hellwig .fh_to_dentry = ext4_fh_to_dentry, 11271b961ac0SChristoph Hellwig .fh_to_parent = ext4_fh_to_parent, 1128617ba13bSMingming Cao .get_parent = ext4_get_parent, 1129ac27a0ecSDave Kleikamp }; 1130ac27a0ecSDave Kleikamp 1131ac27a0ecSDave Kleikamp enum { 1132ac27a0ecSDave Kleikamp Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1133ac27a0ecSDave Kleikamp Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 113472578c33STheodore Ts'o Opt_nouid32, Opt_debug, Opt_removed, 1135ac27a0ecSDave Kleikamp Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 113672578c33STheodore Ts'o Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 113730773840STheodore Ts'o Opt_commit, Opt_min_batch_time, Opt_max_batch_time, 113843e625d8SEric Sandeen Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit, 1139ac27a0ecSDave Kleikamp Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 1140296c355cSTheodore Ts'o Opt_data_err_abort, Opt_data_err_ignore, 1141ac27a0ecSDave Kleikamp Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 11425a20bdfcSJan Kara Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1143ee4a3fcdSTheodore Ts'o Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 1144661aa520SEric Sandeen Opt_usrquota, Opt_grpquota, Opt_i_version, 11451449032bSTheodore Ts'o Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit, 11461449032bSTheodore Ts'o Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 11475328e635SEric Sandeen Opt_inode_readahead_blks, Opt_journal_ioprio, 1148744692dcSJiaying Zhang Opt_dioread_nolock, Opt_dioread_lock, 1149fc6cb1cdSTheodore Ts'o Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1150df981d03STheodore Ts'o Opt_max_dir_size_kb, 1151ac27a0ecSDave Kleikamp }; 1152ac27a0ecSDave Kleikamp 1153a447c093SSteven Whitehouse static const match_table_t tokens = { 1154ac27a0ecSDave Kleikamp {Opt_bsd_df, "bsddf"}, 1155ac27a0ecSDave Kleikamp {Opt_minix_df, "minixdf"}, 1156ac27a0ecSDave Kleikamp {Opt_grpid, "grpid"}, 1157ac27a0ecSDave Kleikamp {Opt_grpid, "bsdgroups"}, 1158ac27a0ecSDave Kleikamp {Opt_nogrpid, "nogrpid"}, 1159ac27a0ecSDave Kleikamp {Opt_nogrpid, "sysvgroups"}, 1160ac27a0ecSDave Kleikamp {Opt_resgid, "resgid=%u"}, 1161ac27a0ecSDave Kleikamp {Opt_resuid, "resuid=%u"}, 1162ac27a0ecSDave Kleikamp {Opt_sb, "sb=%u"}, 1163ac27a0ecSDave Kleikamp {Opt_err_cont, "errors=continue"}, 1164ac27a0ecSDave Kleikamp {Opt_err_panic, "errors=panic"}, 1165ac27a0ecSDave Kleikamp {Opt_err_ro, "errors=remount-ro"}, 1166ac27a0ecSDave Kleikamp {Opt_nouid32, "nouid32"}, 1167ac27a0ecSDave Kleikamp {Opt_debug, "debug"}, 116872578c33STheodore Ts'o {Opt_removed, "oldalloc"}, 116972578c33STheodore Ts'o {Opt_removed, "orlov"}, 1170ac27a0ecSDave Kleikamp {Opt_user_xattr, "user_xattr"}, 1171ac27a0ecSDave Kleikamp {Opt_nouser_xattr, "nouser_xattr"}, 1172ac27a0ecSDave Kleikamp {Opt_acl, "acl"}, 1173ac27a0ecSDave Kleikamp {Opt_noacl, "noacl"}, 1174e3bb52aeSEric Sandeen {Opt_noload, "norecovery"}, 11755a916be1STheodore Ts'o {Opt_noload, "noload"}, 117672578c33STheodore Ts'o {Opt_removed, "nobh"}, 117772578c33STheodore Ts'o {Opt_removed, "bh"}, 1178ac27a0ecSDave Kleikamp {Opt_commit, "commit=%u"}, 117930773840STheodore Ts'o {Opt_min_batch_time, "min_batch_time=%u"}, 118030773840STheodore Ts'o {Opt_max_batch_time, "max_batch_time=%u"}, 1181ac27a0ecSDave Kleikamp {Opt_journal_dev, "journal_dev=%u"}, 1182818d276cSGirish Shilamkar {Opt_journal_checksum, "journal_checksum"}, 1183818d276cSGirish Shilamkar {Opt_journal_async_commit, "journal_async_commit"}, 1184ac27a0ecSDave Kleikamp {Opt_abort, "abort"}, 1185ac27a0ecSDave Kleikamp {Opt_data_journal, "data=journal"}, 1186ac27a0ecSDave Kleikamp {Opt_data_ordered, "data=ordered"}, 1187ac27a0ecSDave Kleikamp {Opt_data_writeback, "data=writeback"}, 11885bf5683aSHidehiro Kawai {Opt_data_err_abort, "data_err=abort"}, 11895bf5683aSHidehiro Kawai {Opt_data_err_ignore, "data_err=ignore"}, 1190ac27a0ecSDave Kleikamp {Opt_offusrjquota, "usrjquota="}, 1191ac27a0ecSDave Kleikamp {Opt_usrjquota, "usrjquota=%s"}, 1192ac27a0ecSDave Kleikamp {Opt_offgrpjquota, "grpjquota="}, 1193ac27a0ecSDave Kleikamp {Opt_grpjquota, "grpjquota=%s"}, 1194ac27a0ecSDave Kleikamp {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1195ac27a0ecSDave Kleikamp {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 11965a20bdfcSJan Kara {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1197ac27a0ecSDave Kleikamp {Opt_grpquota, "grpquota"}, 1198ac27a0ecSDave Kleikamp {Opt_noquota, "noquota"}, 1199ac27a0ecSDave Kleikamp {Opt_quota, "quota"}, 1200ac27a0ecSDave Kleikamp {Opt_usrquota, "usrquota"}, 1201ac27a0ecSDave Kleikamp {Opt_barrier, "barrier=%u"}, 120206705bffSTheodore Ts'o {Opt_barrier, "barrier"}, 120306705bffSTheodore Ts'o {Opt_nobarrier, "nobarrier"}, 120425ec56b5SJean Noel Cordenner {Opt_i_version, "i_version"}, 1205c9de560dSAlex Tomas {Opt_stripe, "stripe=%u"}, 120664769240SAlex Tomas {Opt_delalloc, "delalloc"}, 1207dd919b98SAneesh Kumar K.V {Opt_nodelalloc, "nodelalloc"}, 120836ade451SJan Kara {Opt_removed, "mblk_io_submit"}, 120936ade451SJan Kara {Opt_removed, "nomblk_io_submit"}, 12106fd058f7STheodore Ts'o {Opt_block_validity, "block_validity"}, 12116fd058f7STheodore Ts'o {Opt_noblock_validity, "noblock_validity"}, 1212240799cdSTheodore Ts'o {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1213b3881f74STheodore Ts'o {Opt_journal_ioprio, "journal_ioprio=%u"}, 1214afd4672dSTheodore Ts'o {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 121506705bffSTheodore Ts'o {Opt_auto_da_alloc, "auto_da_alloc"}, 121606705bffSTheodore Ts'o {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1217744692dcSJiaying Zhang {Opt_dioread_nolock, "dioread_nolock"}, 1218744692dcSJiaying Zhang {Opt_dioread_lock, "dioread_lock"}, 12195328e635SEric Sandeen {Opt_discard, "discard"}, 12205328e635SEric Sandeen {Opt_nodiscard, "nodiscard"}, 1221fc6cb1cdSTheodore Ts'o {Opt_init_itable, "init_itable=%u"}, 1222fc6cb1cdSTheodore Ts'o {Opt_init_itable, "init_itable"}, 1223fc6cb1cdSTheodore Ts'o {Opt_noinit_itable, "noinit_itable"}, 1224df981d03STheodore Ts'o {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 1225c7198b9cSTheodore Ts'o {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1226c7198b9cSTheodore Ts'o {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1227c7198b9cSTheodore Ts'o {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1228c7198b9cSTheodore Ts'o {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1229c7198b9cSTheodore Ts'o {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1230f3f12faaSJosef Bacik {Opt_err, NULL}, 1231ac27a0ecSDave Kleikamp }; 1232ac27a0ecSDave Kleikamp 1233617ba13bSMingming Cao static ext4_fsblk_t get_sb_block(void **data) 1234ac27a0ecSDave Kleikamp { 1235617ba13bSMingming Cao ext4_fsblk_t sb_block; 1236ac27a0ecSDave Kleikamp char *options = (char *) *data; 1237ac27a0ecSDave Kleikamp 1238ac27a0ecSDave Kleikamp if (!options || strncmp(options, "sb=", 3) != 0) 1239ac27a0ecSDave Kleikamp return 1; /* Default location */ 12400b8e58a1SAndreas Dilger 1241ac27a0ecSDave Kleikamp options += 3; 12420b8e58a1SAndreas Dilger /* TODO: use simple_strtoll with >32bit ext4 */ 1243ac27a0ecSDave Kleikamp sb_block = simple_strtoul(options, &options, 0); 1244ac27a0ecSDave Kleikamp if (*options && *options != ',') { 12454776004fSTheodore Ts'o printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1246ac27a0ecSDave Kleikamp (char *) *data); 1247ac27a0ecSDave Kleikamp return 1; 1248ac27a0ecSDave Kleikamp } 1249ac27a0ecSDave Kleikamp if (*options == ',') 1250ac27a0ecSDave Kleikamp options++; 1251ac27a0ecSDave Kleikamp *data = (void *) options; 12520b8e58a1SAndreas Dilger 1253ac27a0ecSDave Kleikamp return sb_block; 1254ac27a0ecSDave Kleikamp } 1255ac27a0ecSDave Kleikamp 1256b3881f74STheodore Ts'o #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1257437ca0fdSDmitry Monakhov static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n" 1258437ca0fdSDmitry Monakhov "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1259b3881f74STheodore Ts'o 126056c50f11SDmitry Monakhov #ifdef CONFIG_QUOTA 126156c50f11SDmitry Monakhov static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 126256c50f11SDmitry Monakhov { 126356c50f11SDmitry Monakhov struct ext4_sb_info *sbi = EXT4_SB(sb); 126456c50f11SDmitry Monakhov char *qname; 126503dafb5fSChen Gang int ret = -1; 126656c50f11SDmitry Monakhov 126756c50f11SDmitry Monakhov if (sb_any_quota_loaded(sb) && 126856c50f11SDmitry Monakhov !sbi->s_qf_names[qtype]) { 126956c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 127056c50f11SDmitry Monakhov "Cannot change journaled " 127156c50f11SDmitry Monakhov "quota options when quota turned on"); 127257f73c2cSTheodore Ts'o return -1; 127356c50f11SDmitry Monakhov } 1274262b4662SJan Kara if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) { 1275262b4662SJan Kara ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options " 1276262b4662SJan Kara "when QUOTA feature is enabled"); 1277262b4662SJan Kara return -1; 1278262b4662SJan Kara } 127956c50f11SDmitry Monakhov qname = match_strdup(args); 128056c50f11SDmitry Monakhov if (!qname) { 128156c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 128256c50f11SDmitry Monakhov "Not enough memory for storing quotafile name"); 128357f73c2cSTheodore Ts'o return -1; 128456c50f11SDmitry Monakhov } 128503dafb5fSChen Gang if (sbi->s_qf_names[qtype]) { 128603dafb5fSChen Gang if (strcmp(sbi->s_qf_names[qtype], qname) == 0) 128703dafb5fSChen Gang ret = 1; 128803dafb5fSChen Gang else 128956c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 129003dafb5fSChen Gang "%s quota file already specified", 129103dafb5fSChen Gang QTYPE2NAME(qtype)); 129203dafb5fSChen Gang goto errout; 129356c50f11SDmitry Monakhov } 129403dafb5fSChen Gang if (strchr(qname, '/')) { 129556c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 129656c50f11SDmitry Monakhov "quotafile must be on filesystem root"); 129703dafb5fSChen Gang goto errout; 129856c50f11SDmitry Monakhov } 129903dafb5fSChen Gang sbi->s_qf_names[qtype] = qname; 1300fd8c37ecSTheodore Ts'o set_opt(sb, QUOTA); 130156c50f11SDmitry Monakhov return 1; 130203dafb5fSChen Gang errout: 130303dafb5fSChen Gang kfree(qname); 130403dafb5fSChen Gang return ret; 130556c50f11SDmitry Monakhov } 130656c50f11SDmitry Monakhov 130756c50f11SDmitry Monakhov static int clear_qf_name(struct super_block *sb, int qtype) 130856c50f11SDmitry Monakhov { 130956c50f11SDmitry Monakhov 131056c50f11SDmitry Monakhov struct ext4_sb_info *sbi = EXT4_SB(sb); 131156c50f11SDmitry Monakhov 131256c50f11SDmitry Monakhov if (sb_any_quota_loaded(sb) && 131356c50f11SDmitry Monakhov sbi->s_qf_names[qtype]) { 131456c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 131556c50f11SDmitry Monakhov " when quota turned on"); 131657f73c2cSTheodore Ts'o return -1; 131756c50f11SDmitry Monakhov } 131803dafb5fSChen Gang kfree(sbi->s_qf_names[qtype]); 131956c50f11SDmitry Monakhov sbi->s_qf_names[qtype] = NULL; 132056c50f11SDmitry Monakhov return 1; 132156c50f11SDmitry Monakhov } 132256c50f11SDmitry Monakhov #endif 132356c50f11SDmitry Monakhov 132426092bf5STheodore Ts'o #define MOPT_SET 0x0001 132526092bf5STheodore Ts'o #define MOPT_CLEAR 0x0002 132626092bf5STheodore Ts'o #define MOPT_NOSUPPORT 0x0004 132726092bf5STheodore Ts'o #define MOPT_EXPLICIT 0x0008 132826092bf5STheodore Ts'o #define MOPT_CLEAR_ERR 0x0010 132926092bf5STheodore Ts'o #define MOPT_GTE0 0x0020 133026092bf5STheodore Ts'o #ifdef CONFIG_QUOTA 133126092bf5STheodore Ts'o #define MOPT_Q 0 133226092bf5STheodore Ts'o #define MOPT_QFMT 0x0040 133326092bf5STheodore Ts'o #else 133426092bf5STheodore Ts'o #define MOPT_Q MOPT_NOSUPPORT 133526092bf5STheodore Ts'o #define MOPT_QFMT MOPT_NOSUPPORT 133626092bf5STheodore Ts'o #endif 133726092bf5STheodore Ts'o #define MOPT_DATAJ 0x0080 13388dc0aa8cSTheodore Ts'o #define MOPT_NO_EXT2 0x0100 13398dc0aa8cSTheodore Ts'o #define MOPT_NO_EXT3 0x0200 13408dc0aa8cSTheodore Ts'o #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 134126092bf5STheodore Ts'o 134226092bf5STheodore Ts'o static const struct mount_opts { 134326092bf5STheodore Ts'o int token; 134426092bf5STheodore Ts'o int mount_opt; 134526092bf5STheodore Ts'o int flags; 134626092bf5STheodore Ts'o } ext4_mount_opts[] = { 134726092bf5STheodore Ts'o {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 134826092bf5STheodore Ts'o {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 134926092bf5STheodore Ts'o {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 135026092bf5STheodore Ts'o {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 135126092bf5STheodore Ts'o {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 135226092bf5STheodore Ts'o {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 13538dc0aa8cSTheodore Ts'o {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 13548dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_SET}, 13558dc0aa8cSTheodore Ts'o {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 13568dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_CLEAR}, 135726092bf5STheodore Ts'o {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 135826092bf5STheodore Ts'o {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 13598dc0aa8cSTheodore Ts'o {Opt_delalloc, EXT4_MOUNT_DELALLOC, 13608dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 13618dc0aa8cSTheodore Ts'o {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 1362*59d9fa5cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_CLEAR}, 13638dc0aa8cSTheodore Ts'o {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 13648dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_SET}, 136526092bf5STheodore Ts'o {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 13668dc0aa8cSTheodore Ts'o EXT4_MOUNT_JOURNAL_CHECKSUM), 13678dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_SET}, 13688dc0aa8cSTheodore Ts'o {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 136926092bf5STheodore Ts'o {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 137026092bf5STheodore Ts'o {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 137126092bf5STheodore Ts'o {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 13728dc0aa8cSTheodore Ts'o {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 13738dc0aa8cSTheodore Ts'o MOPT_NO_EXT2 | MOPT_SET}, 13748dc0aa8cSTheodore Ts'o {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 13758dc0aa8cSTheodore Ts'o MOPT_NO_EXT2 | MOPT_CLEAR}, 137626092bf5STheodore Ts'o {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 137726092bf5STheodore Ts'o {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 137826092bf5STheodore Ts'o {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 137926092bf5STheodore Ts'o {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 138026092bf5STheodore Ts'o {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 138126092bf5STheodore Ts'o {Opt_commit, 0, MOPT_GTE0}, 138226092bf5STheodore Ts'o {Opt_max_batch_time, 0, MOPT_GTE0}, 138326092bf5STheodore Ts'o {Opt_min_batch_time, 0, MOPT_GTE0}, 138426092bf5STheodore Ts'o {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 138526092bf5STheodore Ts'o {Opt_init_itable, 0, MOPT_GTE0}, 138626092bf5STheodore Ts'o {Opt_stripe, 0, MOPT_GTE0}, 13870efb3b23SJan Kara {Opt_resuid, 0, MOPT_GTE0}, 13880efb3b23SJan Kara {Opt_resgid, 0, MOPT_GTE0}, 13890efb3b23SJan Kara {Opt_journal_dev, 0, MOPT_GTE0}, 13900efb3b23SJan Kara {Opt_journal_ioprio, 0, MOPT_GTE0}, 13918dc0aa8cSTheodore Ts'o {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 13928dc0aa8cSTheodore Ts'o {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 13938dc0aa8cSTheodore Ts'o {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 13948dc0aa8cSTheodore Ts'o MOPT_NO_EXT2 | MOPT_DATAJ}, 139526092bf5STheodore Ts'o {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 139626092bf5STheodore Ts'o {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 139726092bf5STheodore Ts'o #ifdef CONFIG_EXT4_FS_POSIX_ACL 139826092bf5STheodore Ts'o {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 139926092bf5STheodore Ts'o {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 140026092bf5STheodore Ts'o #else 140126092bf5STheodore Ts'o {Opt_acl, 0, MOPT_NOSUPPORT}, 140226092bf5STheodore Ts'o {Opt_noacl, 0, MOPT_NOSUPPORT}, 140326092bf5STheodore Ts'o #endif 140426092bf5STheodore Ts'o {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 140526092bf5STheodore Ts'o {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 140626092bf5STheodore Ts'o {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 140726092bf5STheodore Ts'o {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 140826092bf5STheodore Ts'o MOPT_SET | MOPT_Q}, 140926092bf5STheodore Ts'o {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 141026092bf5STheodore Ts'o MOPT_SET | MOPT_Q}, 141126092bf5STheodore Ts'o {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 141226092bf5STheodore Ts'o EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q}, 141326092bf5STheodore Ts'o {Opt_usrjquota, 0, MOPT_Q}, 141426092bf5STheodore Ts'o {Opt_grpjquota, 0, MOPT_Q}, 141526092bf5STheodore Ts'o {Opt_offusrjquota, 0, MOPT_Q}, 141626092bf5STheodore Ts'o {Opt_offgrpjquota, 0, MOPT_Q}, 141726092bf5STheodore Ts'o {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 141826092bf5STheodore Ts'o {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 141926092bf5STheodore Ts'o {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 1420df981d03STheodore Ts'o {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 142126092bf5STheodore Ts'o {Opt_err, 0, 0} 142226092bf5STheodore Ts'o }; 142326092bf5STheodore Ts'o 142426092bf5STheodore Ts'o static int handle_mount_opt(struct super_block *sb, char *opt, int token, 142526092bf5STheodore Ts'o substring_t *args, unsigned long *journal_devnum, 142626092bf5STheodore Ts'o unsigned int *journal_ioprio, int is_remount) 142726092bf5STheodore Ts'o { 142826092bf5STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 142926092bf5STheodore Ts'o const struct mount_opts *m; 143008cefc7aSEric W. Biederman kuid_t uid; 143108cefc7aSEric W. Biederman kgid_t gid; 143226092bf5STheodore Ts'o int arg = 0; 143326092bf5STheodore Ts'o 143457f73c2cSTheodore Ts'o #ifdef CONFIG_QUOTA 143557f73c2cSTheodore Ts'o if (token == Opt_usrjquota) 143657f73c2cSTheodore Ts'o return set_qf_name(sb, USRQUOTA, &args[0]); 143757f73c2cSTheodore Ts'o else if (token == Opt_grpjquota) 143857f73c2cSTheodore Ts'o return set_qf_name(sb, GRPQUOTA, &args[0]); 143957f73c2cSTheodore Ts'o else if (token == Opt_offusrjquota) 144057f73c2cSTheodore Ts'o return clear_qf_name(sb, USRQUOTA); 144157f73c2cSTheodore Ts'o else if (token == Opt_offgrpjquota) 144257f73c2cSTheodore Ts'o return clear_qf_name(sb, GRPQUOTA); 144357f73c2cSTheodore Ts'o #endif 144426092bf5STheodore Ts'o switch (token) { 1445f7048605STheodore Ts'o case Opt_noacl: 1446f7048605STheodore Ts'o case Opt_nouser_xattr: 1447f7048605STheodore Ts'o ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 1448f7048605STheodore Ts'o break; 144926092bf5STheodore Ts'o case Opt_sb: 145026092bf5STheodore Ts'o return 1; /* handled by get_sb_block() */ 145126092bf5STheodore Ts'o case Opt_removed: 14525f3633e3SJan Kara ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 145326092bf5STheodore Ts'o return 1; 145426092bf5STheodore Ts'o case Opt_abort: 145526092bf5STheodore Ts'o sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 145626092bf5STheodore Ts'o return 1; 145726092bf5STheodore Ts'o case Opt_i_version: 145826092bf5STheodore Ts'o sb->s_flags |= MS_I_VERSION; 145926092bf5STheodore Ts'o return 1; 146026092bf5STheodore Ts'o } 146126092bf5STheodore Ts'o 14625f3633e3SJan Kara for (m = ext4_mount_opts; m->token != Opt_err; m++) 14635f3633e3SJan Kara if (token == m->token) 14645f3633e3SJan Kara break; 14655f3633e3SJan Kara 14665f3633e3SJan Kara if (m->token == Opt_err) { 14675f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 14685f3633e3SJan Kara "or missing value", opt); 14695f3633e3SJan Kara return -1; 14705f3633e3SJan Kara } 14715f3633e3SJan Kara 14728dc0aa8cSTheodore Ts'o if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 14738dc0aa8cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 14748dc0aa8cSTheodore Ts'o "Mount option \"%s\" incompatible with ext2", opt); 14758dc0aa8cSTheodore Ts'o return -1; 14768dc0aa8cSTheodore Ts'o } 14778dc0aa8cSTheodore Ts'o if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 14788dc0aa8cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 14798dc0aa8cSTheodore Ts'o "Mount option \"%s\" incompatible with ext3", opt); 14808dc0aa8cSTheodore Ts'o return -1; 14818dc0aa8cSTheodore Ts'o } 14828dc0aa8cSTheodore Ts'o 14830efb3b23SJan Kara if (args->from && match_int(args, &arg)) 14840efb3b23SJan Kara return -1; 148526092bf5STheodore Ts'o if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 148626092bf5STheodore Ts'o return -1; 148726092bf5STheodore Ts'o if (m->flags & MOPT_EXPLICIT) 148826092bf5STheodore Ts'o set_opt2(sb, EXPLICIT_DELALLOC); 148926092bf5STheodore Ts'o if (m->flags & MOPT_CLEAR_ERR) 149026092bf5STheodore Ts'o clear_opt(sb, ERRORS_MASK); 149126092bf5STheodore Ts'o if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 149226092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "Cannot change quota " 149326092bf5STheodore Ts'o "options when quota turned on"); 149426092bf5STheodore Ts'o return -1; 149526092bf5STheodore Ts'o } 149626092bf5STheodore Ts'o 149726092bf5STheodore Ts'o if (m->flags & MOPT_NOSUPPORT) { 149826092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 149926092bf5STheodore Ts'o } else if (token == Opt_commit) { 150026092bf5STheodore Ts'o if (arg == 0) 150126092bf5STheodore Ts'o arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 150226092bf5STheodore Ts'o sbi->s_commit_interval = HZ * arg; 150326092bf5STheodore Ts'o } else if (token == Opt_max_batch_time) { 150426092bf5STheodore Ts'o if (arg == 0) 150526092bf5STheodore Ts'o arg = EXT4_DEF_MAX_BATCH_TIME; 150626092bf5STheodore Ts'o sbi->s_max_batch_time = arg; 150726092bf5STheodore Ts'o } else if (token == Opt_min_batch_time) { 150826092bf5STheodore Ts'o sbi->s_min_batch_time = arg; 150926092bf5STheodore Ts'o } else if (token == Opt_inode_readahead_blks) { 1510e33e60eaSJan Kara if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 1511e33e60eaSJan Kara ext4_msg(sb, KERN_ERR, 1512e33e60eaSJan Kara "EXT4-fs: inode_readahead_blks must be " 1513e33e60eaSJan Kara "0 or a power of 2 smaller than 2^31"); 151426092bf5STheodore Ts'o return -1; 151526092bf5STheodore Ts'o } 151626092bf5STheodore Ts'o sbi->s_inode_readahead_blks = arg; 151726092bf5STheodore Ts'o } else if (token == Opt_init_itable) { 151826092bf5STheodore Ts'o set_opt(sb, INIT_INODE_TABLE); 151926092bf5STheodore Ts'o if (!args->from) 152026092bf5STheodore Ts'o arg = EXT4_DEF_LI_WAIT_MULT; 152126092bf5STheodore Ts'o sbi->s_li_wait_mult = arg; 1522df981d03STheodore Ts'o } else if (token == Opt_max_dir_size_kb) { 1523df981d03STheodore Ts'o sbi->s_max_dir_size_kb = arg; 152426092bf5STheodore Ts'o } else if (token == Opt_stripe) { 152526092bf5STheodore Ts'o sbi->s_stripe = arg; 15260efb3b23SJan Kara } else if (token == Opt_resuid) { 15270efb3b23SJan Kara uid = make_kuid(current_user_ns(), arg); 15280efb3b23SJan Kara if (!uid_valid(uid)) { 15295f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 15300efb3b23SJan Kara return -1; 15310efb3b23SJan Kara } 15320efb3b23SJan Kara sbi->s_resuid = uid; 15330efb3b23SJan Kara } else if (token == Opt_resgid) { 15340efb3b23SJan Kara gid = make_kgid(current_user_ns(), arg); 15350efb3b23SJan Kara if (!gid_valid(gid)) { 15365f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 15370efb3b23SJan Kara return -1; 15380efb3b23SJan Kara } 15390efb3b23SJan Kara sbi->s_resgid = gid; 15400efb3b23SJan Kara } else if (token == Opt_journal_dev) { 15410efb3b23SJan Kara if (is_remount) { 15420efb3b23SJan Kara ext4_msg(sb, KERN_ERR, 15430efb3b23SJan Kara "Cannot specify journal on remount"); 15440efb3b23SJan Kara return -1; 15450efb3b23SJan Kara } 15460efb3b23SJan Kara *journal_devnum = arg; 15470efb3b23SJan Kara } else if (token == Opt_journal_ioprio) { 15480efb3b23SJan Kara if (arg > 7) { 15495f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 15500efb3b23SJan Kara " (must be 0-7)"); 15510efb3b23SJan Kara return -1; 15520efb3b23SJan Kara } 15530efb3b23SJan Kara *journal_ioprio = 15540efb3b23SJan Kara IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 155526092bf5STheodore Ts'o } else if (m->flags & MOPT_DATAJ) { 155626092bf5STheodore Ts'o if (is_remount) { 155726092bf5STheodore Ts'o if (!sbi->s_journal) 155826092bf5STheodore Ts'o ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 15595f3633e3SJan Kara else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 156026092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, 156126092bf5STheodore Ts'o "Cannot change data mode on remount"); 156226092bf5STheodore Ts'o return -1; 156326092bf5STheodore Ts'o } 156426092bf5STheodore Ts'o } else { 156526092bf5STheodore Ts'o clear_opt(sb, DATA_FLAGS); 156626092bf5STheodore Ts'o sbi->s_mount_opt |= m->mount_opt; 156726092bf5STheodore Ts'o } 156826092bf5STheodore Ts'o #ifdef CONFIG_QUOTA 156926092bf5STheodore Ts'o } else if (m->flags & MOPT_QFMT) { 157026092bf5STheodore Ts'o if (sb_any_quota_loaded(sb) && 157126092bf5STheodore Ts'o sbi->s_jquota_fmt != m->mount_opt) { 15725f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Cannot change journaled " 15735f3633e3SJan Kara "quota options when quota turned on"); 157426092bf5STheodore Ts'o return -1; 157526092bf5STheodore Ts'o } 1576262b4662SJan Kara if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 1577262b4662SJan Kara EXT4_FEATURE_RO_COMPAT_QUOTA)) { 1578262b4662SJan Kara ext4_msg(sb, KERN_ERR, 1579262b4662SJan Kara "Cannot set journaled quota options " 1580262b4662SJan Kara "when QUOTA feature is enabled"); 1581262b4662SJan Kara return -1; 1582262b4662SJan Kara } 158326092bf5STheodore Ts'o sbi->s_jquota_fmt = m->mount_opt; 158426092bf5STheodore Ts'o #endif 158526092bf5STheodore Ts'o } else { 158626092bf5STheodore Ts'o if (!args->from) 158726092bf5STheodore Ts'o arg = 1; 158826092bf5STheodore Ts'o if (m->flags & MOPT_CLEAR) 158926092bf5STheodore Ts'o arg = !arg; 159026092bf5STheodore Ts'o else if (unlikely(!(m->flags & MOPT_SET))) { 159126092bf5STheodore Ts'o ext4_msg(sb, KERN_WARNING, 159226092bf5STheodore Ts'o "buggy handling of option %s", opt); 159326092bf5STheodore Ts'o WARN_ON(1); 159426092bf5STheodore Ts'o return -1; 159526092bf5STheodore Ts'o } 159626092bf5STheodore Ts'o if (arg != 0) 159726092bf5STheodore Ts'o sbi->s_mount_opt |= m->mount_opt; 159826092bf5STheodore Ts'o else 159926092bf5STheodore Ts'o sbi->s_mount_opt &= ~m->mount_opt; 160026092bf5STheodore Ts'o } 160126092bf5STheodore Ts'o return 1; 160226092bf5STheodore Ts'o } 160326092bf5STheodore Ts'o 1604ac27a0ecSDave Kleikamp static int parse_options(char *options, struct super_block *sb, 1605c3191067STheodore Ts'o unsigned long *journal_devnum, 1606b3881f74STheodore Ts'o unsigned int *journal_ioprio, 1607661aa520SEric Sandeen int is_remount) 1608ac27a0ecSDave Kleikamp { 1609617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 1610ac27a0ecSDave Kleikamp char *p; 1611ac27a0ecSDave Kleikamp substring_t args[MAX_OPT_ARGS]; 161226092bf5STheodore Ts'o int token; 1613ac27a0ecSDave Kleikamp 1614ac27a0ecSDave Kleikamp if (!options) 1615ac27a0ecSDave Kleikamp return 1; 1616ac27a0ecSDave Kleikamp 1617ac27a0ecSDave Kleikamp while ((p = strsep(&options, ",")) != NULL) { 1618ac27a0ecSDave Kleikamp if (!*p) 1619ac27a0ecSDave Kleikamp continue; 162015121c18SEric Sandeen /* 162115121c18SEric Sandeen * Initialize args struct so we know whether arg was 162215121c18SEric Sandeen * found; some options take optional arguments. 162315121c18SEric Sandeen */ 1624caecd0afSSachin Kamat args[0].to = args[0].from = NULL; 1625ac27a0ecSDave Kleikamp token = match_token(p, tokens, args); 162626092bf5STheodore Ts'o if (handle_mount_opt(sb, p, token, args, journal_devnum, 162726092bf5STheodore Ts'o journal_ioprio, is_remount) < 0) 1628ac27a0ecSDave Kleikamp return 0; 1629ac27a0ecSDave Kleikamp } 1630ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1631262b4662SJan Kara if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) && 1632262b4662SJan Kara (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) { 1633262b4662SJan Kara ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA " 1634262b4662SJan Kara "feature is enabled"); 1635262b4662SJan Kara return 0; 1636262b4662SJan Kara } 1637ac27a0ecSDave Kleikamp if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 1638482a7425SDmitry Monakhov if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) 1639fd8c37ecSTheodore Ts'o clear_opt(sb, USRQUOTA); 1640ac27a0ecSDave Kleikamp 1641482a7425SDmitry Monakhov if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) 1642fd8c37ecSTheodore Ts'o clear_opt(sb, GRPQUOTA); 1643ac27a0ecSDave Kleikamp 164456c50f11SDmitry Monakhov if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 1645b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "old and new quota " 1646b31e1552SEric Sandeen "format mixing"); 1647ac27a0ecSDave Kleikamp return 0; 1648ac27a0ecSDave Kleikamp } 1649ac27a0ecSDave Kleikamp 1650ac27a0ecSDave Kleikamp if (!sbi->s_jquota_fmt) { 1651b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journaled quota format " 1652b31e1552SEric Sandeen "not specified"); 1653ac27a0ecSDave Kleikamp return 0; 1654ac27a0ecSDave Kleikamp } 1655ac27a0ecSDave Kleikamp } else { 1656ac27a0ecSDave Kleikamp if (sbi->s_jquota_fmt) { 1657b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journaled quota format " 16582c8be6b2SJan Kara "specified with no journaling " 1659b31e1552SEric Sandeen "enabled"); 1660ac27a0ecSDave Kleikamp return 0; 1661ac27a0ecSDave Kleikamp } 1662ac27a0ecSDave Kleikamp } 1663ac27a0ecSDave Kleikamp #endif 1664261cb20cSJan Kara if (test_opt(sb, DIOREAD_NOLOCK)) { 1665261cb20cSJan Kara int blocksize = 1666261cb20cSJan Kara BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 1667261cb20cSJan Kara 1668261cb20cSJan Kara if (blocksize < PAGE_CACHE_SIZE) { 1669261cb20cSJan Kara ext4_msg(sb, KERN_ERR, "can't mount with " 1670261cb20cSJan Kara "dioread_nolock if block size != PAGE_SIZE"); 1671261cb20cSJan Kara return 0; 1672261cb20cSJan Kara } 1673261cb20cSJan Kara } 1674ac27a0ecSDave Kleikamp return 1; 1675ac27a0ecSDave Kleikamp } 1676ac27a0ecSDave Kleikamp 16772adf6da8STheodore Ts'o static inline void ext4_show_quota_options(struct seq_file *seq, 16782adf6da8STheodore Ts'o struct super_block *sb) 16792adf6da8STheodore Ts'o { 16802adf6da8STheodore Ts'o #if defined(CONFIG_QUOTA) 16812adf6da8STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 16822adf6da8STheodore Ts'o 16832adf6da8STheodore Ts'o if (sbi->s_jquota_fmt) { 16842adf6da8STheodore Ts'o char *fmtname = ""; 16852adf6da8STheodore Ts'o 16862adf6da8STheodore Ts'o switch (sbi->s_jquota_fmt) { 16872adf6da8STheodore Ts'o case QFMT_VFS_OLD: 16882adf6da8STheodore Ts'o fmtname = "vfsold"; 16892adf6da8STheodore Ts'o break; 16902adf6da8STheodore Ts'o case QFMT_VFS_V0: 16912adf6da8STheodore Ts'o fmtname = "vfsv0"; 16922adf6da8STheodore Ts'o break; 16932adf6da8STheodore Ts'o case QFMT_VFS_V1: 16942adf6da8STheodore Ts'o fmtname = "vfsv1"; 16952adf6da8STheodore Ts'o break; 16962adf6da8STheodore Ts'o } 16972adf6da8STheodore Ts'o seq_printf(seq, ",jqfmt=%s", fmtname); 16982adf6da8STheodore Ts'o } 16992adf6da8STheodore Ts'o 17002adf6da8STheodore Ts'o if (sbi->s_qf_names[USRQUOTA]) 17012adf6da8STheodore Ts'o seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]); 17022adf6da8STheodore Ts'o 17032adf6da8STheodore Ts'o if (sbi->s_qf_names[GRPQUOTA]) 17042adf6da8STheodore Ts'o seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]); 17052adf6da8STheodore Ts'o #endif 17062adf6da8STheodore Ts'o } 17072adf6da8STheodore Ts'o 17085a916be1STheodore Ts'o static const char *token2str(int token) 17095a916be1STheodore Ts'o { 171050df9fd5SHerton Ronaldo Krzesinski const struct match_token *t; 17115a916be1STheodore Ts'o 17125a916be1STheodore Ts'o for (t = tokens; t->token != Opt_err; t++) 17135a916be1STheodore Ts'o if (t->token == token && !strchr(t->pattern, '=')) 17145a916be1STheodore Ts'o break; 17155a916be1STheodore Ts'o return t->pattern; 17165a916be1STheodore Ts'o } 17175a916be1STheodore Ts'o 17182adf6da8STheodore Ts'o /* 17192adf6da8STheodore Ts'o * Show an option if 17202adf6da8STheodore Ts'o * - it's set to a non-default value OR 17212adf6da8STheodore Ts'o * - if the per-sb default is different from the global default 17222adf6da8STheodore Ts'o */ 172366acdcf4STheodore Ts'o static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 172466acdcf4STheodore Ts'o int nodefs) 17252adf6da8STheodore Ts'o { 17262adf6da8STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 17272adf6da8STheodore Ts'o struct ext4_super_block *es = sbi->s_es; 172866acdcf4STheodore Ts'o int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt; 17295a916be1STheodore Ts'o const struct mount_opts *m; 173066acdcf4STheodore Ts'o char sep = nodefs ? '\n' : ','; 17312adf6da8STheodore Ts'o 173266acdcf4STheodore Ts'o #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 173366acdcf4STheodore Ts'o #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 17342adf6da8STheodore Ts'o 17352adf6da8STheodore Ts'o if (sbi->s_sb_block != 1) 17365a916be1STheodore Ts'o SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 17375a916be1STheodore Ts'o 17385a916be1STheodore Ts'o for (m = ext4_mount_opts; m->token != Opt_err; m++) { 17395a916be1STheodore Ts'o int want_set = m->flags & MOPT_SET; 17405a916be1STheodore Ts'o if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 17415a916be1STheodore Ts'o (m->flags & MOPT_CLEAR_ERR)) 17425a916be1STheodore Ts'o continue; 174366acdcf4STheodore Ts'o if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 17445a916be1STheodore Ts'o continue; /* skip if same as the default */ 17455a916be1STheodore Ts'o if ((want_set && 17465a916be1STheodore Ts'o (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 17475a916be1STheodore Ts'o (!want_set && (sbi->s_mount_opt & m->mount_opt))) 17485a916be1STheodore Ts'o continue; /* select Opt_noFoo vs Opt_Foo */ 17495a916be1STheodore Ts'o SEQ_OPTS_PRINT("%s", token2str(m->token)); 17505a916be1STheodore Ts'o } 17515a916be1STheodore Ts'o 175208cefc7aSEric W. Biederman if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 17535a916be1STheodore Ts'o le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 175408cefc7aSEric W. Biederman SEQ_OPTS_PRINT("resuid=%u", 175508cefc7aSEric W. Biederman from_kuid_munged(&init_user_ns, sbi->s_resuid)); 175608cefc7aSEric W. Biederman if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 17575a916be1STheodore Ts'o le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 175808cefc7aSEric W. Biederman SEQ_OPTS_PRINT("resgid=%u", 175908cefc7aSEric W. Biederman from_kgid_munged(&init_user_ns, sbi->s_resgid)); 176066acdcf4STheodore Ts'o def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 17615a916be1STheodore Ts'o if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 17625a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=remount-ro"); 17632adf6da8STheodore Ts'o if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 17645a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=continue"); 17652adf6da8STheodore Ts'o if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 17665a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=panic"); 176766acdcf4STheodore Ts'o if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 17685a916be1STheodore Ts'o SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 176966acdcf4STheodore Ts'o if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 17705a916be1STheodore Ts'o SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 177166acdcf4STheodore Ts'o if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 17725a916be1STheodore Ts'o SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 17732adf6da8STheodore Ts'o if (sb->s_flags & MS_I_VERSION) 17745a916be1STheodore Ts'o SEQ_OPTS_PUTS("i_version"); 177566acdcf4STheodore Ts'o if (nodefs || sbi->s_stripe) 17765a916be1STheodore Ts'o SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 177766acdcf4STheodore Ts'o if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) { 17782adf6da8STheodore Ts'o if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 17795a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=journal"); 17802adf6da8STheodore Ts'o else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 17815a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=ordered"); 17822adf6da8STheodore Ts'o else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 17835a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=writeback"); 17845a916be1STheodore Ts'o } 178566acdcf4STheodore Ts'o if (nodefs || 178666acdcf4STheodore Ts'o sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 17875a916be1STheodore Ts'o SEQ_OPTS_PRINT("inode_readahead_blks=%u", 17882adf6da8STheodore Ts'o sbi->s_inode_readahead_blks); 17892adf6da8STheodore Ts'o 179066acdcf4STheodore Ts'o if (nodefs || (test_opt(sb, INIT_INODE_TABLE) && 179166acdcf4STheodore Ts'o (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 17925a916be1STheodore Ts'o SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 1793df981d03STheodore Ts'o if (nodefs || sbi->s_max_dir_size_kb) 1794df981d03STheodore Ts'o SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 17952adf6da8STheodore Ts'o 17962adf6da8STheodore Ts'o ext4_show_quota_options(seq, sb); 17972adf6da8STheodore Ts'o return 0; 17982adf6da8STheodore Ts'o } 17992adf6da8STheodore Ts'o 180066acdcf4STheodore Ts'o static int ext4_show_options(struct seq_file *seq, struct dentry *root) 180166acdcf4STheodore Ts'o { 180266acdcf4STheodore Ts'o return _ext4_show_options(seq, root->d_sb, 0); 180366acdcf4STheodore Ts'o } 180466acdcf4STheodore Ts'o 180566acdcf4STheodore Ts'o static int options_seq_show(struct seq_file *seq, void *offset) 180666acdcf4STheodore Ts'o { 180766acdcf4STheodore Ts'o struct super_block *sb = seq->private; 180866acdcf4STheodore Ts'o int rc; 180966acdcf4STheodore Ts'o 181066acdcf4STheodore Ts'o seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw"); 181166acdcf4STheodore Ts'o rc = _ext4_show_options(seq, sb, 1); 181266acdcf4STheodore Ts'o seq_puts(seq, "\n"); 181366acdcf4STheodore Ts'o return rc; 181466acdcf4STheodore Ts'o } 181566acdcf4STheodore Ts'o 181666acdcf4STheodore Ts'o static int options_open_fs(struct inode *inode, struct file *file) 181766acdcf4STheodore Ts'o { 1818d9dda78bSAl Viro return single_open(file, options_seq_show, PDE_DATA(inode)); 181966acdcf4STheodore Ts'o } 182066acdcf4STheodore Ts'o 182166acdcf4STheodore Ts'o static const struct file_operations ext4_seq_options_fops = { 182266acdcf4STheodore Ts'o .owner = THIS_MODULE, 182366acdcf4STheodore Ts'o .open = options_open_fs, 182466acdcf4STheodore Ts'o .read = seq_read, 182566acdcf4STheodore Ts'o .llseek = seq_lseek, 182666acdcf4STheodore Ts'o .release = single_release, 182766acdcf4STheodore Ts'o }; 182866acdcf4STheodore Ts'o 1829617ba13bSMingming Cao static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 1830ac27a0ecSDave Kleikamp int read_only) 1831ac27a0ecSDave Kleikamp { 1832617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 1833ac27a0ecSDave Kleikamp int res = 0; 1834ac27a0ecSDave Kleikamp 1835617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 1836b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "revision level too high, " 1837b31e1552SEric Sandeen "forcing read-only mode"); 1838ac27a0ecSDave Kleikamp res = MS_RDONLY; 1839ac27a0ecSDave Kleikamp } 1840ac27a0ecSDave Kleikamp if (read_only) 1841281b5995STheodore Ts'o goto done; 1842617ba13bSMingming Cao if (!(sbi->s_mount_state & EXT4_VALID_FS)) 1843b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 1844b31e1552SEric Sandeen "running e2fsck is recommended"); 1845617ba13bSMingming Cao else if ((sbi->s_mount_state & EXT4_ERROR_FS)) 1846b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 1847b31e1552SEric Sandeen "warning: mounting fs with errors, " 1848b31e1552SEric Sandeen "running e2fsck is recommended"); 1849ed3ce80aSTao Ma else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 1850ac27a0ecSDave Kleikamp le16_to_cpu(es->s_mnt_count) >= 1851ac27a0ecSDave Kleikamp (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 1852b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 1853b31e1552SEric Sandeen "warning: maximal mount count reached, " 1854b31e1552SEric Sandeen "running e2fsck is recommended"); 1855ac27a0ecSDave Kleikamp else if (le32_to_cpu(es->s_checkinterval) && 1856ac27a0ecSDave Kleikamp (le32_to_cpu(es->s_lastcheck) + 1857ac27a0ecSDave Kleikamp le32_to_cpu(es->s_checkinterval) <= get_seconds())) 1858b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 1859b31e1552SEric Sandeen "warning: checktime reached, " 1860b31e1552SEric Sandeen "running e2fsck is recommended"); 18610390131bSFrank Mayhar if (!sbi->s_journal) 1862216c34b2SMarcin Slusarz es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 1863ac27a0ecSDave Kleikamp if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 1864617ba13bSMingming Cao es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 1865e8546d06SMarcin Slusarz le16_add_cpu(&es->s_mnt_count, 1); 1866ac27a0ecSDave Kleikamp es->s_mtime = cpu_to_le32(get_seconds()); 1867617ba13bSMingming Cao ext4_update_dynamic_rev(sb); 18680390131bSFrank Mayhar if (sbi->s_journal) 1869617ba13bSMingming Cao EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 1870ac27a0ecSDave Kleikamp 1871e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 1872281b5995STheodore Ts'o done: 1873ac27a0ecSDave Kleikamp if (test_opt(sb, DEBUG)) 1874a9df9a49STheodore Ts'o printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 1875a2595b8aSTheodore Ts'o "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 1876ac27a0ecSDave Kleikamp sb->s_blocksize, 1877ac27a0ecSDave Kleikamp sbi->s_groups_count, 1878617ba13bSMingming Cao EXT4_BLOCKS_PER_GROUP(sb), 1879617ba13bSMingming Cao EXT4_INODES_PER_GROUP(sb), 1880a2595b8aSTheodore Ts'o sbi->s_mount_opt, sbi->s_mount_opt2); 1881ac27a0ecSDave Kleikamp 18827abc52c2SDan Magenheimer cleancache_init_fs(sb); 1883ac27a0ecSDave Kleikamp return res; 1884ac27a0ecSDave Kleikamp } 1885ac27a0ecSDave Kleikamp 1886117fff10STheodore Ts'o int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 1887117fff10STheodore Ts'o { 1888117fff10STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 1889117fff10STheodore Ts'o struct flex_groups *new_groups; 1890117fff10STheodore Ts'o int size; 1891117fff10STheodore Ts'o 1892117fff10STheodore Ts'o if (!sbi->s_log_groups_per_flex) 1893117fff10STheodore Ts'o return 0; 1894117fff10STheodore Ts'o 1895117fff10STheodore Ts'o size = ext4_flex_group(sbi, ngroup - 1) + 1; 1896117fff10STheodore Ts'o if (size <= sbi->s_flex_groups_allocated) 1897117fff10STheodore Ts'o return 0; 1898117fff10STheodore Ts'o 1899117fff10STheodore Ts'o size = roundup_pow_of_two(size * sizeof(struct flex_groups)); 1900117fff10STheodore Ts'o new_groups = ext4_kvzalloc(size, GFP_KERNEL); 1901117fff10STheodore Ts'o if (!new_groups) { 1902117fff10STheodore Ts'o ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups", 1903117fff10STheodore Ts'o size / (int) sizeof(struct flex_groups)); 1904117fff10STheodore Ts'o return -ENOMEM; 1905117fff10STheodore Ts'o } 1906117fff10STheodore Ts'o 1907117fff10STheodore Ts'o if (sbi->s_flex_groups) { 1908117fff10STheodore Ts'o memcpy(new_groups, sbi->s_flex_groups, 1909117fff10STheodore Ts'o (sbi->s_flex_groups_allocated * 1910117fff10STheodore Ts'o sizeof(struct flex_groups))); 1911117fff10STheodore Ts'o ext4_kvfree(sbi->s_flex_groups); 1912117fff10STheodore Ts'o } 1913117fff10STheodore Ts'o sbi->s_flex_groups = new_groups; 1914117fff10STheodore Ts'o sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups); 1915117fff10STheodore Ts'o return 0; 1916117fff10STheodore Ts'o } 1917117fff10STheodore Ts'o 1918772cb7c8SJose R. Santos static int ext4_fill_flex_info(struct super_block *sb) 1919772cb7c8SJose R. Santos { 1920772cb7c8SJose R. Santos struct ext4_sb_info *sbi = EXT4_SB(sb); 1921772cb7c8SJose R. Santos struct ext4_group_desc *gdp = NULL; 1922772cb7c8SJose R. Santos ext4_group_t flex_group; 1923117fff10STheodore Ts'o int i, err; 1924772cb7c8SJose R. Santos 1925503358aeSTheodore Ts'o sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 1926d50f2ab6SXi Wang if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 1927772cb7c8SJose R. Santos sbi->s_log_groups_per_flex = 0; 1928772cb7c8SJose R. Santos return 1; 1929772cb7c8SJose R. Santos } 1930772cb7c8SJose R. Santos 1931117fff10STheodore Ts'o err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 1932117fff10STheodore Ts'o if (err) 1933772cb7c8SJose R. Santos goto failed; 1934772cb7c8SJose R. Santos 1935772cb7c8SJose R. Santos for (i = 0; i < sbi->s_groups_count; i++) { 193688b6edd1STheodore Ts'o gdp = ext4_get_group_desc(sb, i, NULL); 1937772cb7c8SJose R. Santos 1938772cb7c8SJose R. Santos flex_group = ext4_flex_group(sbi, i); 19397ad9bb65STheodore Ts'o atomic_add(ext4_free_inodes_count(sb, gdp), 19407ad9bb65STheodore Ts'o &sbi->s_flex_groups[flex_group].free_inodes); 194190ba983fSTheodore Ts'o atomic64_add(ext4_free_group_clusters(sb, gdp), 194224aaa8efSTheodore Ts'o &sbi->s_flex_groups[flex_group].free_clusters); 19437ad9bb65STheodore Ts'o atomic_add(ext4_used_dirs_count(sb, gdp), 19447ad9bb65STheodore Ts'o &sbi->s_flex_groups[flex_group].used_dirs); 1945772cb7c8SJose R. Santos } 1946772cb7c8SJose R. Santos 1947772cb7c8SJose R. Santos return 1; 1948772cb7c8SJose R. Santos failed: 1949772cb7c8SJose R. Santos return 0; 1950772cb7c8SJose R. Santos } 1951772cb7c8SJose R. Santos 1952feb0ab32SDarrick J. Wong static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group, 1953717d50e4SAndreas Dilger struct ext4_group_desc *gdp) 1954717d50e4SAndreas Dilger { 1955feb0ab32SDarrick J. Wong int offset; 1956717d50e4SAndreas Dilger __u16 crc = 0; 1957717d50e4SAndreas Dilger __le32 le_group = cpu_to_le32(block_group); 1958717d50e4SAndreas Dilger 1959feb0ab32SDarrick J. Wong if ((sbi->s_es->s_feature_ro_compat & 1960feb0ab32SDarrick J. Wong cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) { 1961feb0ab32SDarrick J. Wong /* Use new metadata_csum algorithm */ 1962d6a77105STheodore Ts'o __le16 save_csum; 1963feb0ab32SDarrick J. Wong __u32 csum32; 1964feb0ab32SDarrick J. Wong 1965d6a77105STheodore Ts'o save_csum = gdp->bg_checksum; 1966feb0ab32SDarrick J. Wong gdp->bg_checksum = 0; 1967feb0ab32SDarrick J. Wong csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 1968feb0ab32SDarrick J. Wong sizeof(le_group)); 1969feb0ab32SDarrick J. Wong csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, 1970feb0ab32SDarrick J. Wong sbi->s_desc_size); 1971d6a77105STheodore Ts'o gdp->bg_checksum = save_csum; 1972feb0ab32SDarrick J. Wong 1973feb0ab32SDarrick J. Wong crc = csum32 & 0xFFFF; 1974feb0ab32SDarrick J. Wong goto out; 1975feb0ab32SDarrick J. Wong } 1976feb0ab32SDarrick J. Wong 1977feb0ab32SDarrick J. Wong /* old crc16 code */ 1978feb0ab32SDarrick J. Wong offset = offsetof(struct ext4_group_desc, bg_checksum); 1979feb0ab32SDarrick J. Wong 1980717d50e4SAndreas Dilger crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 1981717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 1982717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)gdp, offset); 1983717d50e4SAndreas Dilger offset += sizeof(gdp->bg_checksum); /* skip checksum */ 1984717d50e4SAndreas Dilger /* for checksum of struct ext4_group_desc do the rest...*/ 1985717d50e4SAndreas Dilger if ((sbi->s_es->s_feature_incompat & 1986717d50e4SAndreas Dilger cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) && 1987717d50e4SAndreas Dilger offset < le16_to_cpu(sbi->s_es->s_desc_size)) 1988717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)gdp + offset, 1989717d50e4SAndreas Dilger le16_to_cpu(sbi->s_es->s_desc_size) - 1990717d50e4SAndreas Dilger offset); 1991717d50e4SAndreas Dilger 1992feb0ab32SDarrick J. Wong out: 1993717d50e4SAndreas Dilger return cpu_to_le16(crc); 1994717d50e4SAndreas Dilger } 1995717d50e4SAndreas Dilger 1996feb0ab32SDarrick J. Wong int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 1997717d50e4SAndreas Dilger struct ext4_group_desc *gdp) 1998717d50e4SAndreas Dilger { 1999feb0ab32SDarrick J. Wong if (ext4_has_group_desc_csum(sb) && 2000feb0ab32SDarrick J. Wong (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb), 2001feb0ab32SDarrick J. Wong block_group, gdp))) 2002717d50e4SAndreas Dilger return 0; 2003717d50e4SAndreas Dilger 2004717d50e4SAndreas Dilger return 1; 2005717d50e4SAndreas Dilger } 2006717d50e4SAndreas Dilger 2007feb0ab32SDarrick J. Wong void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2008feb0ab32SDarrick J. Wong struct ext4_group_desc *gdp) 2009feb0ab32SDarrick J. Wong { 2010feb0ab32SDarrick J. Wong if (!ext4_has_group_desc_csum(sb)) 2011feb0ab32SDarrick J. Wong return; 2012feb0ab32SDarrick J. Wong gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp); 2013feb0ab32SDarrick J. Wong } 2014feb0ab32SDarrick J. Wong 2015ac27a0ecSDave Kleikamp /* Called at mount-time, super-block is locked */ 2016bfff6873SLukas Czerner static int ext4_check_descriptors(struct super_block *sb, 2017bfff6873SLukas Czerner ext4_group_t *first_not_zeroed) 2018ac27a0ecSDave Kleikamp { 2019617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 2020617ba13bSMingming Cao ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2021617ba13bSMingming Cao ext4_fsblk_t last_block; 2022bd81d8eeSLaurent Vivier ext4_fsblk_t block_bitmap; 2023bd81d8eeSLaurent Vivier ext4_fsblk_t inode_bitmap; 2024bd81d8eeSLaurent Vivier ext4_fsblk_t inode_table; 2025ce421581SJose R. Santos int flexbg_flag = 0; 2026bfff6873SLukas Czerner ext4_group_t i, grp = sbi->s_groups_count; 2027ac27a0ecSDave Kleikamp 2028ce421581SJose R. Santos if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) 2029ce421581SJose R. Santos flexbg_flag = 1; 2030ce421581SJose R. Santos 2031617ba13bSMingming Cao ext4_debug("Checking group descriptors"); 2032ac27a0ecSDave Kleikamp 2033197cd65aSAkinobu Mita for (i = 0; i < sbi->s_groups_count; i++) { 2034197cd65aSAkinobu Mita struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2035197cd65aSAkinobu Mita 2036ce421581SJose R. Santos if (i == sbi->s_groups_count - 1 || flexbg_flag) 2037bd81d8eeSLaurent Vivier last_block = ext4_blocks_count(sbi->s_es) - 1; 2038ac27a0ecSDave Kleikamp else 2039ac27a0ecSDave Kleikamp last_block = first_block + 2040617ba13bSMingming Cao (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2041ac27a0ecSDave Kleikamp 2042bfff6873SLukas Czerner if ((grp == sbi->s_groups_count) && 2043bfff6873SLukas Czerner !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2044bfff6873SLukas Czerner grp = i; 2045bfff6873SLukas Czerner 20468fadc143SAlexandre Ratchov block_bitmap = ext4_block_bitmap(sb, gdp); 20472b2d6d01STheodore Ts'o if (block_bitmap < first_block || block_bitmap > last_block) { 2048b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2049a9df9a49STheodore Ts'o "Block bitmap for group %u not in group " 2050b31e1552SEric Sandeen "(block %llu)!", i, block_bitmap); 2051ac27a0ecSDave Kleikamp return 0; 2052ac27a0ecSDave Kleikamp } 20538fadc143SAlexandre Ratchov inode_bitmap = ext4_inode_bitmap(sb, gdp); 20542b2d6d01STheodore Ts'o if (inode_bitmap < first_block || inode_bitmap > last_block) { 2055b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2056a9df9a49STheodore Ts'o "Inode bitmap for group %u not in group " 2057b31e1552SEric Sandeen "(block %llu)!", i, inode_bitmap); 2058ac27a0ecSDave Kleikamp return 0; 2059ac27a0ecSDave Kleikamp } 20608fadc143SAlexandre Ratchov inode_table = ext4_inode_table(sb, gdp); 2061bd81d8eeSLaurent Vivier if (inode_table < first_block || 20622b2d6d01STheodore Ts'o inode_table + sbi->s_itb_per_group - 1 > last_block) { 2063b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2064a9df9a49STheodore Ts'o "Inode table for group %u not in group " 2065b31e1552SEric Sandeen "(block %llu)!", i, inode_table); 2066ac27a0ecSDave Kleikamp return 0; 2067ac27a0ecSDave Kleikamp } 2068955ce5f5SAneesh Kumar K.V ext4_lock_group(sb, i); 2069feb0ab32SDarrick J. Wong if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2070b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2071b31e1552SEric Sandeen "Checksum for group %u failed (%u!=%u)", 2072fd2d4291SAvantika Mathur i, le16_to_cpu(ext4_group_desc_csum(sbi, i, 2073fd2d4291SAvantika Mathur gdp)), le16_to_cpu(gdp->bg_checksum)); 20747ee1ec4cSLi Zefan if (!(sb->s_flags & MS_RDONLY)) { 2075955ce5f5SAneesh Kumar K.V ext4_unlock_group(sb, i); 2076717d50e4SAndreas Dilger return 0; 2077717d50e4SAndreas Dilger } 20787ee1ec4cSLi Zefan } 2079955ce5f5SAneesh Kumar K.V ext4_unlock_group(sb, i); 2080ce421581SJose R. Santos if (!flexbg_flag) 2081617ba13bSMingming Cao first_block += EXT4_BLOCKS_PER_GROUP(sb); 2082ac27a0ecSDave Kleikamp } 2083bfff6873SLukas Czerner if (NULL != first_not_zeroed) 2084bfff6873SLukas Czerner *first_not_zeroed = grp; 2085ac27a0ecSDave Kleikamp 20865dee5437STheodore Ts'o ext4_free_blocks_count_set(sbi->s_es, 20875dee5437STheodore Ts'o EXT4_C2B(sbi, ext4_count_free_clusters(sb))); 2088617ba13bSMingming Cao sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb)); 2089ac27a0ecSDave Kleikamp return 1; 2090ac27a0ecSDave Kleikamp } 2091ac27a0ecSDave Kleikamp 2092617ba13bSMingming Cao /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2093ac27a0ecSDave Kleikamp * the superblock) which were deleted from all directories, but held open by 2094ac27a0ecSDave Kleikamp * a process at the time of a crash. We walk the list and try to delete these 2095ac27a0ecSDave Kleikamp * inodes at recovery time (only with a read-write filesystem). 2096ac27a0ecSDave Kleikamp * 2097ac27a0ecSDave Kleikamp * In order to keep the orphan inode chain consistent during traversal (in 2098ac27a0ecSDave Kleikamp * case of crash during recovery), we link each inode into the superblock 2099ac27a0ecSDave Kleikamp * orphan list_head and handle it the same way as an inode deletion during 2100ac27a0ecSDave Kleikamp * normal operation (which journals the operations for us). 2101ac27a0ecSDave Kleikamp * 2102ac27a0ecSDave Kleikamp * We only do an iget() and an iput() on each inode, which is very safe if we 2103ac27a0ecSDave Kleikamp * accidentally point at an in-use or already deleted inode. The worst that 2104ac27a0ecSDave Kleikamp * can happen in this case is that we get a "bit already cleared" message from 2105617ba13bSMingming Cao * ext4_free_inode(). The only reason we would point at a wrong inode is if 2106ac27a0ecSDave Kleikamp * e2fsck was run on this filesystem, and it must have already done the orphan 2107ac27a0ecSDave Kleikamp * inode cleanup for us, so we can safely abort without any further action. 2108ac27a0ecSDave Kleikamp */ 2109617ba13bSMingming Cao static void ext4_orphan_cleanup(struct super_block *sb, 2110617ba13bSMingming Cao struct ext4_super_block *es) 2111ac27a0ecSDave Kleikamp { 2112ac27a0ecSDave Kleikamp unsigned int s_flags = sb->s_flags; 2113ac27a0ecSDave Kleikamp int nr_orphans = 0, nr_truncates = 0; 2114ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 2115ac27a0ecSDave Kleikamp int i; 2116ac27a0ecSDave Kleikamp #endif 2117ac27a0ecSDave Kleikamp if (!es->s_last_orphan) { 2118ac27a0ecSDave Kleikamp jbd_debug(4, "no orphan inodes to clean up\n"); 2119ac27a0ecSDave Kleikamp return; 2120ac27a0ecSDave Kleikamp } 2121ac27a0ecSDave Kleikamp 2122a8f48a95SEric Sandeen if (bdev_read_only(sb->s_bdev)) { 2123b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "write access " 2124b31e1552SEric Sandeen "unavailable, skipping orphan cleanup"); 2125a8f48a95SEric Sandeen return; 2126a8f48a95SEric Sandeen } 2127a8f48a95SEric Sandeen 2128d39195c3SAmir Goldstein /* Check if feature set would not allow a r/w mount */ 2129d39195c3SAmir Goldstein if (!ext4_feature_set_ok(sb, 0)) { 2130d39195c3SAmir Goldstein ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2131d39195c3SAmir Goldstein "unknown ROCOMPAT features"); 2132d39195c3SAmir Goldstein return; 2133d39195c3SAmir Goldstein } 2134d39195c3SAmir Goldstein 2135617ba13bSMingming Cao if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2136c25f9bc6SEric Sandeen /* don't clear list on RO mount w/ errors */ 2137c25f9bc6SEric Sandeen if (es->s_last_orphan && !(s_flags & MS_RDONLY)) { 2138ac27a0ecSDave Kleikamp jbd_debug(1, "Errors on filesystem, " 2139ac27a0ecSDave Kleikamp "clearing orphan list.\n"); 2140ac27a0ecSDave Kleikamp es->s_last_orphan = 0; 2141c25f9bc6SEric Sandeen } 2142ac27a0ecSDave Kleikamp jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2143ac27a0ecSDave Kleikamp return; 2144ac27a0ecSDave Kleikamp } 2145ac27a0ecSDave Kleikamp 2146ac27a0ecSDave Kleikamp if (s_flags & MS_RDONLY) { 2147b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 2148ac27a0ecSDave Kleikamp sb->s_flags &= ~MS_RDONLY; 2149ac27a0ecSDave Kleikamp } 2150ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 2151ac27a0ecSDave Kleikamp /* Needed for iput() to work correctly and not trash data */ 2152ac27a0ecSDave Kleikamp sb->s_flags |= MS_ACTIVE; 2153ac27a0ecSDave Kleikamp /* Turn on quotas so that they are updated correctly */ 2154ac27a0ecSDave Kleikamp for (i = 0; i < MAXQUOTAS; i++) { 2155617ba13bSMingming Cao if (EXT4_SB(sb)->s_qf_names[i]) { 2156617ba13bSMingming Cao int ret = ext4_quota_on_mount(sb, i); 2157ac27a0ecSDave Kleikamp if (ret < 0) 2158b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 2159b31e1552SEric Sandeen "Cannot turn on journaled " 2160b31e1552SEric Sandeen "quota: error %d", ret); 2161ac27a0ecSDave Kleikamp } 2162ac27a0ecSDave Kleikamp } 2163ac27a0ecSDave Kleikamp #endif 2164ac27a0ecSDave Kleikamp 2165ac27a0ecSDave Kleikamp while (es->s_last_orphan) { 2166ac27a0ecSDave Kleikamp struct inode *inode; 2167ac27a0ecSDave Kleikamp 216897bd42b9SJosef Bacik inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 216997bd42b9SJosef Bacik if (IS_ERR(inode)) { 2170ac27a0ecSDave Kleikamp es->s_last_orphan = 0; 2171ac27a0ecSDave Kleikamp break; 2172ac27a0ecSDave Kleikamp } 2173ac27a0ecSDave Kleikamp 2174617ba13bSMingming Cao list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2175871a2931SChristoph Hellwig dquot_initialize(inode); 2176ac27a0ecSDave Kleikamp if (inode->i_nlink) { 2177566370a2SPaul Taysom if (test_opt(sb, DEBUG)) 2178b31e1552SEric Sandeen ext4_msg(sb, KERN_DEBUG, 2179b31e1552SEric Sandeen "%s: truncating inode %lu to %lld bytes", 218046e665e9SHarvey Harrison __func__, inode->i_ino, inode->i_size); 2181e5f8eab8STheodore Ts'o jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2182ac27a0ecSDave Kleikamp inode->i_ino, inode->i_size); 2183721e3ebaSTheodore Ts'o mutex_lock(&inode->i_mutex); 218455f252c9SLukas Czerner truncate_inode_pages(inode->i_mapping, inode->i_size); 2185617ba13bSMingming Cao ext4_truncate(inode); 2186721e3ebaSTheodore Ts'o mutex_unlock(&inode->i_mutex); 2187ac27a0ecSDave Kleikamp nr_truncates++; 2188ac27a0ecSDave Kleikamp } else { 2189566370a2SPaul Taysom if (test_opt(sb, DEBUG)) 2190b31e1552SEric Sandeen ext4_msg(sb, KERN_DEBUG, 2191b31e1552SEric Sandeen "%s: deleting unreferenced inode %lu", 219246e665e9SHarvey Harrison __func__, inode->i_ino); 2193ac27a0ecSDave Kleikamp jbd_debug(2, "deleting unreferenced inode %lu\n", 2194ac27a0ecSDave Kleikamp inode->i_ino); 2195ac27a0ecSDave Kleikamp nr_orphans++; 2196ac27a0ecSDave Kleikamp } 2197ac27a0ecSDave Kleikamp iput(inode); /* The delete magic happens here! */ 2198ac27a0ecSDave Kleikamp } 2199ac27a0ecSDave Kleikamp 2200ac27a0ecSDave Kleikamp #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 2201ac27a0ecSDave Kleikamp 2202ac27a0ecSDave Kleikamp if (nr_orphans) 2203b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 2204b31e1552SEric Sandeen PLURAL(nr_orphans)); 2205ac27a0ecSDave Kleikamp if (nr_truncates) 2206b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 2207b31e1552SEric Sandeen PLURAL(nr_truncates)); 2208ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 2209ac27a0ecSDave Kleikamp /* Turn quotas off */ 2210ac27a0ecSDave Kleikamp for (i = 0; i < MAXQUOTAS; i++) { 2211ac27a0ecSDave Kleikamp if (sb_dqopt(sb)->files[i]) 2212287a8095SChristoph Hellwig dquot_quota_off(sb, i); 2213ac27a0ecSDave Kleikamp } 2214ac27a0ecSDave Kleikamp #endif 2215ac27a0ecSDave Kleikamp sb->s_flags = s_flags; /* Restore MS_RDONLY status */ 2216ac27a0ecSDave Kleikamp } 22170b8e58a1SAndreas Dilger 2218cd2291a4SEric Sandeen /* 2219cd2291a4SEric Sandeen * Maximal extent format file size. 2220cd2291a4SEric Sandeen * Resulting logical blkno at s_maxbytes must fit in our on-disk 2221cd2291a4SEric Sandeen * extent format containers, within a sector_t, and within i_blocks 2222cd2291a4SEric Sandeen * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2223cd2291a4SEric Sandeen * so that won't be a limiting factor. 2224cd2291a4SEric Sandeen * 2225f17722f9SLukas Czerner * However there is other limiting factor. We do store extents in the form 2226f17722f9SLukas Czerner * of starting block and length, hence the resulting length of the extent 2227f17722f9SLukas Czerner * covering maximum file size must fit into on-disk format containers as 2228f17722f9SLukas Czerner * well. Given that length is always by 1 unit bigger than max unit (because 2229f17722f9SLukas Czerner * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2230f17722f9SLukas Czerner * 2231cd2291a4SEric Sandeen * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2232cd2291a4SEric Sandeen */ 2233f287a1a5STheodore Ts'o static loff_t ext4_max_size(int blkbits, int has_huge_files) 2234cd2291a4SEric Sandeen { 2235cd2291a4SEric Sandeen loff_t res; 2236cd2291a4SEric Sandeen loff_t upper_limit = MAX_LFS_FILESIZE; 2237cd2291a4SEric Sandeen 2238cd2291a4SEric Sandeen /* small i_blocks in vfs inode? */ 2239f287a1a5STheodore Ts'o if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 2240cd2291a4SEric Sandeen /* 224190c699a9SBartlomiej Zolnierkiewicz * CONFIG_LBDAF is not enabled implies the inode 2242cd2291a4SEric Sandeen * i_block represent total blocks in 512 bytes 2243cd2291a4SEric Sandeen * 32 == size of vfs inode i_blocks * 8 2244cd2291a4SEric Sandeen */ 2245cd2291a4SEric Sandeen upper_limit = (1LL << 32) - 1; 2246cd2291a4SEric Sandeen 2247cd2291a4SEric Sandeen /* total blocks in file system block size */ 2248cd2291a4SEric Sandeen upper_limit >>= (blkbits - 9); 2249cd2291a4SEric Sandeen upper_limit <<= blkbits; 2250cd2291a4SEric Sandeen } 2251cd2291a4SEric Sandeen 2252f17722f9SLukas Czerner /* 2253f17722f9SLukas Czerner * 32-bit extent-start container, ee_block. We lower the maxbytes 2254f17722f9SLukas Czerner * by one fs block, so ee_len can cover the extent of maximum file 2255f17722f9SLukas Czerner * size 2256f17722f9SLukas Czerner */ 2257f17722f9SLukas Czerner res = (1LL << 32) - 1; 2258cd2291a4SEric Sandeen res <<= blkbits; 2259cd2291a4SEric Sandeen 2260cd2291a4SEric Sandeen /* Sanity check against vm- & vfs- imposed limits */ 2261cd2291a4SEric Sandeen if (res > upper_limit) 2262cd2291a4SEric Sandeen res = upper_limit; 2263cd2291a4SEric Sandeen 2264cd2291a4SEric Sandeen return res; 2265cd2291a4SEric Sandeen } 2266ac27a0ecSDave Kleikamp 2267ac27a0ecSDave Kleikamp /* 2268cd2291a4SEric Sandeen * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 22690fc1b451SAneesh Kumar K.V * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 22700fc1b451SAneesh Kumar K.V * We need to be 1 filesystem block less than the 2^48 sector limit. 2271ac27a0ecSDave Kleikamp */ 2272f287a1a5STheodore Ts'o static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 2273ac27a0ecSDave Kleikamp { 2274617ba13bSMingming Cao loff_t res = EXT4_NDIR_BLOCKS; 22750fc1b451SAneesh Kumar K.V int meta_blocks; 22760fc1b451SAneesh Kumar K.V loff_t upper_limit; 22770b8e58a1SAndreas Dilger /* This is calculated to be the largest file size for a dense, block 22780b8e58a1SAndreas Dilger * mapped file such that the file's total number of 512-byte sectors, 22790b8e58a1SAndreas Dilger * including data and all indirect blocks, does not exceed (2^48 - 1). 22800b8e58a1SAndreas Dilger * 22810b8e58a1SAndreas Dilger * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 22820b8e58a1SAndreas Dilger * number of 512-byte sectors of the file. 22830fc1b451SAneesh Kumar K.V */ 22840fc1b451SAneesh Kumar K.V 2285f287a1a5STheodore Ts'o if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) { 22860fc1b451SAneesh Kumar K.V /* 228790c699a9SBartlomiej Zolnierkiewicz * !has_huge_files or CONFIG_LBDAF not enabled implies that 22880b8e58a1SAndreas Dilger * the inode i_block field represents total file blocks in 22890b8e58a1SAndreas Dilger * 2^32 512-byte sectors == size of vfs inode i_blocks * 8 22900fc1b451SAneesh Kumar K.V */ 22910fc1b451SAneesh Kumar K.V upper_limit = (1LL << 32) - 1; 22920fc1b451SAneesh Kumar K.V 22930fc1b451SAneesh Kumar K.V /* total blocks in file system block size */ 22940fc1b451SAneesh Kumar K.V upper_limit >>= (bits - 9); 22950fc1b451SAneesh Kumar K.V 22960fc1b451SAneesh Kumar K.V } else { 22978180a562SAneesh Kumar K.V /* 22988180a562SAneesh Kumar K.V * We use 48 bit ext4_inode i_blocks 22998180a562SAneesh Kumar K.V * With EXT4_HUGE_FILE_FL set the i_blocks 23008180a562SAneesh Kumar K.V * represent total number of blocks in 23018180a562SAneesh Kumar K.V * file system block size 23028180a562SAneesh Kumar K.V */ 23030fc1b451SAneesh Kumar K.V upper_limit = (1LL << 48) - 1; 23040fc1b451SAneesh Kumar K.V 23050fc1b451SAneesh Kumar K.V } 23060fc1b451SAneesh Kumar K.V 23070fc1b451SAneesh Kumar K.V /* indirect blocks */ 23080fc1b451SAneesh Kumar K.V meta_blocks = 1; 23090fc1b451SAneesh Kumar K.V /* double indirect blocks */ 23100fc1b451SAneesh Kumar K.V meta_blocks += 1 + (1LL << (bits-2)); 23110fc1b451SAneesh Kumar K.V /* tripple indirect blocks */ 23120fc1b451SAneesh Kumar K.V meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 23130fc1b451SAneesh Kumar K.V 23140fc1b451SAneesh Kumar K.V upper_limit -= meta_blocks; 23150fc1b451SAneesh Kumar K.V upper_limit <<= bits; 2316ac27a0ecSDave Kleikamp 2317ac27a0ecSDave Kleikamp res += 1LL << (bits-2); 2318ac27a0ecSDave Kleikamp res += 1LL << (2*(bits-2)); 2319ac27a0ecSDave Kleikamp res += 1LL << (3*(bits-2)); 2320ac27a0ecSDave Kleikamp res <<= bits; 2321ac27a0ecSDave Kleikamp if (res > upper_limit) 2322ac27a0ecSDave Kleikamp res = upper_limit; 23230fc1b451SAneesh Kumar K.V 23240fc1b451SAneesh Kumar K.V if (res > MAX_LFS_FILESIZE) 23250fc1b451SAneesh Kumar K.V res = MAX_LFS_FILESIZE; 23260fc1b451SAneesh Kumar K.V 2327ac27a0ecSDave Kleikamp return res; 2328ac27a0ecSDave Kleikamp } 2329ac27a0ecSDave Kleikamp 2330617ba13bSMingming Cao static ext4_fsblk_t descriptor_loc(struct super_block *sb, 233170bbb3e0SAndrew Morton ext4_fsblk_t logical_sb_block, int nr) 2332ac27a0ecSDave Kleikamp { 2333617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 2334fd2d4291SAvantika Mathur ext4_group_t bg, first_meta_bg; 2335ac27a0ecSDave Kleikamp int has_super = 0; 2336ac27a0ecSDave Kleikamp 2337ac27a0ecSDave Kleikamp first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 2338ac27a0ecSDave Kleikamp 2339617ba13bSMingming Cao if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) || 2340ac27a0ecSDave Kleikamp nr < first_meta_bg) 234170bbb3e0SAndrew Morton return logical_sb_block + nr + 1; 2342ac27a0ecSDave Kleikamp bg = sbi->s_desc_per_block * nr; 2343617ba13bSMingming Cao if (ext4_bg_has_super(sb, bg)) 2344ac27a0ecSDave Kleikamp has_super = 1; 23450b8e58a1SAndreas Dilger 2346617ba13bSMingming Cao return (has_super + ext4_group_first_block_no(sb, bg)); 2347ac27a0ecSDave Kleikamp } 2348ac27a0ecSDave Kleikamp 2349c9de560dSAlex Tomas /** 2350c9de560dSAlex Tomas * ext4_get_stripe_size: Get the stripe size. 2351c9de560dSAlex Tomas * @sbi: In memory super block info 2352c9de560dSAlex Tomas * 2353c9de560dSAlex Tomas * If we have specified it via mount option, then 2354c9de560dSAlex Tomas * use the mount option value. If the value specified at mount time is 2355c9de560dSAlex Tomas * greater than the blocks per group use the super block value. 2356c9de560dSAlex Tomas * If the super block value is greater than blocks per group return 0. 2357c9de560dSAlex Tomas * Allocator needs it be less than blocks per group. 2358c9de560dSAlex Tomas * 2359c9de560dSAlex Tomas */ 2360c9de560dSAlex Tomas static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 2361c9de560dSAlex Tomas { 2362c9de560dSAlex Tomas unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 2363c9de560dSAlex Tomas unsigned long stripe_width = 2364c9de560dSAlex Tomas le32_to_cpu(sbi->s_es->s_raid_stripe_width); 23653eb08658SDan Ehrenberg int ret; 2366c9de560dSAlex Tomas 2367c9de560dSAlex Tomas if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 23683eb08658SDan Ehrenberg ret = sbi->s_stripe; 23693eb08658SDan Ehrenberg else if (stripe_width <= sbi->s_blocks_per_group) 23703eb08658SDan Ehrenberg ret = stripe_width; 23713eb08658SDan Ehrenberg else if (stride <= sbi->s_blocks_per_group) 23723eb08658SDan Ehrenberg ret = stride; 23733eb08658SDan Ehrenberg else 23743eb08658SDan Ehrenberg ret = 0; 2375c9de560dSAlex Tomas 23763eb08658SDan Ehrenberg /* 23773eb08658SDan Ehrenberg * If the stripe width is 1, this makes no sense and 23783eb08658SDan Ehrenberg * we set it to 0 to turn off stripe handling code. 23793eb08658SDan Ehrenberg */ 23803eb08658SDan Ehrenberg if (ret <= 1) 23813eb08658SDan Ehrenberg ret = 0; 2382c9de560dSAlex Tomas 23833eb08658SDan Ehrenberg return ret; 2384c9de560dSAlex Tomas } 2385ac27a0ecSDave Kleikamp 23863197ebdbSTheodore Ts'o /* sysfs supprt */ 23873197ebdbSTheodore Ts'o 23883197ebdbSTheodore Ts'o struct ext4_attr { 23893197ebdbSTheodore Ts'o struct attribute attr; 23903197ebdbSTheodore Ts'o ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *); 23913197ebdbSTheodore Ts'o ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *, 23923197ebdbSTheodore Ts'o const char *, size_t); 2393f2d50a65SJan Kara union { 23943197ebdbSTheodore Ts'o int offset; 2395f2d50a65SJan Kara int deprecated_val; 2396f2d50a65SJan Kara } u; 23973197ebdbSTheodore Ts'o }; 23983197ebdbSTheodore Ts'o 239927dd4385SLukas Czerner static int parse_strtoull(const char *buf, 240027dd4385SLukas Czerner unsigned long long max, unsigned long long *value) 240127dd4385SLukas Czerner { 240227dd4385SLukas Czerner int ret; 240327dd4385SLukas Czerner 240427dd4385SLukas Czerner ret = kstrtoull(skip_spaces(buf), 0, value); 240527dd4385SLukas Czerner if (!ret && *value > max) 240627dd4385SLukas Czerner ret = -EINVAL; 240727dd4385SLukas Czerner return ret; 240827dd4385SLukas Czerner } 240927dd4385SLukas Czerner 24103197ebdbSTheodore Ts'o static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a, 24113197ebdbSTheodore Ts'o struct ext4_sb_info *sbi, 24123197ebdbSTheodore Ts'o char *buf) 24133197ebdbSTheodore Ts'o { 24143197ebdbSTheodore Ts'o return snprintf(buf, PAGE_SIZE, "%llu\n", 24157b415bf6SAditya Kali (s64) EXT4_C2B(sbi, 24167b415bf6SAditya Kali percpu_counter_sum(&sbi->s_dirtyclusters_counter))); 24173197ebdbSTheodore Ts'o } 24183197ebdbSTheodore Ts'o 24193197ebdbSTheodore Ts'o static ssize_t session_write_kbytes_show(struct ext4_attr *a, 24203197ebdbSTheodore Ts'o struct ext4_sb_info *sbi, char *buf) 24213197ebdbSTheodore Ts'o { 24223197ebdbSTheodore Ts'o struct super_block *sb = sbi->s_buddy_cache->i_sb; 24233197ebdbSTheodore Ts'o 2424f613dfcbSTheodore Ts'o if (!sb->s_bdev->bd_part) 2425f613dfcbSTheodore Ts'o return snprintf(buf, PAGE_SIZE, "0\n"); 24263197ebdbSTheodore Ts'o return snprintf(buf, PAGE_SIZE, "%lu\n", 24273197ebdbSTheodore Ts'o (part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 24283197ebdbSTheodore Ts'o sbi->s_sectors_written_start) >> 1); 24293197ebdbSTheodore Ts'o } 24303197ebdbSTheodore Ts'o 24313197ebdbSTheodore Ts'o static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a, 24323197ebdbSTheodore Ts'o struct ext4_sb_info *sbi, char *buf) 24333197ebdbSTheodore Ts'o { 24343197ebdbSTheodore Ts'o struct super_block *sb = sbi->s_buddy_cache->i_sb; 24353197ebdbSTheodore Ts'o 2436f613dfcbSTheodore Ts'o if (!sb->s_bdev->bd_part) 2437f613dfcbSTheodore Ts'o return snprintf(buf, PAGE_SIZE, "0\n"); 24383197ebdbSTheodore Ts'o return snprintf(buf, PAGE_SIZE, "%llu\n", 2439a6b43e38SAndrew Morton (unsigned long long)(sbi->s_kbytes_written + 24403197ebdbSTheodore Ts'o ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 2441a6b43e38SAndrew Morton EXT4_SB(sb)->s_sectors_written_start) >> 1))); 24423197ebdbSTheodore Ts'o } 24433197ebdbSTheodore Ts'o 24443197ebdbSTheodore Ts'o static ssize_t inode_readahead_blks_store(struct ext4_attr *a, 24453197ebdbSTheodore Ts'o struct ext4_sb_info *sbi, 24463197ebdbSTheodore Ts'o const char *buf, size_t count) 24473197ebdbSTheodore Ts'o { 24483197ebdbSTheodore Ts'o unsigned long t; 2449e1091b15SLukas Czerner int ret; 24503197ebdbSTheodore Ts'o 2451e1091b15SLukas Czerner ret = kstrtoul(skip_spaces(buf), 0, &t); 2452e1091b15SLukas Czerner if (ret) 2453e1091b15SLukas Czerner return ret; 24543197ebdbSTheodore Ts'o 2455e1091b15SLukas Czerner if (t && (!is_power_of_2(t) || t > 0x40000000)) 24563197ebdbSTheodore Ts'o return -EINVAL; 24573197ebdbSTheodore Ts'o 24583197ebdbSTheodore Ts'o sbi->s_inode_readahead_blks = t; 24593197ebdbSTheodore Ts'o return count; 24603197ebdbSTheodore Ts'o } 24613197ebdbSTheodore Ts'o 24623197ebdbSTheodore Ts'o static ssize_t sbi_ui_show(struct ext4_attr *a, 24633197ebdbSTheodore Ts'o struct ext4_sb_info *sbi, char *buf) 24643197ebdbSTheodore Ts'o { 2465f2d50a65SJan Kara unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset); 24663197ebdbSTheodore Ts'o 24673197ebdbSTheodore Ts'o return snprintf(buf, PAGE_SIZE, "%u\n", *ui); 24683197ebdbSTheodore Ts'o } 24693197ebdbSTheodore Ts'o 24703197ebdbSTheodore Ts'o static ssize_t sbi_ui_store(struct ext4_attr *a, 24713197ebdbSTheodore Ts'o struct ext4_sb_info *sbi, 24723197ebdbSTheodore Ts'o const char *buf, size_t count) 24733197ebdbSTheodore Ts'o { 2474f2d50a65SJan Kara unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset); 24753197ebdbSTheodore Ts'o unsigned long t; 2476e1091b15SLukas Czerner int ret; 24773197ebdbSTheodore Ts'o 2478e1091b15SLukas Czerner ret = kstrtoul(skip_spaces(buf), 0, &t); 2479e1091b15SLukas Czerner if (ret) 2480e1091b15SLukas Czerner return ret; 24813197ebdbSTheodore Ts'o *ui = t; 24823197ebdbSTheodore Ts'o return count; 24833197ebdbSTheodore Ts'o } 24843197ebdbSTheodore Ts'o 248527dd4385SLukas Czerner static ssize_t reserved_clusters_show(struct ext4_attr *a, 248627dd4385SLukas Czerner struct ext4_sb_info *sbi, char *buf) 248727dd4385SLukas Czerner { 248827dd4385SLukas Czerner return snprintf(buf, PAGE_SIZE, "%llu\n", 248927dd4385SLukas Czerner (unsigned long long) atomic64_read(&sbi->s_resv_clusters)); 249027dd4385SLukas Czerner } 249127dd4385SLukas Czerner 249227dd4385SLukas Czerner static ssize_t reserved_clusters_store(struct ext4_attr *a, 249327dd4385SLukas Czerner struct ext4_sb_info *sbi, 249427dd4385SLukas Czerner const char *buf, size_t count) 249527dd4385SLukas Czerner { 249627dd4385SLukas Czerner unsigned long long val; 249727dd4385SLukas Czerner int ret; 249827dd4385SLukas Czerner 249927dd4385SLukas Czerner if (parse_strtoull(buf, -1ULL, &val)) 250027dd4385SLukas Czerner return -EINVAL; 250127dd4385SLukas Czerner ret = ext4_reserve_clusters(sbi, val); 250227dd4385SLukas Czerner 250327dd4385SLukas Czerner return ret ? ret : count; 250427dd4385SLukas Czerner } 250527dd4385SLukas Czerner 25062c0544b2STheodore Ts'o static ssize_t trigger_test_error(struct ext4_attr *a, 25072c0544b2STheodore Ts'o struct ext4_sb_info *sbi, 25082c0544b2STheodore Ts'o const char *buf, size_t count) 25092c0544b2STheodore Ts'o { 25102c0544b2STheodore Ts'o int len = count; 25112c0544b2STheodore Ts'o 25122c0544b2STheodore Ts'o if (!capable(CAP_SYS_ADMIN)) 25132c0544b2STheodore Ts'o return -EPERM; 25142c0544b2STheodore Ts'o 25152c0544b2STheodore Ts'o if (len && buf[len-1] == '\n') 25162c0544b2STheodore Ts'o len--; 25172c0544b2STheodore Ts'o 25182c0544b2STheodore Ts'o if (len) 25192c0544b2STheodore Ts'o ext4_error(sbi->s_sb, "%.*s", len, buf); 25202c0544b2STheodore Ts'o return count; 25212c0544b2STheodore Ts'o } 25222c0544b2STheodore Ts'o 2523f2d50a65SJan Kara static ssize_t sbi_deprecated_show(struct ext4_attr *a, 2524f2d50a65SJan Kara struct ext4_sb_info *sbi, char *buf) 2525f2d50a65SJan Kara { 2526f2d50a65SJan Kara return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val); 2527f2d50a65SJan Kara } 2528f2d50a65SJan Kara 25293197ebdbSTheodore Ts'o #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \ 25303197ebdbSTheodore Ts'o static struct ext4_attr ext4_attr_##_name = { \ 25313197ebdbSTheodore Ts'o .attr = {.name = __stringify(_name), .mode = _mode }, \ 25323197ebdbSTheodore Ts'o .show = _show, \ 25333197ebdbSTheodore Ts'o .store = _store, \ 2534f2d50a65SJan Kara .u = { \ 25353197ebdbSTheodore Ts'o .offset = offsetof(struct ext4_sb_info, _elname),\ 2536f2d50a65SJan Kara }, \ 25373197ebdbSTheodore Ts'o } 25383197ebdbSTheodore Ts'o #define EXT4_ATTR(name, mode, show, store) \ 25393197ebdbSTheodore Ts'o static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store) 25403197ebdbSTheodore Ts'o 2541857ac889SLukas Czerner #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL) 25423197ebdbSTheodore Ts'o #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL) 25433197ebdbSTheodore Ts'o #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store) 25443197ebdbSTheodore Ts'o #define EXT4_RW_ATTR_SBI_UI(name, elname) \ 25453197ebdbSTheodore Ts'o EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname) 25463197ebdbSTheodore Ts'o #define ATTR_LIST(name) &ext4_attr_##name.attr 2547f2d50a65SJan Kara #define EXT4_DEPRECATED_ATTR(_name, _val) \ 2548f2d50a65SJan Kara static struct ext4_attr ext4_attr_##_name = { \ 2549f2d50a65SJan Kara .attr = {.name = __stringify(_name), .mode = 0444 }, \ 2550f2d50a65SJan Kara .show = sbi_deprecated_show, \ 2551f2d50a65SJan Kara .u = { \ 2552f2d50a65SJan Kara .deprecated_val = _val, \ 2553f2d50a65SJan Kara }, \ 2554f2d50a65SJan Kara } 25553197ebdbSTheodore Ts'o 25563197ebdbSTheodore Ts'o EXT4_RO_ATTR(delayed_allocation_blocks); 25573197ebdbSTheodore Ts'o EXT4_RO_ATTR(session_write_kbytes); 25583197ebdbSTheodore Ts'o EXT4_RO_ATTR(lifetime_write_kbytes); 255927dd4385SLukas Czerner EXT4_RW_ATTR(reserved_clusters); 25603197ebdbSTheodore Ts'o EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show, 25613197ebdbSTheodore Ts'o inode_readahead_blks_store, s_inode_readahead_blks); 256211013911SAndreas Dilger EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal); 25633197ebdbSTheodore Ts'o EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats); 25643197ebdbSTheodore Ts'o EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan); 25653197ebdbSTheodore Ts'o EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan); 25663197ebdbSTheodore Ts'o EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs); 25673197ebdbSTheodore Ts'o EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request); 25683197ebdbSTheodore Ts'o EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc); 2569f2d50a65SJan Kara EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128); 257067a5da56SZheng Liu EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb); 25712c0544b2STheodore Ts'o EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error); 25723197ebdbSTheodore Ts'o 25733197ebdbSTheodore Ts'o static struct attribute *ext4_attrs[] = { 25743197ebdbSTheodore Ts'o ATTR_LIST(delayed_allocation_blocks), 25753197ebdbSTheodore Ts'o ATTR_LIST(session_write_kbytes), 25763197ebdbSTheodore Ts'o ATTR_LIST(lifetime_write_kbytes), 257727dd4385SLukas Czerner ATTR_LIST(reserved_clusters), 25783197ebdbSTheodore Ts'o ATTR_LIST(inode_readahead_blks), 257911013911SAndreas Dilger ATTR_LIST(inode_goal), 25803197ebdbSTheodore Ts'o ATTR_LIST(mb_stats), 25813197ebdbSTheodore Ts'o ATTR_LIST(mb_max_to_scan), 25823197ebdbSTheodore Ts'o ATTR_LIST(mb_min_to_scan), 25833197ebdbSTheodore Ts'o ATTR_LIST(mb_order2_req), 25843197ebdbSTheodore Ts'o ATTR_LIST(mb_stream_req), 25853197ebdbSTheodore Ts'o ATTR_LIST(mb_group_prealloc), 258655138e0bSTheodore Ts'o ATTR_LIST(max_writeback_mb_bump), 258767a5da56SZheng Liu ATTR_LIST(extent_max_zeroout_kb), 25882c0544b2STheodore Ts'o ATTR_LIST(trigger_fs_error), 25893197ebdbSTheodore Ts'o NULL, 25903197ebdbSTheodore Ts'o }; 25913197ebdbSTheodore Ts'o 2592857ac889SLukas Czerner /* Features this copy of ext4 supports */ 2593857ac889SLukas Czerner EXT4_INFO_ATTR(lazy_itable_init); 259427ee40dfSLukas Czerner EXT4_INFO_ATTR(batched_discard); 25955e7bbef1STheodore Ts'o EXT4_INFO_ATTR(meta_bg_resize); 2596857ac889SLukas Czerner 2597857ac889SLukas Czerner static struct attribute *ext4_feat_attrs[] = { 2598857ac889SLukas Czerner ATTR_LIST(lazy_itable_init), 259927ee40dfSLukas Czerner ATTR_LIST(batched_discard), 26005e7bbef1STheodore Ts'o ATTR_LIST(meta_bg_resize), 2601857ac889SLukas Czerner NULL, 2602857ac889SLukas Czerner }; 2603857ac889SLukas Czerner 26043197ebdbSTheodore Ts'o static ssize_t ext4_attr_show(struct kobject *kobj, 26053197ebdbSTheodore Ts'o struct attribute *attr, char *buf) 26063197ebdbSTheodore Ts'o { 26073197ebdbSTheodore Ts'o struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 26083197ebdbSTheodore Ts'o s_kobj); 26093197ebdbSTheodore Ts'o struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); 26103197ebdbSTheodore Ts'o 26113197ebdbSTheodore Ts'o return a->show ? a->show(a, sbi, buf) : 0; 26123197ebdbSTheodore Ts'o } 26133197ebdbSTheodore Ts'o 26143197ebdbSTheodore Ts'o static ssize_t ext4_attr_store(struct kobject *kobj, 26153197ebdbSTheodore Ts'o struct attribute *attr, 26163197ebdbSTheodore Ts'o const char *buf, size_t len) 26173197ebdbSTheodore Ts'o { 26183197ebdbSTheodore Ts'o struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 26193197ebdbSTheodore Ts'o s_kobj); 26203197ebdbSTheodore Ts'o struct ext4_attr *a = container_of(attr, struct ext4_attr, attr); 26213197ebdbSTheodore Ts'o 26223197ebdbSTheodore Ts'o return a->store ? a->store(a, sbi, buf, len) : 0; 26233197ebdbSTheodore Ts'o } 26243197ebdbSTheodore Ts'o 26253197ebdbSTheodore Ts'o static void ext4_sb_release(struct kobject *kobj) 26263197ebdbSTheodore Ts'o { 26273197ebdbSTheodore Ts'o struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info, 26283197ebdbSTheodore Ts'o s_kobj); 26293197ebdbSTheodore Ts'o complete(&sbi->s_kobj_unregister); 26303197ebdbSTheodore Ts'o } 26313197ebdbSTheodore Ts'o 263252cf25d0SEmese Revfy static const struct sysfs_ops ext4_attr_ops = { 26333197ebdbSTheodore Ts'o .show = ext4_attr_show, 26343197ebdbSTheodore Ts'o .store = ext4_attr_store, 26353197ebdbSTheodore Ts'o }; 26363197ebdbSTheodore Ts'o 26373197ebdbSTheodore Ts'o static struct kobj_type ext4_ktype = { 26383197ebdbSTheodore Ts'o .default_attrs = ext4_attrs, 26393197ebdbSTheodore Ts'o .sysfs_ops = &ext4_attr_ops, 26403197ebdbSTheodore Ts'o .release = ext4_sb_release, 26413197ebdbSTheodore Ts'o }; 26423197ebdbSTheodore Ts'o 2643857ac889SLukas Czerner static void ext4_feat_release(struct kobject *kobj) 2644857ac889SLukas Czerner { 2645857ac889SLukas Czerner complete(&ext4_feat->f_kobj_unregister); 2646857ac889SLukas Czerner } 2647857ac889SLukas Czerner 2648857ac889SLukas Czerner static struct kobj_type ext4_feat_ktype = { 2649857ac889SLukas Czerner .default_attrs = ext4_feat_attrs, 2650857ac889SLukas Czerner .sysfs_ops = &ext4_attr_ops, 2651857ac889SLukas Czerner .release = ext4_feat_release, 2652857ac889SLukas Czerner }; 2653857ac889SLukas Czerner 2654a13fb1a4SEric Sandeen /* 2655a13fb1a4SEric Sandeen * Check whether this filesystem can be mounted based on 2656a13fb1a4SEric Sandeen * the features present and the RDONLY/RDWR mount requested. 2657a13fb1a4SEric Sandeen * Returns 1 if this filesystem can be mounted as requested, 2658a13fb1a4SEric Sandeen * 0 if it cannot be. 2659a13fb1a4SEric Sandeen */ 2660a13fb1a4SEric Sandeen static int ext4_feature_set_ok(struct super_block *sb, int readonly) 2661a13fb1a4SEric Sandeen { 2662a13fb1a4SEric Sandeen if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) { 2663a13fb1a4SEric Sandeen ext4_msg(sb, KERN_ERR, 2664a13fb1a4SEric Sandeen "Couldn't mount because of " 2665a13fb1a4SEric Sandeen "unsupported optional features (%x)", 2666a13fb1a4SEric Sandeen (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 2667a13fb1a4SEric Sandeen ~EXT4_FEATURE_INCOMPAT_SUPP)); 2668a13fb1a4SEric Sandeen return 0; 2669a13fb1a4SEric Sandeen } 2670a13fb1a4SEric Sandeen 2671a13fb1a4SEric Sandeen if (readonly) 2672a13fb1a4SEric Sandeen return 1; 2673a13fb1a4SEric Sandeen 2674a13fb1a4SEric Sandeen /* Check that feature set is OK for a read-write mount */ 2675a13fb1a4SEric Sandeen if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) { 2676a13fb1a4SEric Sandeen ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 2677a13fb1a4SEric Sandeen "unsupported optional features (%x)", 2678a13fb1a4SEric Sandeen (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 2679a13fb1a4SEric Sandeen ~EXT4_FEATURE_RO_COMPAT_SUPP)); 2680a13fb1a4SEric Sandeen return 0; 2681a13fb1a4SEric Sandeen } 2682a13fb1a4SEric Sandeen /* 2683a13fb1a4SEric Sandeen * Large file size enabled file system can only be mounted 2684a13fb1a4SEric Sandeen * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF 2685a13fb1a4SEric Sandeen */ 2686a13fb1a4SEric Sandeen if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) { 2687a13fb1a4SEric Sandeen if (sizeof(blkcnt_t) < sizeof(u64)) { 2688a13fb1a4SEric Sandeen ext4_msg(sb, KERN_ERR, "Filesystem with huge files " 2689a13fb1a4SEric Sandeen "cannot be mounted RDWR without " 2690a13fb1a4SEric Sandeen "CONFIG_LBDAF"); 2691a13fb1a4SEric Sandeen return 0; 2692a13fb1a4SEric Sandeen } 2693a13fb1a4SEric Sandeen } 2694bab08ab9STheodore Ts'o if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) && 2695bab08ab9STheodore Ts'o !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) { 2696bab08ab9STheodore Ts'o ext4_msg(sb, KERN_ERR, 2697bab08ab9STheodore Ts'o "Can't support bigalloc feature without " 2698bab08ab9STheodore Ts'o "extents feature\n"); 2699bab08ab9STheodore Ts'o return 0; 2700bab08ab9STheodore Ts'o } 27017c319d32SAditya Kali 27027c319d32SAditya Kali #ifndef CONFIG_QUOTA 27037c319d32SAditya Kali if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) && 27047c319d32SAditya Kali !readonly) { 27057c319d32SAditya Kali ext4_msg(sb, KERN_ERR, 27067c319d32SAditya Kali "Filesystem with quota feature cannot be mounted RDWR " 27077c319d32SAditya Kali "without CONFIG_QUOTA"); 27087c319d32SAditya Kali return 0; 27097c319d32SAditya Kali } 27107c319d32SAditya Kali #endif /* CONFIG_QUOTA */ 2711a13fb1a4SEric Sandeen return 1; 2712a13fb1a4SEric Sandeen } 2713a13fb1a4SEric Sandeen 271466e61a9eSTheodore Ts'o /* 271566e61a9eSTheodore Ts'o * This function is called once a day if we have errors logged 271666e61a9eSTheodore Ts'o * on the file system 271766e61a9eSTheodore Ts'o */ 271866e61a9eSTheodore Ts'o static void print_daily_error_info(unsigned long arg) 271966e61a9eSTheodore Ts'o { 272066e61a9eSTheodore Ts'o struct super_block *sb = (struct super_block *) arg; 272166e61a9eSTheodore Ts'o struct ext4_sb_info *sbi; 272266e61a9eSTheodore Ts'o struct ext4_super_block *es; 272366e61a9eSTheodore Ts'o 272466e61a9eSTheodore Ts'o sbi = EXT4_SB(sb); 272566e61a9eSTheodore Ts'o es = sbi->s_es; 272666e61a9eSTheodore Ts'o 272766e61a9eSTheodore Ts'o if (es->s_error_count) 272866e61a9eSTheodore Ts'o ext4_msg(sb, KERN_NOTICE, "error count: %u", 272966e61a9eSTheodore Ts'o le32_to_cpu(es->s_error_count)); 273066e61a9eSTheodore Ts'o if (es->s_first_error_time) { 273166e61a9eSTheodore Ts'o printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d", 273266e61a9eSTheodore Ts'o sb->s_id, le32_to_cpu(es->s_first_error_time), 273366e61a9eSTheodore Ts'o (int) sizeof(es->s_first_error_func), 273466e61a9eSTheodore Ts'o es->s_first_error_func, 273566e61a9eSTheodore Ts'o le32_to_cpu(es->s_first_error_line)); 273666e61a9eSTheodore Ts'o if (es->s_first_error_ino) 273766e61a9eSTheodore Ts'o printk(": inode %u", 273866e61a9eSTheodore Ts'o le32_to_cpu(es->s_first_error_ino)); 273966e61a9eSTheodore Ts'o if (es->s_first_error_block) 274066e61a9eSTheodore Ts'o printk(": block %llu", (unsigned long long) 274166e61a9eSTheodore Ts'o le64_to_cpu(es->s_first_error_block)); 274266e61a9eSTheodore Ts'o printk("\n"); 274366e61a9eSTheodore Ts'o } 274466e61a9eSTheodore Ts'o if (es->s_last_error_time) { 274566e61a9eSTheodore Ts'o printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d", 274666e61a9eSTheodore Ts'o sb->s_id, le32_to_cpu(es->s_last_error_time), 274766e61a9eSTheodore Ts'o (int) sizeof(es->s_last_error_func), 274866e61a9eSTheodore Ts'o es->s_last_error_func, 274966e61a9eSTheodore Ts'o le32_to_cpu(es->s_last_error_line)); 275066e61a9eSTheodore Ts'o if (es->s_last_error_ino) 275166e61a9eSTheodore Ts'o printk(": inode %u", 275266e61a9eSTheodore Ts'o le32_to_cpu(es->s_last_error_ino)); 275366e61a9eSTheodore Ts'o if (es->s_last_error_block) 275466e61a9eSTheodore Ts'o printk(": block %llu", (unsigned long long) 275566e61a9eSTheodore Ts'o le64_to_cpu(es->s_last_error_block)); 275666e61a9eSTheodore Ts'o printk("\n"); 275766e61a9eSTheodore Ts'o } 275866e61a9eSTheodore Ts'o mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 275966e61a9eSTheodore Ts'o } 276066e61a9eSTheodore Ts'o 2761bfff6873SLukas Czerner /* Find next suitable group and run ext4_init_inode_table */ 2762bfff6873SLukas Czerner static int ext4_run_li_request(struct ext4_li_request *elr) 2763bfff6873SLukas Czerner { 2764bfff6873SLukas Czerner struct ext4_group_desc *gdp = NULL; 2765bfff6873SLukas Czerner ext4_group_t group, ngroups; 2766bfff6873SLukas Czerner struct super_block *sb; 2767bfff6873SLukas Czerner unsigned long timeout = 0; 2768bfff6873SLukas Czerner int ret = 0; 2769bfff6873SLukas Czerner 2770bfff6873SLukas Czerner sb = elr->lr_super; 2771bfff6873SLukas Czerner ngroups = EXT4_SB(sb)->s_groups_count; 2772bfff6873SLukas Czerner 27738e8ad8a5SJan Kara sb_start_write(sb); 2774bfff6873SLukas Czerner for (group = elr->lr_next_group; group < ngroups; group++) { 2775bfff6873SLukas Czerner gdp = ext4_get_group_desc(sb, group, NULL); 2776bfff6873SLukas Czerner if (!gdp) { 2777bfff6873SLukas Czerner ret = 1; 2778bfff6873SLukas Czerner break; 2779bfff6873SLukas Czerner } 2780bfff6873SLukas Czerner 2781bfff6873SLukas Czerner if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2782bfff6873SLukas Czerner break; 2783bfff6873SLukas Czerner } 2784bfff6873SLukas Czerner 27857f511862STheodore Ts'o if (group >= ngroups) 2786bfff6873SLukas Czerner ret = 1; 2787bfff6873SLukas Czerner 2788bfff6873SLukas Czerner if (!ret) { 2789bfff6873SLukas Czerner timeout = jiffies; 2790bfff6873SLukas Czerner ret = ext4_init_inode_table(sb, group, 2791bfff6873SLukas Czerner elr->lr_timeout ? 0 : 1); 2792bfff6873SLukas Czerner if (elr->lr_timeout == 0) { 279351ce6511SLukas Czerner timeout = (jiffies - timeout) * 279451ce6511SLukas Czerner elr->lr_sbi->s_li_wait_mult; 2795bfff6873SLukas Czerner elr->lr_timeout = timeout; 2796bfff6873SLukas Czerner } 2797bfff6873SLukas Czerner elr->lr_next_sched = jiffies + elr->lr_timeout; 2798bfff6873SLukas Czerner elr->lr_next_group = group + 1; 2799bfff6873SLukas Czerner } 28008e8ad8a5SJan Kara sb_end_write(sb); 2801bfff6873SLukas Czerner 2802bfff6873SLukas Czerner return ret; 2803bfff6873SLukas Czerner } 2804bfff6873SLukas Czerner 2805bfff6873SLukas Czerner /* 2806bfff6873SLukas Czerner * Remove lr_request from the list_request and free the 28074ed5c033SLukas Czerner * request structure. Should be called with li_list_mtx held 2808bfff6873SLukas Czerner */ 2809bfff6873SLukas Czerner static void ext4_remove_li_request(struct ext4_li_request *elr) 2810bfff6873SLukas Czerner { 2811bfff6873SLukas Czerner struct ext4_sb_info *sbi; 2812bfff6873SLukas Czerner 2813bfff6873SLukas Czerner if (!elr) 2814bfff6873SLukas Czerner return; 2815bfff6873SLukas Czerner 2816bfff6873SLukas Czerner sbi = elr->lr_sbi; 2817bfff6873SLukas Czerner 2818bfff6873SLukas Czerner list_del(&elr->lr_request); 2819bfff6873SLukas Czerner sbi->s_li_request = NULL; 2820bfff6873SLukas Czerner kfree(elr); 2821bfff6873SLukas Czerner } 2822bfff6873SLukas Czerner 2823bfff6873SLukas Czerner static void ext4_unregister_li_request(struct super_block *sb) 2824bfff6873SLukas Czerner { 28251bb933fbSLukas Czerner mutex_lock(&ext4_li_mtx); 28261bb933fbSLukas Czerner if (!ext4_li_info) { 28271bb933fbSLukas Czerner mutex_unlock(&ext4_li_mtx); 2828bfff6873SLukas Czerner return; 28291bb933fbSLukas Czerner } 2830bfff6873SLukas Czerner 2831bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 28321bb933fbSLukas Czerner ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 2833bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 28341bb933fbSLukas Czerner mutex_unlock(&ext4_li_mtx); 2835bfff6873SLukas Czerner } 2836bfff6873SLukas Czerner 28378f1f7453SEric Sandeen static struct task_struct *ext4_lazyinit_task; 28388f1f7453SEric Sandeen 2839bfff6873SLukas Czerner /* 2840bfff6873SLukas Czerner * This is the function where ext4lazyinit thread lives. It walks 2841bfff6873SLukas Czerner * through the request list searching for next scheduled filesystem. 2842bfff6873SLukas Czerner * When such a fs is found, run the lazy initialization request 2843bfff6873SLukas Czerner * (ext4_rn_li_request) and keep track of the time spend in this 2844bfff6873SLukas Czerner * function. Based on that time we compute next schedule time of 2845bfff6873SLukas Czerner * the request. When walking through the list is complete, compute 2846bfff6873SLukas Czerner * next waking time and put itself into sleep. 2847bfff6873SLukas Czerner */ 2848bfff6873SLukas Czerner static int ext4_lazyinit_thread(void *arg) 2849bfff6873SLukas Czerner { 2850bfff6873SLukas Czerner struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 2851bfff6873SLukas Czerner struct list_head *pos, *n; 2852bfff6873SLukas Czerner struct ext4_li_request *elr; 28534ed5c033SLukas Czerner unsigned long next_wakeup, cur; 2854bfff6873SLukas Czerner 2855bfff6873SLukas Czerner BUG_ON(NULL == eli); 2856bfff6873SLukas Czerner 2857bfff6873SLukas Czerner cont_thread: 2858bfff6873SLukas Czerner while (true) { 2859bfff6873SLukas Czerner next_wakeup = MAX_JIFFY_OFFSET; 2860bfff6873SLukas Czerner 2861bfff6873SLukas Czerner mutex_lock(&eli->li_list_mtx); 2862bfff6873SLukas Czerner if (list_empty(&eli->li_request_list)) { 2863bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 2864bfff6873SLukas Czerner goto exit_thread; 2865bfff6873SLukas Czerner } 2866bfff6873SLukas Czerner 2867bfff6873SLukas Czerner list_for_each_safe(pos, n, &eli->li_request_list) { 2868bfff6873SLukas Czerner elr = list_entry(pos, struct ext4_li_request, 2869bfff6873SLukas Czerner lr_request); 2870bfff6873SLukas Czerner 2871b2c78cd0STheodore Ts'o if (time_after_eq(jiffies, elr->lr_next_sched)) { 2872b2c78cd0STheodore Ts'o if (ext4_run_li_request(elr) != 0) { 2873b2c78cd0STheodore Ts'o /* error, remove the lazy_init job */ 2874bfff6873SLukas Czerner ext4_remove_li_request(elr); 2875bfff6873SLukas Czerner continue; 2876bfff6873SLukas Czerner } 2877b2c78cd0STheodore Ts'o } 2878bfff6873SLukas Czerner 2879bfff6873SLukas Czerner if (time_before(elr->lr_next_sched, next_wakeup)) 2880bfff6873SLukas Czerner next_wakeup = elr->lr_next_sched; 2881bfff6873SLukas Czerner } 2882bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 2883bfff6873SLukas Czerner 2884a0acae0eSTejun Heo try_to_freeze(); 2885bfff6873SLukas Czerner 28864ed5c033SLukas Czerner cur = jiffies; 28874ed5c033SLukas Czerner if ((time_after_eq(cur, next_wakeup)) || 2888f4245bd4SLukas Czerner (MAX_JIFFY_OFFSET == next_wakeup)) { 2889bfff6873SLukas Czerner cond_resched(); 2890bfff6873SLukas Czerner continue; 2891bfff6873SLukas Czerner } 2892bfff6873SLukas Czerner 28934ed5c033SLukas Czerner schedule_timeout_interruptible(next_wakeup - cur); 28944ed5c033SLukas Czerner 28958f1f7453SEric Sandeen if (kthread_should_stop()) { 28968f1f7453SEric Sandeen ext4_clear_request_list(); 28978f1f7453SEric Sandeen goto exit_thread; 28988f1f7453SEric Sandeen } 2899bfff6873SLukas Czerner } 2900bfff6873SLukas Czerner 2901bfff6873SLukas Czerner exit_thread: 2902bfff6873SLukas Czerner /* 2903bfff6873SLukas Czerner * It looks like the request list is empty, but we need 2904bfff6873SLukas Czerner * to check it under the li_list_mtx lock, to prevent any 2905bfff6873SLukas Czerner * additions into it, and of course we should lock ext4_li_mtx 2906bfff6873SLukas Czerner * to atomically free the list and ext4_li_info, because at 2907bfff6873SLukas Czerner * this point another ext4 filesystem could be registering 2908bfff6873SLukas Czerner * new one. 2909bfff6873SLukas Czerner */ 2910bfff6873SLukas Czerner mutex_lock(&ext4_li_mtx); 2911bfff6873SLukas Czerner mutex_lock(&eli->li_list_mtx); 2912bfff6873SLukas Czerner if (!list_empty(&eli->li_request_list)) { 2913bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 2914bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 2915bfff6873SLukas Czerner goto cont_thread; 2916bfff6873SLukas Czerner } 2917bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 2918bfff6873SLukas Czerner kfree(ext4_li_info); 2919bfff6873SLukas Czerner ext4_li_info = NULL; 2920bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 2921bfff6873SLukas Czerner 2922bfff6873SLukas Czerner return 0; 2923bfff6873SLukas Czerner } 2924bfff6873SLukas Czerner 2925bfff6873SLukas Czerner static void ext4_clear_request_list(void) 2926bfff6873SLukas Czerner { 2927bfff6873SLukas Czerner struct list_head *pos, *n; 2928bfff6873SLukas Czerner struct ext4_li_request *elr; 2929bfff6873SLukas Czerner 2930bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 2931bfff6873SLukas Czerner list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 2932bfff6873SLukas Czerner elr = list_entry(pos, struct ext4_li_request, 2933bfff6873SLukas Czerner lr_request); 2934bfff6873SLukas Czerner ext4_remove_li_request(elr); 2935bfff6873SLukas Czerner } 2936bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 2937bfff6873SLukas Czerner } 2938bfff6873SLukas Czerner 2939bfff6873SLukas Czerner static int ext4_run_lazyinit_thread(void) 2940bfff6873SLukas Czerner { 29418f1f7453SEric Sandeen ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 29428f1f7453SEric Sandeen ext4_li_info, "ext4lazyinit"); 29438f1f7453SEric Sandeen if (IS_ERR(ext4_lazyinit_task)) { 29448f1f7453SEric Sandeen int err = PTR_ERR(ext4_lazyinit_task); 2945bfff6873SLukas Czerner ext4_clear_request_list(); 2946bfff6873SLukas Czerner kfree(ext4_li_info); 2947bfff6873SLukas Czerner ext4_li_info = NULL; 294892b97816STheodore Ts'o printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 2949bfff6873SLukas Czerner "initialization thread\n", 2950bfff6873SLukas Czerner err); 2951bfff6873SLukas Czerner return err; 2952bfff6873SLukas Czerner } 2953bfff6873SLukas Czerner ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 2954bfff6873SLukas Czerner return 0; 2955bfff6873SLukas Czerner } 2956bfff6873SLukas Czerner 2957bfff6873SLukas Czerner /* 2958bfff6873SLukas Czerner * Check whether it make sense to run itable init. thread or not. 2959bfff6873SLukas Czerner * If there is at least one uninitialized inode table, return 2960bfff6873SLukas Czerner * corresponding group number, else the loop goes through all 2961bfff6873SLukas Czerner * groups and return total number of groups. 2962bfff6873SLukas Czerner */ 2963bfff6873SLukas Czerner static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 2964bfff6873SLukas Czerner { 2965bfff6873SLukas Czerner ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 2966bfff6873SLukas Czerner struct ext4_group_desc *gdp = NULL; 2967bfff6873SLukas Czerner 2968bfff6873SLukas Czerner for (group = 0; group < ngroups; group++) { 2969bfff6873SLukas Czerner gdp = ext4_get_group_desc(sb, group, NULL); 2970bfff6873SLukas Czerner if (!gdp) 2971bfff6873SLukas Czerner continue; 2972bfff6873SLukas Czerner 2973bfff6873SLukas Czerner if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2974bfff6873SLukas Czerner break; 2975bfff6873SLukas Czerner } 2976bfff6873SLukas Czerner 2977bfff6873SLukas Czerner return group; 2978bfff6873SLukas Czerner } 2979bfff6873SLukas Czerner 2980bfff6873SLukas Czerner static int ext4_li_info_new(void) 2981bfff6873SLukas Czerner { 2982bfff6873SLukas Czerner struct ext4_lazy_init *eli = NULL; 2983bfff6873SLukas Czerner 2984bfff6873SLukas Czerner eli = kzalloc(sizeof(*eli), GFP_KERNEL); 2985bfff6873SLukas Czerner if (!eli) 2986bfff6873SLukas Czerner return -ENOMEM; 2987bfff6873SLukas Czerner 2988bfff6873SLukas Czerner INIT_LIST_HEAD(&eli->li_request_list); 2989bfff6873SLukas Czerner mutex_init(&eli->li_list_mtx); 2990bfff6873SLukas Czerner 2991bfff6873SLukas Czerner eli->li_state |= EXT4_LAZYINIT_QUIT; 2992bfff6873SLukas Czerner 2993bfff6873SLukas Czerner ext4_li_info = eli; 2994bfff6873SLukas Czerner 2995bfff6873SLukas Czerner return 0; 2996bfff6873SLukas Czerner } 2997bfff6873SLukas Czerner 2998bfff6873SLukas Czerner static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 2999bfff6873SLukas Czerner ext4_group_t start) 3000bfff6873SLukas Czerner { 3001bfff6873SLukas Czerner struct ext4_sb_info *sbi = EXT4_SB(sb); 3002bfff6873SLukas Czerner struct ext4_li_request *elr; 3003bfff6873SLukas Czerner unsigned long rnd; 3004bfff6873SLukas Czerner 3005bfff6873SLukas Czerner elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3006bfff6873SLukas Czerner if (!elr) 3007bfff6873SLukas Czerner return NULL; 3008bfff6873SLukas Czerner 3009bfff6873SLukas Czerner elr->lr_super = sb; 3010bfff6873SLukas Czerner elr->lr_sbi = sbi; 3011bfff6873SLukas Czerner elr->lr_next_group = start; 3012bfff6873SLukas Czerner 3013bfff6873SLukas Czerner /* 3014bfff6873SLukas Czerner * Randomize first schedule time of the request to 3015bfff6873SLukas Czerner * spread the inode table initialization requests 3016bfff6873SLukas Czerner * better. 3017bfff6873SLukas Czerner */ 3018bfff6873SLukas Czerner get_random_bytes(&rnd, sizeof(rnd)); 3019bfff6873SLukas Czerner elr->lr_next_sched = jiffies + (unsigned long)rnd % 3020bfff6873SLukas Czerner (EXT4_DEF_LI_MAX_START_DELAY * HZ); 3021bfff6873SLukas Czerner 3022bfff6873SLukas Czerner return elr; 3023bfff6873SLukas Czerner } 3024bfff6873SLukas Czerner 30257f511862STheodore Ts'o int ext4_register_li_request(struct super_block *sb, 3026bfff6873SLukas Czerner ext4_group_t first_not_zeroed) 3027bfff6873SLukas Czerner { 3028bfff6873SLukas Czerner struct ext4_sb_info *sbi = EXT4_SB(sb); 30297f511862STheodore Ts'o struct ext4_li_request *elr = NULL; 3030bfff6873SLukas Czerner ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 30316c5a6cb9SAndrew Morton int ret = 0; 3032bfff6873SLukas Czerner 30337f511862STheodore Ts'o mutex_lock(&ext4_li_mtx); 303451ce6511SLukas Czerner if (sbi->s_li_request != NULL) { 303551ce6511SLukas Czerner /* 303651ce6511SLukas Czerner * Reset timeout so it can be computed again, because 303751ce6511SLukas Czerner * s_li_wait_mult might have changed. 303851ce6511SLukas Czerner */ 303951ce6511SLukas Czerner sbi->s_li_request->lr_timeout = 0; 30407f511862STheodore Ts'o goto out; 304151ce6511SLukas Czerner } 3042bfff6873SLukas Czerner 3043bfff6873SLukas Czerner if (first_not_zeroed == ngroups || 3044bfff6873SLukas Czerner (sb->s_flags & MS_RDONLY) || 304555ff3840STao Ma !test_opt(sb, INIT_INODE_TABLE)) 30467f511862STheodore Ts'o goto out; 3047bfff6873SLukas Czerner 3048bfff6873SLukas Czerner elr = ext4_li_request_new(sb, first_not_zeroed); 30497f511862STheodore Ts'o if (!elr) { 30507f511862STheodore Ts'o ret = -ENOMEM; 30517f511862STheodore Ts'o goto out; 30527f511862STheodore Ts'o } 3053bfff6873SLukas Czerner 3054bfff6873SLukas Czerner if (NULL == ext4_li_info) { 3055bfff6873SLukas Czerner ret = ext4_li_info_new(); 3056bfff6873SLukas Czerner if (ret) 3057bfff6873SLukas Czerner goto out; 3058bfff6873SLukas Czerner } 3059bfff6873SLukas Czerner 3060bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 3061bfff6873SLukas Czerner list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3062bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 3063bfff6873SLukas Czerner 3064bfff6873SLukas Czerner sbi->s_li_request = elr; 306546e4690bSTao Ma /* 306646e4690bSTao Ma * set elr to NULL here since it has been inserted to 306746e4690bSTao Ma * the request_list and the removal and free of it is 306846e4690bSTao Ma * handled by ext4_clear_request_list from now on. 306946e4690bSTao Ma */ 307046e4690bSTao Ma elr = NULL; 3071bfff6873SLukas Czerner 3072bfff6873SLukas Czerner if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3073bfff6873SLukas Czerner ret = ext4_run_lazyinit_thread(); 3074bfff6873SLukas Czerner if (ret) 3075bfff6873SLukas Czerner goto out; 3076bfff6873SLukas Czerner } 3077bfff6873SLukas Czerner out: 3078bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 3079beed5ecbSNicolas Kaiser if (ret) 3080bfff6873SLukas Czerner kfree(elr); 3081bfff6873SLukas Czerner return ret; 3082bfff6873SLukas Czerner } 3083bfff6873SLukas Czerner 3084bfff6873SLukas Czerner /* 3085bfff6873SLukas Czerner * We do not need to lock anything since this is called on 3086bfff6873SLukas Czerner * module unload. 3087bfff6873SLukas Czerner */ 3088bfff6873SLukas Czerner static void ext4_destroy_lazyinit_thread(void) 3089bfff6873SLukas Czerner { 3090bfff6873SLukas Czerner /* 3091bfff6873SLukas Czerner * If thread exited earlier 3092bfff6873SLukas Czerner * there's nothing to be done. 3093bfff6873SLukas Czerner */ 30948f1f7453SEric Sandeen if (!ext4_li_info || !ext4_lazyinit_task) 3095bfff6873SLukas Czerner return; 3096bfff6873SLukas Czerner 30978f1f7453SEric Sandeen kthread_stop(ext4_lazyinit_task); 3098bfff6873SLukas Czerner } 3099bfff6873SLukas Czerner 310025ed6e8aSDarrick J. Wong static int set_journal_csum_feature_set(struct super_block *sb) 310125ed6e8aSDarrick J. Wong { 310225ed6e8aSDarrick J. Wong int ret = 1; 310325ed6e8aSDarrick J. Wong int compat, incompat; 310425ed6e8aSDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 310525ed6e8aSDarrick J. Wong 310625ed6e8aSDarrick J. Wong if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 310725ed6e8aSDarrick J. Wong EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) { 310825ed6e8aSDarrick J. Wong /* journal checksum v2 */ 310925ed6e8aSDarrick J. Wong compat = 0; 311025ed6e8aSDarrick J. Wong incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2; 311125ed6e8aSDarrick J. Wong } else { 311225ed6e8aSDarrick J. Wong /* journal checksum v1 */ 311325ed6e8aSDarrick J. Wong compat = JBD2_FEATURE_COMPAT_CHECKSUM; 311425ed6e8aSDarrick J. Wong incompat = 0; 311525ed6e8aSDarrick J. Wong } 311625ed6e8aSDarrick J. Wong 311725ed6e8aSDarrick J. Wong if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 311825ed6e8aSDarrick J. Wong ret = jbd2_journal_set_features(sbi->s_journal, 311925ed6e8aSDarrick J. Wong compat, 0, 312025ed6e8aSDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 312125ed6e8aSDarrick J. Wong incompat); 312225ed6e8aSDarrick J. Wong } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 312325ed6e8aSDarrick J. Wong ret = jbd2_journal_set_features(sbi->s_journal, 312425ed6e8aSDarrick J. Wong compat, 0, 312525ed6e8aSDarrick J. Wong incompat); 312625ed6e8aSDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 0, 0, 312725ed6e8aSDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 312825ed6e8aSDarrick J. Wong } else { 312925ed6e8aSDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 313025ed6e8aSDarrick J. Wong JBD2_FEATURE_COMPAT_CHECKSUM, 0, 313125ed6e8aSDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 313225ed6e8aSDarrick J. Wong JBD2_FEATURE_INCOMPAT_CSUM_V2); 313325ed6e8aSDarrick J. Wong } 313425ed6e8aSDarrick J. Wong 313525ed6e8aSDarrick J. Wong return ret; 313625ed6e8aSDarrick J. Wong } 313725ed6e8aSDarrick J. Wong 3138952fc18eSTheodore Ts'o /* 3139952fc18eSTheodore Ts'o * Note: calculating the overhead so we can be compatible with 3140952fc18eSTheodore Ts'o * historical BSD practice is quite difficult in the face of 3141952fc18eSTheodore Ts'o * clusters/bigalloc. This is because multiple metadata blocks from 3142952fc18eSTheodore Ts'o * different block group can end up in the same allocation cluster. 3143952fc18eSTheodore Ts'o * Calculating the exact overhead in the face of clustered allocation 3144952fc18eSTheodore Ts'o * requires either O(all block bitmaps) in memory or O(number of block 3145952fc18eSTheodore Ts'o * groups**2) in time. We will still calculate the superblock for 3146952fc18eSTheodore Ts'o * older file systems --- and if we come across with a bigalloc file 3147952fc18eSTheodore Ts'o * system with zero in s_overhead_clusters the estimate will be close to 3148952fc18eSTheodore Ts'o * correct especially for very large cluster sizes --- but for newer 3149952fc18eSTheodore Ts'o * file systems, it's better to calculate this figure once at mkfs 3150952fc18eSTheodore Ts'o * time, and store it in the superblock. If the superblock value is 3151952fc18eSTheodore Ts'o * present (even for non-bigalloc file systems), we will use it. 3152952fc18eSTheodore Ts'o */ 3153952fc18eSTheodore Ts'o static int count_overhead(struct super_block *sb, ext4_group_t grp, 3154952fc18eSTheodore Ts'o char *buf) 3155952fc18eSTheodore Ts'o { 3156952fc18eSTheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 3157952fc18eSTheodore Ts'o struct ext4_group_desc *gdp; 3158952fc18eSTheodore Ts'o ext4_fsblk_t first_block, last_block, b; 3159952fc18eSTheodore Ts'o ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3160952fc18eSTheodore Ts'o int s, j, count = 0; 3161952fc18eSTheodore Ts'o 31620548bbb8STheodore Ts'o if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC)) 31630548bbb8STheodore Ts'o return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 31640548bbb8STheodore Ts'o sbi->s_itb_per_group + 2); 31650548bbb8STheodore Ts'o 3166952fc18eSTheodore Ts'o first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3167952fc18eSTheodore Ts'o (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3168952fc18eSTheodore Ts'o last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3169952fc18eSTheodore Ts'o for (i = 0; i < ngroups; i++) { 3170952fc18eSTheodore Ts'o gdp = ext4_get_group_desc(sb, i, NULL); 3171952fc18eSTheodore Ts'o b = ext4_block_bitmap(sb, gdp); 3172952fc18eSTheodore Ts'o if (b >= first_block && b <= last_block) { 3173952fc18eSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3174952fc18eSTheodore Ts'o count++; 3175952fc18eSTheodore Ts'o } 3176952fc18eSTheodore Ts'o b = ext4_inode_bitmap(sb, gdp); 3177952fc18eSTheodore Ts'o if (b >= first_block && b <= last_block) { 3178952fc18eSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3179952fc18eSTheodore Ts'o count++; 3180952fc18eSTheodore Ts'o } 3181952fc18eSTheodore Ts'o b = ext4_inode_table(sb, gdp); 3182952fc18eSTheodore Ts'o if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3183952fc18eSTheodore Ts'o for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3184952fc18eSTheodore Ts'o int c = EXT4_B2C(sbi, b - first_block); 3185952fc18eSTheodore Ts'o ext4_set_bit(c, buf); 3186952fc18eSTheodore Ts'o count++; 3187952fc18eSTheodore Ts'o } 3188952fc18eSTheodore Ts'o if (i != grp) 3189952fc18eSTheodore Ts'o continue; 3190952fc18eSTheodore Ts'o s = 0; 3191952fc18eSTheodore Ts'o if (ext4_bg_has_super(sb, grp)) { 3192952fc18eSTheodore Ts'o ext4_set_bit(s++, buf); 3193952fc18eSTheodore Ts'o count++; 3194952fc18eSTheodore Ts'o } 3195952fc18eSTheodore Ts'o for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) { 3196952fc18eSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3197952fc18eSTheodore Ts'o count++; 3198952fc18eSTheodore Ts'o } 3199952fc18eSTheodore Ts'o } 3200952fc18eSTheodore Ts'o if (!count) 3201952fc18eSTheodore Ts'o return 0; 3202952fc18eSTheodore Ts'o return EXT4_CLUSTERS_PER_GROUP(sb) - 3203952fc18eSTheodore Ts'o ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3204952fc18eSTheodore Ts'o } 3205952fc18eSTheodore Ts'o 3206952fc18eSTheodore Ts'o /* 3207952fc18eSTheodore Ts'o * Compute the overhead and stash it in sbi->s_overhead 3208952fc18eSTheodore Ts'o */ 3209952fc18eSTheodore Ts'o int ext4_calculate_overhead(struct super_block *sb) 3210952fc18eSTheodore Ts'o { 3211952fc18eSTheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 3212952fc18eSTheodore Ts'o struct ext4_super_block *es = sbi->s_es; 3213952fc18eSTheodore Ts'o ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3214952fc18eSTheodore Ts'o ext4_fsblk_t overhead = 0; 3215952fc18eSTheodore Ts'o char *buf = (char *) get_zeroed_page(GFP_KERNEL); 3216952fc18eSTheodore Ts'o 3217952fc18eSTheodore Ts'o if (!buf) 3218952fc18eSTheodore Ts'o return -ENOMEM; 3219952fc18eSTheodore Ts'o 3220952fc18eSTheodore Ts'o /* 3221952fc18eSTheodore Ts'o * Compute the overhead (FS structures). This is constant 3222952fc18eSTheodore Ts'o * for a given filesystem unless the number of block groups 3223952fc18eSTheodore Ts'o * changes so we cache the previous value until it does. 3224952fc18eSTheodore Ts'o */ 3225952fc18eSTheodore Ts'o 3226952fc18eSTheodore Ts'o /* 3227952fc18eSTheodore Ts'o * All of the blocks before first_data_block are overhead 3228952fc18eSTheodore Ts'o */ 3229952fc18eSTheodore Ts'o overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3230952fc18eSTheodore Ts'o 3231952fc18eSTheodore Ts'o /* 3232952fc18eSTheodore Ts'o * Add the overhead found in each block group 3233952fc18eSTheodore Ts'o */ 3234952fc18eSTheodore Ts'o for (i = 0; i < ngroups; i++) { 3235952fc18eSTheodore Ts'o int blks; 3236952fc18eSTheodore Ts'o 3237952fc18eSTheodore Ts'o blks = count_overhead(sb, i, buf); 3238952fc18eSTheodore Ts'o overhead += blks; 3239952fc18eSTheodore Ts'o if (blks) 3240952fc18eSTheodore Ts'o memset(buf, 0, PAGE_SIZE); 3241952fc18eSTheodore Ts'o cond_resched(); 3242952fc18eSTheodore Ts'o } 32430875a2b4SEric Sandeen /* Add the journal blocks as well */ 32440875a2b4SEric Sandeen if (sbi->s_journal) 3245810da240SLukas Czerner overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen); 32460875a2b4SEric Sandeen 3247952fc18eSTheodore Ts'o sbi->s_overhead = overhead; 3248952fc18eSTheodore Ts'o smp_wmb(); 3249952fc18eSTheodore Ts'o free_page((unsigned long) buf); 3250952fc18eSTheodore Ts'o return 0; 3251952fc18eSTheodore Ts'o } 3252952fc18eSTheodore Ts'o 325327dd4385SLukas Czerner 325427dd4385SLukas Czerner static ext4_fsblk_t ext4_calculate_resv_clusters(struct ext4_sb_info *sbi) 325527dd4385SLukas Czerner { 325627dd4385SLukas Czerner ext4_fsblk_t resv_clusters; 325727dd4385SLukas Czerner 325827dd4385SLukas Czerner /* 325927dd4385SLukas Czerner * By default we reserve 2% or 4096 clusters, whichever is smaller. 326027dd4385SLukas Czerner * This should cover the situations where we can not afford to run 326127dd4385SLukas Czerner * out of space like for example punch hole, or converting 326227dd4385SLukas Czerner * uninitialized extents in delalloc path. In most cases such 326327dd4385SLukas Czerner * allocation would require 1, or 2 blocks, higher numbers are 326427dd4385SLukas Czerner * very rare. 326527dd4385SLukas Czerner */ 326627dd4385SLukas Czerner resv_clusters = ext4_blocks_count(sbi->s_es) >> sbi->s_cluster_bits; 326727dd4385SLukas Czerner 326827dd4385SLukas Czerner do_div(resv_clusters, 50); 326927dd4385SLukas Czerner resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 327027dd4385SLukas Czerner 327127dd4385SLukas Czerner return resv_clusters; 327227dd4385SLukas Czerner } 327327dd4385SLukas Czerner 327427dd4385SLukas Czerner 327527dd4385SLukas Czerner static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count) 327627dd4385SLukas Czerner { 327727dd4385SLukas Czerner ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >> 327827dd4385SLukas Czerner sbi->s_cluster_bits; 327927dd4385SLukas Czerner 328027dd4385SLukas Czerner if (count >= clusters) 328127dd4385SLukas Czerner return -EINVAL; 328227dd4385SLukas Czerner 328327dd4385SLukas Czerner atomic64_set(&sbi->s_resv_clusters, count); 328427dd4385SLukas Czerner return 0; 328527dd4385SLukas Czerner } 328627dd4385SLukas Czerner 3287617ba13bSMingming Cao static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3288ac27a0ecSDave Kleikamp { 3289d4c402d9SCurt Wohlgemuth char *orig_data = kstrdup(data, GFP_KERNEL); 3290ac27a0ecSDave Kleikamp struct buffer_head *bh; 3291617ba13bSMingming Cao struct ext4_super_block *es = NULL; 3292617ba13bSMingming Cao struct ext4_sb_info *sbi; 3293617ba13bSMingming Cao ext4_fsblk_t block; 3294617ba13bSMingming Cao ext4_fsblk_t sb_block = get_sb_block(&data); 329570bbb3e0SAndrew Morton ext4_fsblk_t logical_sb_block; 3296ac27a0ecSDave Kleikamp unsigned long offset = 0; 3297ac27a0ecSDave Kleikamp unsigned long journal_devnum = 0; 3298ac27a0ecSDave Kleikamp unsigned long def_mount_opts; 3299ac27a0ecSDave Kleikamp struct inode *root; 33009f6200bbSTheodore Ts'o char *cp; 33010390131bSFrank Mayhar const char *descr; 3302dcc7dae3SCyrill Gorcunov int ret = -ENOMEM; 3303281b5995STheodore Ts'o int blocksize, clustersize; 33044ec11028STheodore Ts'o unsigned int db_count; 33054ec11028STheodore Ts'o unsigned int i; 3306281b5995STheodore Ts'o int needs_recovery, has_huge_files, has_bigalloc; 3307bd81d8eeSLaurent Vivier __u64 blocks_count; 330807aa2ea1SLukas Czerner int err = 0; 3309b3881f74STheodore Ts'o unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3310bfff6873SLukas Czerner ext4_group_t first_not_zeroed; 3311ac27a0ecSDave Kleikamp 3312ac27a0ecSDave Kleikamp sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 3313ac27a0ecSDave Kleikamp if (!sbi) 3314dcc7dae3SCyrill Gorcunov goto out_free_orig; 3315705895b6SPekka Enberg 3316705895b6SPekka Enberg sbi->s_blockgroup_lock = 3317705895b6SPekka Enberg kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 3318705895b6SPekka Enberg if (!sbi->s_blockgroup_lock) { 3319705895b6SPekka Enberg kfree(sbi); 3320dcc7dae3SCyrill Gorcunov goto out_free_orig; 3321705895b6SPekka Enberg } 3322ac27a0ecSDave Kleikamp sb->s_fs_info = sbi; 33232c0544b2STheodore Ts'o sbi->s_sb = sb; 3324240799cdSTheodore Ts'o sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3325d9c9bef1SMiklos Szeredi sbi->s_sb_block = sb_block; 3326f613dfcbSTheodore Ts'o if (sb->s_bdev->bd_part) 3327f613dfcbSTheodore Ts'o sbi->s_sectors_written_start = 3328f613dfcbSTheodore Ts'o part_stat_read(sb->s_bdev->bd_part, sectors[1]); 3329ac27a0ecSDave Kleikamp 33309f6200bbSTheodore Ts'o /* Cleanup superblock name */ 33319f6200bbSTheodore Ts'o for (cp = sb->s_id; (cp = strchr(cp, '/'));) 33329f6200bbSTheodore Ts'o *cp = '!'; 33339f6200bbSTheodore Ts'o 333407aa2ea1SLukas Czerner /* -EINVAL is default */ 3335dcc7dae3SCyrill Gorcunov ret = -EINVAL; 3336617ba13bSMingming Cao blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3337ac27a0ecSDave Kleikamp if (!blocksize) { 3338b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3339ac27a0ecSDave Kleikamp goto out_fail; 3340ac27a0ecSDave Kleikamp } 3341ac27a0ecSDave Kleikamp 3342ac27a0ecSDave Kleikamp /* 3343617ba13bSMingming Cao * The ext4 superblock will not be buffer aligned for other than 1kB 3344ac27a0ecSDave Kleikamp * block sizes. We need to calculate the offset from buffer start. 3345ac27a0ecSDave Kleikamp */ 3346617ba13bSMingming Cao if (blocksize != EXT4_MIN_BLOCK_SIZE) { 334770bbb3e0SAndrew Morton logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 334870bbb3e0SAndrew Morton offset = do_div(logical_sb_block, blocksize); 3349ac27a0ecSDave Kleikamp } else { 335070bbb3e0SAndrew Morton logical_sb_block = sb_block; 3351ac27a0ecSDave Kleikamp } 3352ac27a0ecSDave Kleikamp 335370bbb3e0SAndrew Morton if (!(bh = sb_bread(sb, logical_sb_block))) { 3354b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3355ac27a0ecSDave Kleikamp goto out_fail; 3356ac27a0ecSDave Kleikamp } 3357ac27a0ecSDave Kleikamp /* 3358ac27a0ecSDave Kleikamp * Note: s_es must be initialized as soon as possible because 3359617ba13bSMingming Cao * some ext4 macro-instructions depend on its value 3360ac27a0ecSDave Kleikamp */ 33612716b802STheodore Ts'o es = (struct ext4_super_block *) (bh->b_data + offset); 3362ac27a0ecSDave Kleikamp sbi->s_es = es; 3363ac27a0ecSDave Kleikamp sb->s_magic = le16_to_cpu(es->s_magic); 3364617ba13bSMingming Cao if (sb->s_magic != EXT4_SUPER_MAGIC) 3365617ba13bSMingming Cao goto cantfind_ext4; 3366afc32f7eSTheodore Ts'o sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3367ac27a0ecSDave Kleikamp 3368feb0ab32SDarrick J. Wong /* Warn if metadata_csum and gdt_csum are both set. */ 3369feb0ab32SDarrick J. Wong if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3370feb0ab32SDarrick J. Wong EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 3371feb0ab32SDarrick J. Wong EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) 3372feb0ab32SDarrick J. Wong ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are " 3373feb0ab32SDarrick J. Wong "redundant flags; please run fsck."); 3374feb0ab32SDarrick J. Wong 3375d25425f8SDarrick J. Wong /* Check for a known checksum algorithm */ 3376d25425f8SDarrick J. Wong if (!ext4_verify_csum_type(sb, es)) { 3377d25425f8SDarrick J. Wong ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3378d25425f8SDarrick J. Wong "unknown checksum algorithm."); 3379d25425f8SDarrick J. Wong silent = 1; 3380d25425f8SDarrick J. Wong goto cantfind_ext4; 3381d25425f8SDarrick J. Wong } 3382d25425f8SDarrick J. Wong 33830441984aSDarrick J. Wong /* Load the checksum driver */ 33840441984aSDarrick J. Wong if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 33850441984aSDarrick J. Wong EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) { 33860441984aSDarrick J. Wong sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 33870441984aSDarrick J. Wong if (IS_ERR(sbi->s_chksum_driver)) { 33880441984aSDarrick J. Wong ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 33890441984aSDarrick J. Wong ret = PTR_ERR(sbi->s_chksum_driver); 33900441984aSDarrick J. Wong sbi->s_chksum_driver = NULL; 33910441984aSDarrick J. Wong goto failed_mount; 33920441984aSDarrick J. Wong } 33930441984aSDarrick J. Wong } 33940441984aSDarrick J. Wong 3395a9c47317SDarrick J. Wong /* Check superblock checksum */ 3396a9c47317SDarrick J. Wong if (!ext4_superblock_csum_verify(sb, es)) { 3397a9c47317SDarrick J. Wong ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3398a9c47317SDarrick J. Wong "invalid superblock checksum. Run e2fsck?"); 3399a9c47317SDarrick J. Wong silent = 1; 3400a9c47317SDarrick J. Wong goto cantfind_ext4; 3401a9c47317SDarrick J. Wong } 3402a9c47317SDarrick J. Wong 3403a9c47317SDarrick J. Wong /* Precompute checksum seed for all metadata */ 3404a9c47317SDarrick J. Wong if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3405a9c47317SDarrick J. Wong EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) 3406a9c47317SDarrick J. Wong sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 3407a9c47317SDarrick J. Wong sizeof(es->s_uuid)); 3408a9c47317SDarrick J. Wong 3409ac27a0ecSDave Kleikamp /* Set defaults before we parse the mount options */ 3410ac27a0ecSDave Kleikamp def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3411fd8c37ecSTheodore Ts'o set_opt(sb, INIT_INODE_TABLE); 3412617ba13bSMingming Cao if (def_mount_opts & EXT4_DEFM_DEBUG) 3413fd8c37ecSTheodore Ts'o set_opt(sb, DEBUG); 341487f26807STheodore Ts'o if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 3415fd8c37ecSTheodore Ts'o set_opt(sb, GRPID); 3416617ba13bSMingming Cao if (def_mount_opts & EXT4_DEFM_UID16) 3417fd8c37ecSTheodore Ts'o set_opt(sb, NO_UID32); 3418ea663336SEric Sandeen /* xattr user namespace & acls are now defaulted on */ 3419fd8c37ecSTheodore Ts'o set_opt(sb, XATTR_USER); 342003010a33STheodore Ts'o #ifdef CONFIG_EXT4_FS_POSIX_ACL 3421fd8c37ecSTheodore Ts'o set_opt(sb, POSIX_ACL); 34222e7842b8SHugh Dickins #endif 3423617ba13bSMingming Cao if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 3424fd8c37ecSTheodore Ts'o set_opt(sb, JOURNAL_DATA); 3425617ba13bSMingming Cao else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 3426fd8c37ecSTheodore Ts'o set_opt(sb, ORDERED_DATA); 3427617ba13bSMingming Cao else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 3428fd8c37ecSTheodore Ts'o set_opt(sb, WRITEBACK_DATA); 3429ac27a0ecSDave Kleikamp 3430617ba13bSMingming Cao if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 3431fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_PANIC); 3432bb4f397aSAneesh Kumar K.V else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 3433fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_CONT); 3434bb4f397aSAneesh Kumar K.V else 3435fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_RO); 34368b67f04aSTheodore Ts'o if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY) 3437fd8c37ecSTheodore Ts'o set_opt(sb, BLOCK_VALIDITY); 34388b67f04aSTheodore Ts'o if (def_mount_opts & EXT4_DEFM_DISCARD) 3439fd8c37ecSTheodore Ts'o set_opt(sb, DISCARD); 3440ac27a0ecSDave Kleikamp 344108cefc7aSEric W. Biederman sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 344208cefc7aSEric W. Biederman sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 344330773840STheodore Ts'o sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 344430773840STheodore Ts'o sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 344530773840STheodore Ts'o sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3446ac27a0ecSDave Kleikamp 34478b67f04aSTheodore Ts'o if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 3448fd8c37ecSTheodore Ts'o set_opt(sb, BARRIER); 3449ac27a0ecSDave Kleikamp 34501e2462f9SMingming Cao /* 3451dd919b98SAneesh Kumar K.V * enable delayed allocation by default 3452dd919b98SAneesh Kumar K.V * Use -o nodelalloc to turn it off 3453dd919b98SAneesh Kumar K.V */ 3454bc0b75f7STheodore Ts'o if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 34558b67f04aSTheodore Ts'o ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 3456fd8c37ecSTheodore Ts'o set_opt(sb, DELALLOC); 3457dd919b98SAneesh Kumar K.V 345851ce6511SLukas Czerner /* 345951ce6511SLukas Czerner * set default s_li_wait_mult for lazyinit, for the case there is 346051ce6511SLukas Czerner * no mount option specified. 346151ce6511SLukas Czerner */ 346251ce6511SLukas Czerner sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 346351ce6511SLukas Czerner 34648b67f04aSTheodore Ts'o if (!parse_options((char *) sbi->s_es->s_mount_opts, sb, 3465661aa520SEric Sandeen &journal_devnum, &journal_ioprio, 0)) { 34668b67f04aSTheodore Ts'o ext4_msg(sb, KERN_WARNING, 34678b67f04aSTheodore Ts'o "failed to parse options in superblock: %s", 34688b67f04aSTheodore Ts'o sbi->s_es->s_mount_opts); 34698b67f04aSTheodore Ts'o } 34705a916be1STheodore Ts'o sbi->s_def_mount_opt = sbi->s_mount_opt; 3471b3881f74STheodore Ts'o if (!parse_options((char *) data, sb, &journal_devnum, 3472661aa520SEric Sandeen &journal_ioprio, 0)) 3473ac27a0ecSDave Kleikamp goto failed_mount; 3474ac27a0ecSDave Kleikamp 347556889787STheodore Ts'o if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 347656889787STheodore Ts'o printk_once(KERN_WARNING "EXT4-fs: Warning: mounting " 347756889787STheodore Ts'o "with data=journal disables delayed " 347856889787STheodore Ts'o "allocation and O_DIRECT support!\n"); 347956889787STheodore Ts'o if (test_opt2(sb, EXPLICIT_DELALLOC)) { 348056889787STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't mount with " 348156889787STheodore Ts'o "both data=journal and delalloc"); 348256889787STheodore Ts'o goto failed_mount; 348356889787STheodore Ts'o } 348456889787STheodore Ts'o if (test_opt(sb, DIOREAD_NOLOCK)) { 348556889787STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't mount with " 348656889787STheodore Ts'o "both data=journal and delalloc"); 348756889787STheodore Ts'o goto failed_mount; 348856889787STheodore Ts'o } 348956889787STheodore Ts'o if (test_opt(sb, DELALLOC)) 349056889787STheodore Ts'o clear_opt(sb, DELALLOC); 349156889787STheodore Ts'o } 349256889787STheodore Ts'o 3493ac27a0ecSDave Kleikamp sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 3494482a7425SDmitry Monakhov (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 3495ac27a0ecSDave Kleikamp 3496617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 3497617ba13bSMingming Cao (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) || 3498617ba13bSMingming Cao EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) || 3499617ba13bSMingming Cao EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U))) 3500b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 3501b31e1552SEric Sandeen "feature flags set on rev 0 fs, " 3502b31e1552SEric Sandeen "running e2fsck is recommended"); 3503469108ffSTheodore Tso 35042035e776STheodore Ts'o if (IS_EXT2_SB(sb)) { 35052035e776STheodore Ts'o if (ext2_feature_set_ok(sb)) 35062035e776STheodore Ts'o ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 35072035e776STheodore Ts'o "using the ext4 subsystem"); 35082035e776STheodore Ts'o else { 35092035e776STheodore Ts'o ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 35102035e776STheodore Ts'o "to feature incompatibilities"); 35112035e776STheodore Ts'o goto failed_mount; 35122035e776STheodore Ts'o } 35132035e776STheodore Ts'o } 35142035e776STheodore Ts'o 35152035e776STheodore Ts'o if (IS_EXT3_SB(sb)) { 35162035e776STheodore Ts'o if (ext3_feature_set_ok(sb)) 35172035e776STheodore Ts'o ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 35182035e776STheodore Ts'o "using the ext4 subsystem"); 35192035e776STheodore Ts'o else { 35202035e776STheodore Ts'o ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 35212035e776STheodore Ts'o "to feature incompatibilities"); 35222035e776STheodore Ts'o goto failed_mount; 35232035e776STheodore Ts'o } 35242035e776STheodore Ts'o } 35252035e776STheodore Ts'o 3526469108ffSTheodore Tso /* 3527ac27a0ecSDave Kleikamp * Check feature flags regardless of the revision level, since we 3528ac27a0ecSDave Kleikamp * previously didn't change the revision level when setting the flags, 3529ac27a0ecSDave Kleikamp * so there is a chance incompat flags are set on a rev 0 filesystem. 3530ac27a0ecSDave Kleikamp */ 3531a13fb1a4SEric Sandeen if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY))) 3532ac27a0ecSDave Kleikamp goto failed_mount; 3533a13fb1a4SEric Sandeen 3534261cb20cSJan Kara blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 3535617ba13bSMingming Cao if (blocksize < EXT4_MIN_BLOCK_SIZE || 3536617ba13bSMingming Cao blocksize > EXT4_MAX_BLOCK_SIZE) { 3537b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3538b31e1552SEric Sandeen "Unsupported filesystem blocksize %d", blocksize); 3539ac27a0ecSDave Kleikamp goto failed_mount; 3540ac27a0ecSDave Kleikamp } 3541ac27a0ecSDave Kleikamp 3542ac27a0ecSDave Kleikamp if (sb->s_blocksize != blocksize) { 3543ce40733cSAneesh Kumar K.V /* Validate the filesystem blocksize */ 3544ce40733cSAneesh Kumar K.V if (!sb_set_blocksize(sb, blocksize)) { 3545b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "bad block size %d", 3546ce40733cSAneesh Kumar K.V blocksize); 3547ac27a0ecSDave Kleikamp goto failed_mount; 3548ac27a0ecSDave Kleikamp } 3549ac27a0ecSDave Kleikamp 3550ac27a0ecSDave Kleikamp brelse(bh); 355170bbb3e0SAndrew Morton logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 355270bbb3e0SAndrew Morton offset = do_div(logical_sb_block, blocksize); 355370bbb3e0SAndrew Morton bh = sb_bread(sb, logical_sb_block); 3554ac27a0ecSDave Kleikamp if (!bh) { 3555b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3556b31e1552SEric Sandeen "Can't read superblock on 2nd try"); 3557ac27a0ecSDave Kleikamp goto failed_mount; 3558ac27a0ecSDave Kleikamp } 35592716b802STheodore Ts'o es = (struct ext4_super_block *)(bh->b_data + offset); 3560ac27a0ecSDave Kleikamp sbi->s_es = es; 3561617ba13bSMingming Cao if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 3562b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3563b31e1552SEric Sandeen "Magic mismatch, very weird!"); 3564ac27a0ecSDave Kleikamp goto failed_mount; 3565ac27a0ecSDave Kleikamp } 3566ac27a0ecSDave Kleikamp } 3567ac27a0ecSDave Kleikamp 3568a13fb1a4SEric Sandeen has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb, 3569a13fb1a4SEric Sandeen EXT4_FEATURE_RO_COMPAT_HUGE_FILE); 3570f287a1a5STheodore Ts'o sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 3571f287a1a5STheodore Ts'o has_huge_files); 3572f287a1a5STheodore Ts'o sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 3573ac27a0ecSDave Kleikamp 3574617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 3575617ba13bSMingming Cao sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 3576617ba13bSMingming Cao sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 3577ac27a0ecSDave Kleikamp } else { 3578ac27a0ecSDave Kleikamp sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 3579ac27a0ecSDave Kleikamp sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 3580617ba13bSMingming Cao if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 35811330593eSVignesh Babu (!is_power_of_2(sbi->s_inode_size)) || 3582ac27a0ecSDave Kleikamp (sbi->s_inode_size > blocksize)) { 3583b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3584b31e1552SEric Sandeen "unsupported inode size: %d", 3585ac27a0ecSDave Kleikamp sbi->s_inode_size); 3586ac27a0ecSDave Kleikamp goto failed_mount; 3587ac27a0ecSDave Kleikamp } 3588ef7f3835SKalpak Shah if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) 3589ef7f3835SKalpak Shah sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2); 3590ac27a0ecSDave Kleikamp } 35910b8e58a1SAndreas Dilger 35920d1ee42fSAlexandre Ratchov sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 35930d1ee42fSAlexandre Ratchov if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) { 35948fadc143SAlexandre Ratchov if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 35950d1ee42fSAlexandre Ratchov sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 3596d8ea6cf8Svignesh babu !is_power_of_2(sbi->s_desc_size)) { 3597b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3598b31e1552SEric Sandeen "unsupported descriptor size %lu", 35990d1ee42fSAlexandre Ratchov sbi->s_desc_size); 36000d1ee42fSAlexandre Ratchov goto failed_mount; 36010d1ee42fSAlexandre Ratchov } 36020d1ee42fSAlexandre Ratchov } else 36030d1ee42fSAlexandre Ratchov sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 36040b8e58a1SAndreas Dilger 3605ac27a0ecSDave Kleikamp sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 3606ac27a0ecSDave Kleikamp sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 3607b47b6f38SAndries E. Brouwer if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0) 3608617ba13bSMingming Cao goto cantfind_ext4; 36090b8e58a1SAndreas Dilger 3610617ba13bSMingming Cao sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 3611ac27a0ecSDave Kleikamp if (sbi->s_inodes_per_block == 0) 3612617ba13bSMingming Cao goto cantfind_ext4; 3613ac27a0ecSDave Kleikamp sbi->s_itb_per_group = sbi->s_inodes_per_group / 3614ac27a0ecSDave Kleikamp sbi->s_inodes_per_block; 36150d1ee42fSAlexandre Ratchov sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 3616ac27a0ecSDave Kleikamp sbi->s_sbh = bh; 3617ac27a0ecSDave Kleikamp sbi->s_mount_state = le16_to_cpu(es->s_state); 3618e57aa839SFengguang Wu sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 3619e57aa839SFengguang Wu sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 36200b8e58a1SAndreas Dilger 3621ac27a0ecSDave Kleikamp for (i = 0; i < 4; i++) 3622ac27a0ecSDave Kleikamp sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 3623ac27a0ecSDave Kleikamp sbi->s_def_hash_version = es->s_def_hash_version; 3624f99b2589STheodore Ts'o i = le32_to_cpu(es->s_flags); 3625f99b2589STheodore Ts'o if (i & EXT2_FLAGS_UNSIGNED_HASH) 3626f99b2589STheodore Ts'o sbi->s_hash_unsigned = 3; 3627f99b2589STheodore Ts'o else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 3628f99b2589STheodore Ts'o #ifdef __CHAR_UNSIGNED__ 3629f99b2589STheodore Ts'o es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 3630f99b2589STheodore Ts'o sbi->s_hash_unsigned = 3; 3631f99b2589STheodore Ts'o #else 3632f99b2589STheodore Ts'o es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 3633f99b2589STheodore Ts'o #endif 3634f99b2589STheodore Ts'o } 3635ac27a0ecSDave Kleikamp 3636281b5995STheodore Ts'o /* Handle clustersize */ 3637281b5995STheodore Ts'o clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 3638281b5995STheodore Ts'o has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb, 3639281b5995STheodore Ts'o EXT4_FEATURE_RO_COMPAT_BIGALLOC); 3640281b5995STheodore Ts'o if (has_bigalloc) { 3641281b5995STheodore Ts'o if (clustersize < blocksize) { 3642281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, 3643281b5995STheodore Ts'o "cluster size (%d) smaller than " 3644281b5995STheodore Ts'o "block size (%d)", clustersize, blocksize); 3645281b5995STheodore Ts'o goto failed_mount; 3646281b5995STheodore Ts'o } 3647281b5995STheodore Ts'o sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 3648281b5995STheodore Ts'o le32_to_cpu(es->s_log_block_size); 3649281b5995STheodore Ts'o sbi->s_clusters_per_group = 3650281b5995STheodore Ts'o le32_to_cpu(es->s_clusters_per_group); 3651281b5995STheodore Ts'o if (sbi->s_clusters_per_group > blocksize * 8) { 3652281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, 3653281b5995STheodore Ts'o "#clusters per group too big: %lu", 3654281b5995STheodore Ts'o sbi->s_clusters_per_group); 3655281b5995STheodore Ts'o goto failed_mount; 3656281b5995STheodore Ts'o } 3657281b5995STheodore Ts'o if (sbi->s_blocks_per_group != 3658281b5995STheodore Ts'o (sbi->s_clusters_per_group * (clustersize / blocksize))) { 3659281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 3660281b5995STheodore Ts'o "clusters per group (%lu) inconsistent", 3661281b5995STheodore Ts'o sbi->s_blocks_per_group, 3662281b5995STheodore Ts'o sbi->s_clusters_per_group); 3663281b5995STheodore Ts'o goto failed_mount; 3664281b5995STheodore Ts'o } 3665281b5995STheodore Ts'o } else { 3666281b5995STheodore Ts'o if (clustersize != blocksize) { 3667281b5995STheodore Ts'o ext4_warning(sb, "fragment/cluster size (%d) != " 3668281b5995STheodore Ts'o "block size (%d)", clustersize, 3669281b5995STheodore Ts'o blocksize); 3670281b5995STheodore Ts'o clustersize = blocksize; 3671281b5995STheodore Ts'o } 3672ac27a0ecSDave Kleikamp if (sbi->s_blocks_per_group > blocksize * 8) { 3673b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3674b31e1552SEric Sandeen "#blocks per group too big: %lu", 3675ac27a0ecSDave Kleikamp sbi->s_blocks_per_group); 3676ac27a0ecSDave Kleikamp goto failed_mount; 3677ac27a0ecSDave Kleikamp } 3678281b5995STheodore Ts'o sbi->s_clusters_per_group = sbi->s_blocks_per_group; 3679281b5995STheodore Ts'o sbi->s_cluster_bits = 0; 3680281b5995STheodore Ts'o } 3681281b5995STheodore Ts'o sbi->s_cluster_ratio = clustersize / blocksize; 3682281b5995STheodore Ts'o 3683ac27a0ecSDave Kleikamp if (sbi->s_inodes_per_group > blocksize * 8) { 3684b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3685b31e1552SEric Sandeen "#inodes per group too big: %lu", 3686ac27a0ecSDave Kleikamp sbi->s_inodes_per_group); 3687ac27a0ecSDave Kleikamp goto failed_mount; 3688ac27a0ecSDave Kleikamp } 3689ac27a0ecSDave Kleikamp 3690960fd856STheodore Ts'o /* Do we have standard group size of clustersize * 8 blocks ? */ 3691960fd856STheodore Ts'o if (sbi->s_blocks_per_group == clustersize << 3) 3692960fd856STheodore Ts'o set_opt2(sb, STD_GROUP_SIZE); 3693960fd856STheodore Ts'o 3694bf43d84bSEric Sandeen /* 3695bf43d84bSEric Sandeen * Test whether we have more sectors than will fit in sector_t, 3696bf43d84bSEric Sandeen * and whether the max offset is addressable by the page cache. 3697bf43d84bSEric Sandeen */ 36985a9ae68aSDarrick J. Wong err = generic_check_addressable(sb->s_blocksize_bits, 369930ca22c7SPatrick J. LoPresti ext4_blocks_count(es)); 37005a9ae68aSDarrick J. Wong if (err) { 3701b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "filesystem" 3702bf43d84bSEric Sandeen " too large to mount safely on this system"); 3703ac27a0ecSDave Kleikamp if (sizeof(sector_t) < 8) 370490c699a9SBartlomiej Zolnierkiewicz ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); 3705ac27a0ecSDave Kleikamp goto failed_mount; 3706ac27a0ecSDave Kleikamp } 3707ac27a0ecSDave Kleikamp 3708617ba13bSMingming Cao if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 3709617ba13bSMingming Cao goto cantfind_ext4; 3710e7c95593SEric Sandeen 37110f2ddca6SFrom: Thiemo Nagel /* check blocks count against device size */ 37120f2ddca6SFrom: Thiemo Nagel blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 37130f2ddca6SFrom: Thiemo Nagel if (blocks_count && ext4_blocks_count(es) > blocks_count) { 3714b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 3715b31e1552SEric Sandeen "exceeds size of device (%llu blocks)", 37160f2ddca6SFrom: Thiemo Nagel ext4_blocks_count(es), blocks_count); 37170f2ddca6SFrom: Thiemo Nagel goto failed_mount; 37180f2ddca6SFrom: Thiemo Nagel } 37190f2ddca6SFrom: Thiemo Nagel 37204ec11028STheodore Ts'o /* 37214ec11028STheodore Ts'o * It makes no sense for the first data block to be beyond the end 37224ec11028STheodore Ts'o * of the filesystem. 37234ec11028STheodore Ts'o */ 37244ec11028STheodore Ts'o if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 3725b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 3726b31e1552SEric Sandeen "block %u is beyond end of filesystem (%llu)", 3727e7c95593SEric Sandeen le32_to_cpu(es->s_first_data_block), 37284ec11028STheodore Ts'o ext4_blocks_count(es)); 3729e7c95593SEric Sandeen goto failed_mount; 3730e7c95593SEric Sandeen } 3731bd81d8eeSLaurent Vivier blocks_count = (ext4_blocks_count(es) - 3732bd81d8eeSLaurent Vivier le32_to_cpu(es->s_first_data_block) + 3733bd81d8eeSLaurent Vivier EXT4_BLOCKS_PER_GROUP(sb) - 1); 3734bd81d8eeSLaurent Vivier do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 37354ec11028STheodore Ts'o if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 3736b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "groups count too large: %u " 37374ec11028STheodore Ts'o "(block count %llu, first data block %u, " 3738b31e1552SEric Sandeen "blocks per group %lu)", sbi->s_groups_count, 37394ec11028STheodore Ts'o ext4_blocks_count(es), 37404ec11028STheodore Ts'o le32_to_cpu(es->s_first_data_block), 37414ec11028STheodore Ts'o EXT4_BLOCKS_PER_GROUP(sb)); 37424ec11028STheodore Ts'o goto failed_mount; 37434ec11028STheodore Ts'o } 3744bd81d8eeSLaurent Vivier sbi->s_groups_count = blocks_count; 3745fb0a387dSEric Sandeen sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 3746fb0a387dSEric Sandeen (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 3747617ba13bSMingming Cao db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 3748617ba13bSMingming Cao EXT4_DESC_PER_BLOCK(sb); 3749f18a5f21STheodore Ts'o sbi->s_group_desc = ext4_kvmalloc(db_count * 3750f18a5f21STheodore Ts'o sizeof(struct buffer_head *), 3751ac27a0ecSDave Kleikamp GFP_KERNEL); 3752ac27a0ecSDave Kleikamp if (sbi->s_group_desc == NULL) { 3753b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "not enough memory"); 37542cde417dSTheodore Ts'o ret = -ENOMEM; 3755ac27a0ecSDave Kleikamp goto failed_mount; 3756ac27a0ecSDave Kleikamp } 3757ac27a0ecSDave Kleikamp 37589f6200bbSTheodore Ts'o if (ext4_proc_root) 37599f6200bbSTheodore Ts'o sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root); 3760240799cdSTheodore Ts'o 376166acdcf4STheodore Ts'o if (sbi->s_proc) 376266acdcf4STheodore Ts'o proc_create_data("options", S_IRUGO, sbi->s_proc, 376366acdcf4STheodore Ts'o &ext4_seq_options_fops, sb); 376466acdcf4STheodore Ts'o 3765705895b6SPekka Enberg bgl_lock_init(sbi->s_blockgroup_lock); 3766ac27a0ecSDave Kleikamp 3767ac27a0ecSDave Kleikamp for (i = 0; i < db_count; i++) { 376870bbb3e0SAndrew Morton block = descriptor_loc(sb, logical_sb_block, i); 3769ac27a0ecSDave Kleikamp sbi->s_group_desc[i] = sb_bread(sb, block); 3770ac27a0ecSDave Kleikamp if (!sbi->s_group_desc[i]) { 3771b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3772b31e1552SEric Sandeen "can't read group descriptor %d", i); 3773ac27a0ecSDave Kleikamp db_count = i; 3774ac27a0ecSDave Kleikamp goto failed_mount2; 3775ac27a0ecSDave Kleikamp } 3776ac27a0ecSDave Kleikamp } 3777bfff6873SLukas Czerner if (!ext4_check_descriptors(sb, &first_not_zeroed)) { 3778b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 3779ac27a0ecSDave Kleikamp goto failed_mount2; 3780ac27a0ecSDave Kleikamp } 3781772cb7c8SJose R. Santos if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) 3782772cb7c8SJose R. Santos if (!ext4_fill_flex_info(sb)) { 3783b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3784b31e1552SEric Sandeen "unable to initialize " 3785b31e1552SEric Sandeen "flex_bg meta info!"); 3786772cb7c8SJose R. Santos goto failed_mount2; 3787772cb7c8SJose R. Santos } 3788772cb7c8SJose R. Santos 3789ac27a0ecSDave Kleikamp sbi->s_gdb_count = db_count; 3790ac27a0ecSDave Kleikamp get_random_bytes(&sbi->s_next_generation, sizeof(u32)); 3791ac27a0ecSDave Kleikamp spin_lock_init(&sbi->s_next_gen_lock); 3792ac27a0ecSDave Kleikamp 379304496411STao Ma init_timer(&sbi->s_err_report); 379404496411STao Ma sbi->s_err_report.function = print_daily_error_info; 379504496411STao Ma sbi->s_err_report.data = (unsigned long) sb; 379604496411STao Ma 3797a75ae78fSDmitry Monakhov /* Register extent status tree shrinker */ 3798d3922a77SZheng Liu ext4_es_register_shrinker(sbi); 3799a75ae78fSDmitry Monakhov 380057042651STheodore Ts'o err = percpu_counter_init(&sbi->s_freeclusters_counter, 38015dee5437STheodore Ts'o ext4_count_free_clusters(sb)); 3802ce7e010aSTheodore Ts'o if (!err) { 3803ce7e010aSTheodore Ts'o err = percpu_counter_init(&sbi->s_freeinodes_counter, 3804ce7e010aSTheodore Ts'o ext4_count_free_inodes(sb)); 3805ce7e010aSTheodore Ts'o } 3806ce7e010aSTheodore Ts'o if (!err) { 3807ce7e010aSTheodore Ts'o err = percpu_counter_init(&sbi->s_dirs_counter, 3808ce7e010aSTheodore Ts'o ext4_count_dirs(sb)); 3809ce7e010aSTheodore Ts'o } 3810ce7e010aSTheodore Ts'o if (!err) { 381157042651STheodore Ts'o err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0); 3812ce7e010aSTheodore Ts'o } 38131ac6466fSTheodore Ts'o if (!err) { 38141ac6466fSTheodore Ts'o err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0); 38151ac6466fSTheodore Ts'o } 3816ce7e010aSTheodore Ts'o if (err) { 3817ce7e010aSTheodore Ts'o ext4_msg(sb, KERN_ERR, "insufficient memory"); 3818ce7e010aSTheodore Ts'o goto failed_mount3; 3819ce7e010aSTheodore Ts'o } 3820ce7e010aSTheodore Ts'o 3821c9de560dSAlex Tomas sbi->s_stripe = ext4_get_stripe_size(sbi); 382267a5da56SZheng Liu sbi->s_extent_max_zeroout_kb = 32; 3823c9de560dSAlex Tomas 3824ac27a0ecSDave Kleikamp /* 3825ac27a0ecSDave Kleikamp * set up enough so that it can read an inode 3826ac27a0ecSDave Kleikamp */ 38279ca92389STheodore Ts'o if (!test_opt(sb, NOLOAD) && 38289ca92389STheodore Ts'o EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) 3829617ba13bSMingming Cao sb->s_op = &ext4_sops; 38309ca92389STheodore Ts'o else 38319ca92389STheodore Ts'o sb->s_op = &ext4_nojournal_sops; 3832617ba13bSMingming Cao sb->s_export_op = &ext4_export_ops; 3833617ba13bSMingming Cao sb->s_xattr = ext4_xattr_handlers; 3834ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 3835617ba13bSMingming Cao sb->dq_op = &ext4_quota_operations; 3836262b4662SJan Kara if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) 38377c319d32SAditya Kali sb->s_qcop = &ext4_qctl_sysfile_operations; 3838262b4662SJan Kara else 3839262b4662SJan Kara sb->s_qcop = &ext4_qctl_operations; 3840ac27a0ecSDave Kleikamp #endif 3841f2fa2ffcSAneesh Kumar K.V memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 3842f2fa2ffcSAneesh Kumar K.V 3843ac27a0ecSDave Kleikamp INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 38443b9d4ed2STheodore Ts'o mutex_init(&sbi->s_orphan_lock); 3845ac27a0ecSDave Kleikamp 3846ac27a0ecSDave Kleikamp sb->s_root = NULL; 3847ac27a0ecSDave Kleikamp 3848ac27a0ecSDave Kleikamp needs_recovery = (es->s_last_orphan != 0 || 3849617ba13bSMingming Cao EXT4_HAS_INCOMPAT_FEATURE(sb, 3850617ba13bSMingming Cao EXT4_FEATURE_INCOMPAT_RECOVER)); 3851ac27a0ecSDave Kleikamp 3852c5e06d10SJohann Lombardi if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) && 3853c5e06d10SJohann Lombardi !(sb->s_flags & MS_RDONLY)) 3854c5e06d10SJohann Lombardi if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 3855c5e06d10SJohann Lombardi goto failed_mount3; 3856c5e06d10SJohann Lombardi 3857ac27a0ecSDave Kleikamp /* 3858ac27a0ecSDave Kleikamp * The first inode we look at is the journal inode. Don't try 3859ac27a0ecSDave Kleikamp * root first: it may be modified in the journal! 3860ac27a0ecSDave Kleikamp */ 3861ac27a0ecSDave Kleikamp if (!test_opt(sb, NOLOAD) && 3862617ba13bSMingming Cao EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { 3863617ba13bSMingming Cao if (ext4_load_journal(sb, es, journal_devnum)) 3864ac27a0ecSDave Kleikamp goto failed_mount3; 38650390131bSFrank Mayhar } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) && 38660390131bSFrank Mayhar EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { 3867b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "required journal recovery " 3868b31e1552SEric Sandeen "suppressed and not mounted read-only"); 3869744692dcSJiaying Zhang goto failed_mount_wq; 3870ac27a0ecSDave Kleikamp } else { 3871fd8c37ecSTheodore Ts'o clear_opt(sb, DATA_FLAGS); 38720390131bSFrank Mayhar sbi->s_journal = NULL; 38730390131bSFrank Mayhar needs_recovery = 0; 38740390131bSFrank Mayhar goto no_journal; 3875ac27a0ecSDave Kleikamp } 3876ac27a0ecSDave Kleikamp 3877f32aaf2dSTheodore Ts'o if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) && 3878eb40a09cSJose R. Santos !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 3879eb40a09cSJose R. Santos JBD2_FEATURE_INCOMPAT_64BIT)) { 3880b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 3881744692dcSJiaying Zhang goto failed_mount_wq; 3882eb40a09cSJose R. Santos } 3883eb40a09cSJose R. Santos 388425ed6e8aSDarrick J. Wong if (!set_journal_csum_feature_set(sb)) { 388525ed6e8aSDarrick J. Wong ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 388625ed6e8aSDarrick J. Wong "feature set"); 388725ed6e8aSDarrick J. Wong goto failed_mount_wq; 3888d4da6c9cSLinus Torvalds } 3889818d276cSGirish Shilamkar 3890ac27a0ecSDave Kleikamp /* We have now updated the journal if required, so we can 3891ac27a0ecSDave Kleikamp * validate the data journaling mode. */ 3892ac27a0ecSDave Kleikamp switch (test_opt(sb, DATA_FLAGS)) { 3893ac27a0ecSDave Kleikamp case 0: 3894ac27a0ecSDave Kleikamp /* No mode set, assume a default based on the journal 389563f57933SAndrew Morton * capabilities: ORDERED_DATA if the journal can 389663f57933SAndrew Morton * cope, else JOURNAL_DATA 389763f57933SAndrew Morton */ 3898dab291afSMingming Cao if (jbd2_journal_check_available_features 3899dab291afSMingming Cao (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) 3900fd8c37ecSTheodore Ts'o set_opt(sb, ORDERED_DATA); 3901ac27a0ecSDave Kleikamp else 3902fd8c37ecSTheodore Ts'o set_opt(sb, JOURNAL_DATA); 3903ac27a0ecSDave Kleikamp break; 3904ac27a0ecSDave Kleikamp 3905617ba13bSMingming Cao case EXT4_MOUNT_ORDERED_DATA: 3906617ba13bSMingming Cao case EXT4_MOUNT_WRITEBACK_DATA: 3907dab291afSMingming Cao if (!jbd2_journal_check_available_features 3908dab291afSMingming Cao (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 3909b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Journal does not support " 3910b31e1552SEric Sandeen "requested data journaling mode"); 3911744692dcSJiaying Zhang goto failed_mount_wq; 3912ac27a0ecSDave Kleikamp } 3913ac27a0ecSDave Kleikamp default: 3914ac27a0ecSDave Kleikamp break; 3915ac27a0ecSDave Kleikamp } 3916b3881f74STheodore Ts'o set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 3917ac27a0ecSDave Kleikamp 391818aadd47SBobi Jam sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; 391918aadd47SBobi Jam 3920ce7e010aSTheodore Ts'o /* 3921ce7e010aSTheodore Ts'o * The journal may have updated the bg summary counts, so we 3922ce7e010aSTheodore Ts'o * need to update the global counters. 3923ce7e010aSTheodore Ts'o */ 392457042651STheodore Ts'o percpu_counter_set(&sbi->s_freeclusters_counter, 39255dee5437STheodore Ts'o ext4_count_free_clusters(sb)); 3926ce7e010aSTheodore Ts'o percpu_counter_set(&sbi->s_freeinodes_counter, 392784061e07SDmitry Monakhov ext4_count_free_inodes(sb)); 3928ce7e010aSTheodore Ts'o percpu_counter_set(&sbi->s_dirs_counter, 392984061e07SDmitry Monakhov ext4_count_dirs(sb)); 393057042651STheodore Ts'o percpu_counter_set(&sbi->s_dirtyclusters_counter, 0); 3931206f7ab4SChristoph Hellwig 3932ce7e010aSTheodore Ts'o no_journal: 3933fd89d5f2STejun Heo /* 3934952fc18eSTheodore Ts'o * Get the # of file system overhead blocks from the 3935952fc18eSTheodore Ts'o * superblock if present. 3936952fc18eSTheodore Ts'o */ 3937952fc18eSTheodore Ts'o if (es->s_overhead_clusters) 3938952fc18eSTheodore Ts'o sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 3939952fc18eSTheodore Ts'o else { 394007aa2ea1SLukas Czerner err = ext4_calculate_overhead(sb); 394107aa2ea1SLukas Czerner if (err) 3942952fc18eSTheodore Ts'o goto failed_mount_wq; 3943952fc18eSTheodore Ts'o } 3944952fc18eSTheodore Ts'o 3945952fc18eSTheodore Ts'o /* 3946fd89d5f2STejun Heo * The maximum number of concurrent works can be high and 3947fd89d5f2STejun Heo * concurrency isn't really necessary. Limit it to 1. 3948fd89d5f2STejun Heo */ 39492e8fa54eSJan Kara EXT4_SB(sb)->rsv_conversion_wq = 39502e8fa54eSJan Kara alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 39512e8fa54eSJan Kara if (!EXT4_SB(sb)->rsv_conversion_wq) { 39522e8fa54eSJan Kara printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 395307aa2ea1SLukas Czerner ret = -ENOMEM; 39542e8fa54eSJan Kara goto failed_mount4; 39552e8fa54eSJan Kara } 39562e8fa54eSJan Kara 39572e8fa54eSJan Kara EXT4_SB(sb)->unrsv_conversion_wq = 39582e8fa54eSJan Kara alloc_workqueue("ext4-unrsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 39592e8fa54eSJan Kara if (!EXT4_SB(sb)->unrsv_conversion_wq) { 39602e8fa54eSJan Kara printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 39612e8fa54eSJan Kara ret = -ENOMEM; 39622e8fa54eSJan Kara goto failed_mount4; 39634c0425ffSMingming Cao } 39644c0425ffSMingming Cao 3965ac27a0ecSDave Kleikamp /* 3966dab291afSMingming Cao * The jbd2_journal_load will have done any necessary log recovery, 3967ac27a0ecSDave Kleikamp * so we can safely mount the rest of the filesystem now. 3968ac27a0ecSDave Kleikamp */ 3969ac27a0ecSDave Kleikamp 39701d1fe1eeSDavid Howells root = ext4_iget(sb, EXT4_ROOT_INO); 39711d1fe1eeSDavid Howells if (IS_ERR(root)) { 3972b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "get root inode failed"); 39731d1fe1eeSDavid Howells ret = PTR_ERR(root); 397432a9bb57SManish Katiyar root = NULL; 3975ac27a0ecSDave Kleikamp goto failed_mount4; 3976ac27a0ecSDave Kleikamp } 3977ac27a0ecSDave Kleikamp if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 3978b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 397994bf608aSAl Viro iput(root); 3980ac27a0ecSDave Kleikamp goto failed_mount4; 3981ac27a0ecSDave Kleikamp } 398248fde701SAl Viro sb->s_root = d_make_root(root); 39831d1fe1eeSDavid Howells if (!sb->s_root) { 3984b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "get root dentry failed"); 39851d1fe1eeSDavid Howells ret = -ENOMEM; 39861d1fe1eeSDavid Howells goto failed_mount4; 39871d1fe1eeSDavid Howells } 3988ac27a0ecSDave Kleikamp 39897e84b621SEric Sandeen if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY)) 39907e84b621SEric Sandeen sb->s_flags |= MS_RDONLY; 3991ef7f3835SKalpak Shah 3992ef7f3835SKalpak Shah /* determine the minimum size of new large inodes, if present */ 3993ef7f3835SKalpak Shah if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 3994ef7f3835SKalpak Shah sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 3995ef7f3835SKalpak Shah EXT4_GOOD_OLD_INODE_SIZE; 3996ef7f3835SKalpak Shah if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 3997ef7f3835SKalpak Shah EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) { 3998ef7f3835SKalpak Shah if (sbi->s_want_extra_isize < 3999ef7f3835SKalpak Shah le16_to_cpu(es->s_want_extra_isize)) 4000ef7f3835SKalpak Shah sbi->s_want_extra_isize = 4001ef7f3835SKalpak Shah le16_to_cpu(es->s_want_extra_isize); 4002ef7f3835SKalpak Shah if (sbi->s_want_extra_isize < 4003ef7f3835SKalpak Shah le16_to_cpu(es->s_min_extra_isize)) 4004ef7f3835SKalpak Shah sbi->s_want_extra_isize = 4005ef7f3835SKalpak Shah le16_to_cpu(es->s_min_extra_isize); 4006ef7f3835SKalpak Shah } 4007ef7f3835SKalpak Shah } 4008ef7f3835SKalpak Shah /* Check if enough inode space is available */ 4009ef7f3835SKalpak Shah if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize > 4010ef7f3835SKalpak Shah sbi->s_inode_size) { 4011ef7f3835SKalpak Shah sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 4012ef7f3835SKalpak Shah EXT4_GOOD_OLD_INODE_SIZE; 4013b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "required extra inode space not" 4014b31e1552SEric Sandeen "available"); 4015ef7f3835SKalpak Shah } 4016ef7f3835SKalpak Shah 401727dd4385SLukas Czerner err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sbi)); 401827dd4385SLukas Czerner if (err) { 401927dd4385SLukas Czerner ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for " 402027dd4385SLukas Czerner "reserved pool", ext4_calculate_resv_clusters(sbi)); 402127dd4385SLukas Czerner goto failed_mount4a; 402227dd4385SLukas Czerner } 402327dd4385SLukas Czerner 40246fd058f7STheodore Ts'o err = ext4_setup_system_zone(sb); 40256fd058f7STheodore Ts'o if (err) { 4026b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "failed to initialize system " 4027fbe845ddSCurt Wohlgemuth "zone (%d)", err); 402894bf608aSAl Viro goto failed_mount4a; 40296fd058f7STheodore Ts'o } 40306fd058f7STheodore Ts'o 4031c2774d84SAneesh Kumar K.V ext4_ext_init(sb); 40329d99012fSAkira Fujita err = ext4_mb_init(sb); 4033c2774d84SAneesh Kumar K.V if (err) { 4034421f91d2SUwe Kleine-König ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 4035c2774d84SAneesh Kumar K.V err); 4036dcf2d804STao Ma goto failed_mount5; 4037c2774d84SAneesh Kumar K.V } 4038c2774d84SAneesh Kumar K.V 4039bfff6873SLukas Czerner err = ext4_register_li_request(sb, first_not_zeroed); 4040bfff6873SLukas Czerner if (err) 4041dcf2d804STao Ma goto failed_mount6; 4042bfff6873SLukas Czerner 40433197ebdbSTheodore Ts'o sbi->s_kobj.kset = ext4_kset; 40443197ebdbSTheodore Ts'o init_completion(&sbi->s_kobj_unregister); 40453197ebdbSTheodore Ts'o err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL, 40463197ebdbSTheodore Ts'o "%s", sb->s_id); 4047dcf2d804STao Ma if (err) 4048dcf2d804STao Ma goto failed_mount7; 40493197ebdbSTheodore Ts'o 40509b2ff357SJan Kara #ifdef CONFIG_QUOTA 40519b2ff357SJan Kara /* Enable quota usage during mount. */ 40529b2ff357SJan Kara if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) && 40539b2ff357SJan Kara !(sb->s_flags & MS_RDONLY)) { 40549b2ff357SJan Kara err = ext4_enable_quotas(sb); 40559b2ff357SJan Kara if (err) 40569b2ff357SJan Kara goto failed_mount8; 40579b2ff357SJan Kara } 40589b2ff357SJan Kara #endif /* CONFIG_QUOTA */ 40599b2ff357SJan Kara 4060617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 4061617ba13bSMingming Cao ext4_orphan_cleanup(sb, es); 4062617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 40630390131bSFrank Mayhar if (needs_recovery) { 4064b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "recovery complete"); 4065617ba13bSMingming Cao ext4_mark_recovery_complete(sb, es); 40660390131bSFrank Mayhar } 40670390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal) { 40680390131bSFrank Mayhar if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 40690390131bSFrank Mayhar descr = " journalled data mode"; 40700390131bSFrank Mayhar else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 40710390131bSFrank Mayhar descr = " ordered data mode"; 40720390131bSFrank Mayhar else 40730390131bSFrank Mayhar descr = " writeback data mode"; 40740390131bSFrank Mayhar } else 40750390131bSFrank Mayhar descr = "out journal"; 40760390131bSFrank Mayhar 407779add3a3SLukas Czerner if (test_opt(sb, DISCARD)) { 407879add3a3SLukas Czerner struct request_queue *q = bdev_get_queue(sb->s_bdev); 407979add3a3SLukas Czerner if (!blk_queue_discard(q)) 408079add3a3SLukas Czerner ext4_msg(sb, KERN_WARNING, 408179add3a3SLukas Czerner "mounting with \"discard\" option, but " 408279add3a3SLukas Czerner "the device does not support discard"); 408379add3a3SLukas Czerner } 408479add3a3SLukas Czerner 4085d4c402d9SCurt Wohlgemuth ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 40868b67f04aSTheodore Ts'o "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts, 40878b67f04aSTheodore Ts'o *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 4088ac27a0ecSDave Kleikamp 408966e61a9eSTheodore Ts'o if (es->s_error_count) 409066e61a9eSTheodore Ts'o mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 4091ac27a0ecSDave Kleikamp 4092d4c402d9SCurt Wohlgemuth kfree(orig_data); 4093ac27a0ecSDave Kleikamp return 0; 4094ac27a0ecSDave Kleikamp 4095617ba13bSMingming Cao cantfind_ext4: 4096ac27a0ecSDave Kleikamp if (!silent) 4097b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 4098ac27a0ecSDave Kleikamp goto failed_mount; 4099ac27a0ecSDave Kleikamp 410072ba7450STheodore Ts'o #ifdef CONFIG_QUOTA 410172ba7450STheodore Ts'o failed_mount8: 410272ba7450STheodore Ts'o kobject_del(&sbi->s_kobj); 410372ba7450STheodore Ts'o #endif 4104dcf2d804STao Ma failed_mount7: 4105dcf2d804STao Ma ext4_unregister_li_request(sb); 4106dcf2d804STao Ma failed_mount6: 4107dcf2d804STao Ma ext4_mb_release(sb); 410894bf608aSAl Viro failed_mount5: 410994bf608aSAl Viro ext4_ext_release(sb); 4110dcf2d804STao Ma ext4_release_system_zone(sb); 411194bf608aSAl Viro failed_mount4a: 411294bf608aSAl Viro dput(sb->s_root); 411332a9bb57SManish Katiyar sb->s_root = NULL; 411494bf608aSAl Viro failed_mount4: 4115b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "mount failed"); 41162e8fa54eSJan Kara if (EXT4_SB(sb)->rsv_conversion_wq) 41172e8fa54eSJan Kara destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 41182e8fa54eSJan Kara if (EXT4_SB(sb)->unrsv_conversion_wq) 41192e8fa54eSJan Kara destroy_workqueue(EXT4_SB(sb)->unrsv_conversion_wq); 41204c0425ffSMingming Cao failed_mount_wq: 41210390131bSFrank Mayhar if (sbi->s_journal) { 4122dab291afSMingming Cao jbd2_journal_destroy(sbi->s_journal); 412347b4a50bSJan Kara sbi->s_journal = NULL; 41240390131bSFrank Mayhar } 4125ac27a0ecSDave Kleikamp failed_mount3: 4126d3922a77SZheng Liu ext4_es_unregister_shrinker(sbi); 412704496411STao Ma del_timer(&sbi->s_err_report); 41289933fc0aSTheodore Ts'o if (sbi->s_flex_groups) 41299933fc0aSTheodore Ts'o ext4_kvfree(sbi->s_flex_groups); 413057042651STheodore Ts'o percpu_counter_destroy(&sbi->s_freeclusters_counter); 4131ce7e010aSTheodore Ts'o percpu_counter_destroy(&sbi->s_freeinodes_counter); 4132ce7e010aSTheodore Ts'o percpu_counter_destroy(&sbi->s_dirs_counter); 413357042651STheodore Ts'o percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 41341ac6466fSTheodore Ts'o percpu_counter_destroy(&sbi->s_extent_cache_cnt); 4135c5e06d10SJohann Lombardi if (sbi->s_mmp_tsk) 4136c5e06d10SJohann Lombardi kthread_stop(sbi->s_mmp_tsk); 4137ac27a0ecSDave Kleikamp failed_mount2: 4138ac27a0ecSDave Kleikamp for (i = 0; i < db_count; i++) 4139ac27a0ecSDave Kleikamp brelse(sbi->s_group_desc[i]); 4140f18a5f21STheodore Ts'o ext4_kvfree(sbi->s_group_desc); 4141ac27a0ecSDave Kleikamp failed_mount: 41420441984aSDarrick J. Wong if (sbi->s_chksum_driver) 41430441984aSDarrick J. Wong crypto_free_shash(sbi->s_chksum_driver); 4144240799cdSTheodore Ts'o if (sbi->s_proc) { 414566acdcf4STheodore Ts'o remove_proc_entry("options", sbi->s_proc); 41469f6200bbSTheodore Ts'o remove_proc_entry(sb->s_id, ext4_proc_root); 4147240799cdSTheodore Ts'o } 4148ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4149ac27a0ecSDave Kleikamp for (i = 0; i < MAXQUOTAS; i++) 4150ac27a0ecSDave Kleikamp kfree(sbi->s_qf_names[i]); 4151ac27a0ecSDave Kleikamp #endif 4152617ba13bSMingming Cao ext4_blkdev_remove(sbi); 4153ac27a0ecSDave Kleikamp brelse(bh); 4154ac27a0ecSDave Kleikamp out_fail: 4155ac27a0ecSDave Kleikamp sb->s_fs_info = NULL; 4156f6830165SManish Katiyar kfree(sbi->s_blockgroup_lock); 4157ac27a0ecSDave Kleikamp kfree(sbi); 4158dcc7dae3SCyrill Gorcunov out_free_orig: 4159d4c402d9SCurt Wohlgemuth kfree(orig_data); 416007aa2ea1SLukas Czerner return err ? err : ret; 4161ac27a0ecSDave Kleikamp } 4162ac27a0ecSDave Kleikamp 4163ac27a0ecSDave Kleikamp /* 4164ac27a0ecSDave Kleikamp * Setup any per-fs journal parameters now. We'll do this both on 4165ac27a0ecSDave Kleikamp * initial mount, once the journal has been initialised but before we've 4166ac27a0ecSDave Kleikamp * done any recovery; and again on any subsequent remount. 4167ac27a0ecSDave Kleikamp */ 4168617ba13bSMingming Cao static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 4169ac27a0ecSDave Kleikamp { 4170617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 4171ac27a0ecSDave Kleikamp 4172ac27a0ecSDave Kleikamp journal->j_commit_interval = sbi->s_commit_interval; 417330773840STheodore Ts'o journal->j_min_batch_time = sbi->s_min_batch_time; 417430773840STheodore Ts'o journal->j_max_batch_time = sbi->s_max_batch_time; 4175ac27a0ecSDave Kleikamp 4176a931da6aSTheodore Ts'o write_lock(&journal->j_state_lock); 4177ac27a0ecSDave Kleikamp if (test_opt(sb, BARRIER)) 4178dab291afSMingming Cao journal->j_flags |= JBD2_BARRIER; 4179ac27a0ecSDave Kleikamp else 4180dab291afSMingming Cao journal->j_flags &= ~JBD2_BARRIER; 41815bf5683aSHidehiro Kawai if (test_opt(sb, DATA_ERR_ABORT)) 41825bf5683aSHidehiro Kawai journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 41835bf5683aSHidehiro Kawai else 41845bf5683aSHidehiro Kawai journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 4185a931da6aSTheodore Ts'o write_unlock(&journal->j_state_lock); 4186ac27a0ecSDave Kleikamp } 4187ac27a0ecSDave Kleikamp 4188617ba13bSMingming Cao static journal_t *ext4_get_journal(struct super_block *sb, 4189ac27a0ecSDave Kleikamp unsigned int journal_inum) 4190ac27a0ecSDave Kleikamp { 4191ac27a0ecSDave Kleikamp struct inode *journal_inode; 4192ac27a0ecSDave Kleikamp journal_t *journal; 4193ac27a0ecSDave Kleikamp 41940390131bSFrank Mayhar BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 41950390131bSFrank Mayhar 4196ac27a0ecSDave Kleikamp /* First, test for the existence of a valid inode on disk. Bad 4197ac27a0ecSDave Kleikamp * things happen if we iget() an unused inode, as the subsequent 4198ac27a0ecSDave Kleikamp * iput() will try to delete it. */ 4199ac27a0ecSDave Kleikamp 42001d1fe1eeSDavid Howells journal_inode = ext4_iget(sb, journal_inum); 42011d1fe1eeSDavid Howells if (IS_ERR(journal_inode)) { 4202b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "no journal found"); 4203ac27a0ecSDave Kleikamp return NULL; 4204ac27a0ecSDave Kleikamp } 4205ac27a0ecSDave Kleikamp if (!journal_inode->i_nlink) { 4206ac27a0ecSDave Kleikamp make_bad_inode(journal_inode); 4207ac27a0ecSDave Kleikamp iput(journal_inode); 4208b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 4209ac27a0ecSDave Kleikamp return NULL; 4210ac27a0ecSDave Kleikamp } 4211ac27a0ecSDave Kleikamp 4212e5f8eab8STheodore Ts'o jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 4213ac27a0ecSDave Kleikamp journal_inode, journal_inode->i_size); 42141d1fe1eeSDavid Howells if (!S_ISREG(journal_inode->i_mode)) { 4215b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "invalid journal inode"); 4216ac27a0ecSDave Kleikamp iput(journal_inode); 4217ac27a0ecSDave Kleikamp return NULL; 4218ac27a0ecSDave Kleikamp } 4219ac27a0ecSDave Kleikamp 4220dab291afSMingming Cao journal = jbd2_journal_init_inode(journal_inode); 4221ac27a0ecSDave Kleikamp if (!journal) { 4222b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 4223ac27a0ecSDave Kleikamp iput(journal_inode); 4224ac27a0ecSDave Kleikamp return NULL; 4225ac27a0ecSDave Kleikamp } 4226ac27a0ecSDave Kleikamp journal->j_private = sb; 4227617ba13bSMingming Cao ext4_init_journal_params(sb, journal); 4228ac27a0ecSDave Kleikamp return journal; 4229ac27a0ecSDave Kleikamp } 4230ac27a0ecSDave Kleikamp 4231617ba13bSMingming Cao static journal_t *ext4_get_dev_journal(struct super_block *sb, 4232ac27a0ecSDave Kleikamp dev_t j_dev) 4233ac27a0ecSDave Kleikamp { 4234ac27a0ecSDave Kleikamp struct buffer_head *bh; 4235ac27a0ecSDave Kleikamp journal_t *journal; 4236617ba13bSMingming Cao ext4_fsblk_t start; 4237617ba13bSMingming Cao ext4_fsblk_t len; 4238ac27a0ecSDave Kleikamp int hblock, blocksize; 4239617ba13bSMingming Cao ext4_fsblk_t sb_block; 4240ac27a0ecSDave Kleikamp unsigned long offset; 4241617ba13bSMingming Cao struct ext4_super_block *es; 4242ac27a0ecSDave Kleikamp struct block_device *bdev; 4243ac27a0ecSDave Kleikamp 42440390131bSFrank Mayhar BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 42450390131bSFrank Mayhar 4246b31e1552SEric Sandeen bdev = ext4_blkdev_get(j_dev, sb); 4247ac27a0ecSDave Kleikamp if (bdev == NULL) 4248ac27a0ecSDave Kleikamp return NULL; 4249ac27a0ecSDave Kleikamp 4250ac27a0ecSDave Kleikamp blocksize = sb->s_blocksize; 4251e1defc4fSMartin K. Petersen hblock = bdev_logical_block_size(bdev); 4252ac27a0ecSDave Kleikamp if (blocksize < hblock) { 4253b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4254b31e1552SEric Sandeen "blocksize too small for journal device"); 4255ac27a0ecSDave Kleikamp goto out_bdev; 4256ac27a0ecSDave Kleikamp } 4257ac27a0ecSDave Kleikamp 4258617ba13bSMingming Cao sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 4259617ba13bSMingming Cao offset = EXT4_MIN_BLOCK_SIZE % blocksize; 4260ac27a0ecSDave Kleikamp set_blocksize(bdev, blocksize); 4261ac27a0ecSDave Kleikamp if (!(bh = __bread(bdev, sb_block, blocksize))) { 4262b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 4263b31e1552SEric Sandeen "external journal"); 4264ac27a0ecSDave Kleikamp goto out_bdev; 4265ac27a0ecSDave Kleikamp } 4266ac27a0ecSDave Kleikamp 42672716b802STheodore Ts'o es = (struct ext4_super_block *) (bh->b_data + offset); 4268617ba13bSMingming Cao if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 4269ac27a0ecSDave Kleikamp !(le32_to_cpu(es->s_feature_incompat) & 4270617ba13bSMingming Cao EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 4271b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "external journal has " 4272b31e1552SEric Sandeen "bad superblock"); 4273ac27a0ecSDave Kleikamp brelse(bh); 4274ac27a0ecSDave Kleikamp goto out_bdev; 4275ac27a0ecSDave Kleikamp } 4276ac27a0ecSDave Kleikamp 4277617ba13bSMingming Cao if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 4278b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 4279ac27a0ecSDave Kleikamp brelse(bh); 4280ac27a0ecSDave Kleikamp goto out_bdev; 4281ac27a0ecSDave Kleikamp } 4282ac27a0ecSDave Kleikamp 4283bd81d8eeSLaurent Vivier len = ext4_blocks_count(es); 4284ac27a0ecSDave Kleikamp start = sb_block + 1; 4285ac27a0ecSDave Kleikamp brelse(bh); /* we're done with the superblock */ 4286ac27a0ecSDave Kleikamp 4287dab291afSMingming Cao journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 4288ac27a0ecSDave Kleikamp start, len, blocksize); 4289ac27a0ecSDave Kleikamp if (!journal) { 4290b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "failed to create device journal"); 4291ac27a0ecSDave Kleikamp goto out_bdev; 4292ac27a0ecSDave Kleikamp } 4293ac27a0ecSDave Kleikamp journal->j_private = sb; 42949f203507STheodore Ts'o ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer); 4295ac27a0ecSDave Kleikamp wait_on_buffer(journal->j_sb_buffer); 4296ac27a0ecSDave Kleikamp if (!buffer_uptodate(journal->j_sb_buffer)) { 4297b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 4298ac27a0ecSDave Kleikamp goto out_journal; 4299ac27a0ecSDave Kleikamp } 4300ac27a0ecSDave Kleikamp if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 4301b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "External journal has more than one " 4302b31e1552SEric Sandeen "user (unsupported) - %d", 4303ac27a0ecSDave Kleikamp be32_to_cpu(journal->j_superblock->s_nr_users)); 4304ac27a0ecSDave Kleikamp goto out_journal; 4305ac27a0ecSDave Kleikamp } 4306617ba13bSMingming Cao EXT4_SB(sb)->journal_bdev = bdev; 4307617ba13bSMingming Cao ext4_init_journal_params(sb, journal); 4308ac27a0ecSDave Kleikamp return journal; 43090b8e58a1SAndreas Dilger 4310ac27a0ecSDave Kleikamp out_journal: 4311dab291afSMingming Cao jbd2_journal_destroy(journal); 4312ac27a0ecSDave Kleikamp out_bdev: 4313617ba13bSMingming Cao ext4_blkdev_put(bdev); 4314ac27a0ecSDave Kleikamp return NULL; 4315ac27a0ecSDave Kleikamp } 4316ac27a0ecSDave Kleikamp 4317617ba13bSMingming Cao static int ext4_load_journal(struct super_block *sb, 4318617ba13bSMingming Cao struct ext4_super_block *es, 4319ac27a0ecSDave Kleikamp unsigned long journal_devnum) 4320ac27a0ecSDave Kleikamp { 4321ac27a0ecSDave Kleikamp journal_t *journal; 4322ac27a0ecSDave Kleikamp unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 4323ac27a0ecSDave Kleikamp dev_t journal_dev; 4324ac27a0ecSDave Kleikamp int err = 0; 4325ac27a0ecSDave Kleikamp int really_read_only; 4326ac27a0ecSDave Kleikamp 43270390131bSFrank Mayhar BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 43280390131bSFrank Mayhar 4329ac27a0ecSDave Kleikamp if (journal_devnum && 4330ac27a0ecSDave Kleikamp journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4331b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "external journal device major/minor " 4332b31e1552SEric Sandeen "numbers have changed"); 4333ac27a0ecSDave Kleikamp journal_dev = new_decode_dev(journal_devnum); 4334ac27a0ecSDave Kleikamp } else 4335ac27a0ecSDave Kleikamp journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 4336ac27a0ecSDave Kleikamp 4337ac27a0ecSDave Kleikamp really_read_only = bdev_read_only(sb->s_bdev); 4338ac27a0ecSDave Kleikamp 4339ac27a0ecSDave Kleikamp /* 4340ac27a0ecSDave Kleikamp * Are we loading a blank journal or performing recovery after a 4341ac27a0ecSDave Kleikamp * crash? For recovery, we need to check in advance whether we 4342ac27a0ecSDave Kleikamp * can get read-write access to the device. 4343ac27a0ecSDave Kleikamp */ 4344617ba13bSMingming Cao if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) { 4345ac27a0ecSDave Kleikamp if (sb->s_flags & MS_RDONLY) { 4346b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "INFO: recovery " 4347b31e1552SEric Sandeen "required on readonly filesystem"); 4348ac27a0ecSDave Kleikamp if (really_read_only) { 4349b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "write access " 4350b31e1552SEric Sandeen "unavailable, cannot proceed"); 4351ac27a0ecSDave Kleikamp return -EROFS; 4352ac27a0ecSDave Kleikamp } 4353b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "write access will " 4354b31e1552SEric Sandeen "be enabled during recovery"); 4355ac27a0ecSDave Kleikamp } 4356ac27a0ecSDave Kleikamp } 4357ac27a0ecSDave Kleikamp 4358ac27a0ecSDave Kleikamp if (journal_inum && journal_dev) { 4359b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "filesystem has both journal " 4360b31e1552SEric Sandeen "and inode journals!"); 4361ac27a0ecSDave Kleikamp return -EINVAL; 4362ac27a0ecSDave Kleikamp } 4363ac27a0ecSDave Kleikamp 4364ac27a0ecSDave Kleikamp if (journal_inum) { 4365617ba13bSMingming Cao if (!(journal = ext4_get_journal(sb, journal_inum))) 4366ac27a0ecSDave Kleikamp return -EINVAL; 4367ac27a0ecSDave Kleikamp } else { 4368617ba13bSMingming Cao if (!(journal = ext4_get_dev_journal(sb, journal_dev))) 4369ac27a0ecSDave Kleikamp return -EINVAL; 4370ac27a0ecSDave Kleikamp } 4371ac27a0ecSDave Kleikamp 437290576c0bSTheodore Ts'o if (!(journal->j_flags & JBD2_BARRIER)) 4373b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "barriers disabled"); 43744776004fSTheodore Ts'o 4375617ba13bSMingming Cao if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) 4376dab291afSMingming Cao err = jbd2_journal_wipe(journal, !really_read_only); 43771c13d5c0STheodore Ts'o if (!err) { 43781c13d5c0STheodore Ts'o char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 43791c13d5c0STheodore Ts'o if (save) 43801c13d5c0STheodore Ts'o memcpy(save, ((char *) es) + 43811c13d5c0STheodore Ts'o EXT4_S_ERR_START, EXT4_S_ERR_LEN); 4382dab291afSMingming Cao err = jbd2_journal_load(journal); 43831c13d5c0STheodore Ts'o if (save) 43841c13d5c0STheodore Ts'o memcpy(((char *) es) + EXT4_S_ERR_START, 43851c13d5c0STheodore Ts'o save, EXT4_S_ERR_LEN); 43861c13d5c0STheodore Ts'o kfree(save); 43871c13d5c0STheodore Ts'o } 4388ac27a0ecSDave Kleikamp 4389ac27a0ecSDave Kleikamp if (err) { 4390b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "error loading journal"); 4391dab291afSMingming Cao jbd2_journal_destroy(journal); 4392ac27a0ecSDave Kleikamp return err; 4393ac27a0ecSDave Kleikamp } 4394ac27a0ecSDave Kleikamp 4395617ba13bSMingming Cao EXT4_SB(sb)->s_journal = journal; 4396617ba13bSMingming Cao ext4_clear_journal_err(sb, es); 4397ac27a0ecSDave Kleikamp 4398c41303ceSMaciej Żenczykowski if (!really_read_only && journal_devnum && 4399ac27a0ecSDave Kleikamp journal_devnum != le32_to_cpu(es->s_journal_dev)) { 4400ac27a0ecSDave Kleikamp es->s_journal_dev = cpu_to_le32(journal_devnum); 4401ac27a0ecSDave Kleikamp 4402ac27a0ecSDave Kleikamp /* Make sure we flush the recovery flag to disk. */ 4403e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 4404ac27a0ecSDave Kleikamp } 4405ac27a0ecSDave Kleikamp 4406ac27a0ecSDave Kleikamp return 0; 4407ac27a0ecSDave Kleikamp } 4408ac27a0ecSDave Kleikamp 4409e2d67052STheodore Ts'o static int ext4_commit_super(struct super_block *sb, int sync) 4410ac27a0ecSDave Kleikamp { 4411e2d67052STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 4412617ba13bSMingming Cao struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 4413c4be0c1dSTakashi Sato int error = 0; 4414ac27a0ecSDave Kleikamp 44157c2e7087STheodore Ts'o if (!sbh || block_device_ejected(sb)) 4416c4be0c1dSTakashi Sato return error; 4417914258bfSTheodore Ts'o if (buffer_write_io_error(sbh)) { 4418914258bfSTheodore Ts'o /* 4419914258bfSTheodore Ts'o * Oh, dear. A previous attempt to write the 4420914258bfSTheodore Ts'o * superblock failed. This could happen because the 4421914258bfSTheodore Ts'o * USB device was yanked out. Or it could happen to 4422914258bfSTheodore Ts'o * be a transient write error and maybe the block will 4423914258bfSTheodore Ts'o * be remapped. Nothing we can do but to retry the 4424914258bfSTheodore Ts'o * write and hope for the best. 4425914258bfSTheodore Ts'o */ 4426b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "previous I/O error to " 4427b31e1552SEric Sandeen "superblock detected"); 4428914258bfSTheodore Ts'o clear_buffer_write_io_error(sbh); 4429914258bfSTheodore Ts'o set_buffer_uptodate(sbh); 4430914258bfSTheodore Ts'o } 443171290b36STheodore Ts'o /* 443271290b36STheodore Ts'o * If the file system is mounted read-only, don't update the 443371290b36STheodore Ts'o * superblock write time. This avoids updating the superblock 443471290b36STheodore Ts'o * write time when we are mounting the root file system 443571290b36STheodore Ts'o * read/only but we need to replay the journal; at that point, 443671290b36STheodore Ts'o * for people who are east of GMT and who make their clock 443771290b36STheodore Ts'o * tick in localtime for Windows bug-for-bug compatibility, 443871290b36STheodore Ts'o * the clock is set in the future, and this will cause e2fsck 443971290b36STheodore Ts'o * to complain and force a full file system check. 444071290b36STheodore Ts'o */ 444171290b36STheodore Ts'o if (!(sb->s_flags & MS_RDONLY)) 4442ac27a0ecSDave Kleikamp es->s_wtime = cpu_to_le32(get_seconds()); 4443f613dfcbSTheodore Ts'o if (sb->s_bdev->bd_part) 4444afc32f7eSTheodore Ts'o es->s_kbytes_written = 4445afc32f7eSTheodore Ts'o cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 4446afc32f7eSTheodore Ts'o ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 4447afc32f7eSTheodore Ts'o EXT4_SB(sb)->s_sectors_written_start) >> 1)); 4448f613dfcbSTheodore Ts'o else 4449f613dfcbSTheodore Ts'o es->s_kbytes_written = 4450f613dfcbSTheodore Ts'o cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 445157042651STheodore Ts'o ext4_free_blocks_count_set(es, 445257042651STheodore Ts'o EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 445357042651STheodore Ts'o &EXT4_SB(sb)->s_freeclusters_counter))); 44547f93cff9STheodore Ts'o es->s_free_inodes_count = 44557f93cff9STheodore Ts'o cpu_to_le32(percpu_counter_sum_positive( 44565d1b1b3fSAneesh Kumar K.V &EXT4_SB(sb)->s_freeinodes_counter)); 4457ac27a0ecSDave Kleikamp BUFFER_TRACE(sbh, "marking dirty"); 445806db49e6STheodore Ts'o ext4_superblock_csum_set(sb); 4459ac27a0ecSDave Kleikamp mark_buffer_dirty(sbh); 4460914258bfSTheodore Ts'o if (sync) { 4461c4be0c1dSTakashi Sato error = sync_dirty_buffer(sbh); 4462c4be0c1dSTakashi Sato if (error) 4463c4be0c1dSTakashi Sato return error; 4464c4be0c1dSTakashi Sato 4465c4be0c1dSTakashi Sato error = buffer_write_io_error(sbh); 4466c4be0c1dSTakashi Sato if (error) { 4467b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "I/O error while writing " 4468b31e1552SEric Sandeen "superblock"); 4469914258bfSTheodore Ts'o clear_buffer_write_io_error(sbh); 4470914258bfSTheodore Ts'o set_buffer_uptodate(sbh); 4471914258bfSTheodore Ts'o } 4472914258bfSTheodore Ts'o } 4473c4be0c1dSTakashi Sato return error; 4474ac27a0ecSDave Kleikamp } 4475ac27a0ecSDave Kleikamp 4476ac27a0ecSDave Kleikamp /* 4477ac27a0ecSDave Kleikamp * Have we just finished recovery? If so, and if we are mounting (or 4478ac27a0ecSDave Kleikamp * remounting) the filesystem readonly, then we will end up with a 4479ac27a0ecSDave Kleikamp * consistent fs on disk. Record that fact. 4480ac27a0ecSDave Kleikamp */ 4481617ba13bSMingming Cao static void ext4_mark_recovery_complete(struct super_block *sb, 4482617ba13bSMingming Cao struct ext4_super_block *es) 4483ac27a0ecSDave Kleikamp { 4484617ba13bSMingming Cao journal_t *journal = EXT4_SB(sb)->s_journal; 4485ac27a0ecSDave Kleikamp 44860390131bSFrank Mayhar if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) { 44870390131bSFrank Mayhar BUG_ON(journal != NULL); 44880390131bSFrank Mayhar return; 44890390131bSFrank Mayhar } 4490dab291afSMingming Cao jbd2_journal_lock_updates(journal); 44917ffe1ea8SHidehiro Kawai if (jbd2_journal_flush(journal) < 0) 44927ffe1ea8SHidehiro Kawai goto out; 44937ffe1ea8SHidehiro Kawai 4494617ba13bSMingming Cao if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) && 4495ac27a0ecSDave Kleikamp sb->s_flags & MS_RDONLY) { 4496617ba13bSMingming Cao EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4497e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 4498ac27a0ecSDave Kleikamp } 44997ffe1ea8SHidehiro Kawai 45007ffe1ea8SHidehiro Kawai out: 4501dab291afSMingming Cao jbd2_journal_unlock_updates(journal); 4502ac27a0ecSDave Kleikamp } 4503ac27a0ecSDave Kleikamp 4504ac27a0ecSDave Kleikamp /* 4505ac27a0ecSDave Kleikamp * If we are mounting (or read-write remounting) a filesystem whose journal 4506ac27a0ecSDave Kleikamp * has recorded an error from a previous lifetime, move that error to the 4507ac27a0ecSDave Kleikamp * main filesystem now. 4508ac27a0ecSDave Kleikamp */ 4509617ba13bSMingming Cao static void ext4_clear_journal_err(struct super_block *sb, 4510617ba13bSMingming Cao struct ext4_super_block *es) 4511ac27a0ecSDave Kleikamp { 4512ac27a0ecSDave Kleikamp journal_t *journal; 4513ac27a0ecSDave Kleikamp int j_errno; 4514ac27a0ecSDave Kleikamp const char *errstr; 4515ac27a0ecSDave Kleikamp 45160390131bSFrank Mayhar BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)); 45170390131bSFrank Mayhar 4518617ba13bSMingming Cao journal = EXT4_SB(sb)->s_journal; 4519ac27a0ecSDave Kleikamp 4520ac27a0ecSDave Kleikamp /* 4521ac27a0ecSDave Kleikamp * Now check for any error status which may have been recorded in the 4522617ba13bSMingming Cao * journal by a prior ext4_error() or ext4_abort() 4523ac27a0ecSDave Kleikamp */ 4524ac27a0ecSDave Kleikamp 4525dab291afSMingming Cao j_errno = jbd2_journal_errno(journal); 4526ac27a0ecSDave Kleikamp if (j_errno) { 4527ac27a0ecSDave Kleikamp char nbuf[16]; 4528ac27a0ecSDave Kleikamp 4529617ba13bSMingming Cao errstr = ext4_decode_error(sb, j_errno, nbuf); 453012062dddSEric Sandeen ext4_warning(sb, "Filesystem error recorded " 4531ac27a0ecSDave Kleikamp "from previous mount: %s", errstr); 453212062dddSEric Sandeen ext4_warning(sb, "Marking fs in need of filesystem check."); 4533ac27a0ecSDave Kleikamp 4534617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 4535617ba13bSMingming Cao es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 4536e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 4537ac27a0ecSDave Kleikamp 4538dab291afSMingming Cao jbd2_journal_clear_err(journal); 4539d796c52eSTheodore Ts'o jbd2_journal_update_sb_errno(journal); 4540ac27a0ecSDave Kleikamp } 4541ac27a0ecSDave Kleikamp } 4542ac27a0ecSDave Kleikamp 4543ac27a0ecSDave Kleikamp /* 4544ac27a0ecSDave Kleikamp * Force the running and committing transactions to commit, 4545ac27a0ecSDave Kleikamp * and wait on the commit. 4546ac27a0ecSDave Kleikamp */ 4547617ba13bSMingming Cao int ext4_force_commit(struct super_block *sb) 4548ac27a0ecSDave Kleikamp { 4549ac27a0ecSDave Kleikamp journal_t *journal; 4550ac27a0ecSDave Kleikamp 4551ac27a0ecSDave Kleikamp if (sb->s_flags & MS_RDONLY) 4552ac27a0ecSDave Kleikamp return 0; 4553ac27a0ecSDave Kleikamp 4554617ba13bSMingming Cao journal = EXT4_SB(sb)->s_journal; 4555b1deefc9SGuo Chao return ext4_journal_force_commit(journal); 4556ac27a0ecSDave Kleikamp } 4557ac27a0ecSDave Kleikamp 4558617ba13bSMingming Cao static int ext4_sync_fs(struct super_block *sb, int wait) 4559ac27a0ecSDave Kleikamp { 456014ce0cb4STheodore Ts'o int ret = 0; 45619eddacf9SJan Kara tid_t target; 456206a407f1SDmitry Monakhov bool needs_barrier = false; 45638d5d02e6SMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 4564ac27a0ecSDave Kleikamp 45659bffad1eSTheodore Ts'o trace_ext4_sync_fs(sb, wait); 45662e8fa54eSJan Kara flush_workqueue(sbi->rsv_conversion_wq); 45672e8fa54eSJan Kara flush_workqueue(sbi->unrsv_conversion_wq); 4568a1177825SJan Kara /* 4569a1177825SJan Kara * Writeback quota in non-journalled quota case - journalled quota has 4570a1177825SJan Kara * no dirty dquots 4571a1177825SJan Kara */ 4572a1177825SJan Kara dquot_writeback_dquots(sb, -1); 457306a407f1SDmitry Monakhov /* 457406a407f1SDmitry Monakhov * Data writeback is possible w/o journal transaction, so barrier must 457506a407f1SDmitry Monakhov * being sent at the end of the function. But we can skip it if 457606a407f1SDmitry Monakhov * transaction_commit will do it for us. 457706a407f1SDmitry Monakhov */ 457806a407f1SDmitry Monakhov target = jbd2_get_latest_transaction(sbi->s_journal); 457906a407f1SDmitry Monakhov if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 458006a407f1SDmitry Monakhov !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 458106a407f1SDmitry Monakhov needs_barrier = true; 458206a407f1SDmitry Monakhov 45838d5d02e6SMingming Cao if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 4584ac27a0ecSDave Kleikamp if (wait) 458506a407f1SDmitry Monakhov ret = jbd2_log_wait_commit(sbi->s_journal, target); 45860390131bSFrank Mayhar } 458706a407f1SDmitry Monakhov if (needs_barrier) { 458806a407f1SDmitry Monakhov int err; 458906a407f1SDmitry Monakhov err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); 459006a407f1SDmitry Monakhov if (!ret) 459106a407f1SDmitry Monakhov ret = err; 459206a407f1SDmitry Monakhov } 459306a407f1SDmitry Monakhov 459406a407f1SDmitry Monakhov return ret; 459506a407f1SDmitry Monakhov } 459606a407f1SDmitry Monakhov 459706a407f1SDmitry Monakhov static int ext4_sync_fs_nojournal(struct super_block *sb, int wait) 459806a407f1SDmitry Monakhov { 459906a407f1SDmitry Monakhov int ret = 0; 460006a407f1SDmitry Monakhov 460106a407f1SDmitry Monakhov trace_ext4_sync_fs(sb, wait); 460206a407f1SDmitry Monakhov flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 460306a407f1SDmitry Monakhov flush_workqueue(EXT4_SB(sb)->unrsv_conversion_wq); 460406a407f1SDmitry Monakhov dquot_writeback_dquots(sb, -1); 460506a407f1SDmitry Monakhov if (wait && test_opt(sb, BARRIER)) 460606a407f1SDmitry Monakhov ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); 460706a407f1SDmitry Monakhov 460814ce0cb4STheodore Ts'o return ret; 4609ac27a0ecSDave Kleikamp } 4610ac27a0ecSDave Kleikamp 4611ac27a0ecSDave Kleikamp /* 4612ac27a0ecSDave Kleikamp * LVM calls this function before a (read-only) snapshot is created. This 4613ac27a0ecSDave Kleikamp * gives us a chance to flush the journal completely and mark the fs clean. 4614be4f27d3SYongqiang Yang * 4615be4f27d3SYongqiang Yang * Note that only this function cannot bring a filesystem to be in a clean 46168e8ad8a5SJan Kara * state independently. It relies on upper layer to stop all data & metadata 46178e8ad8a5SJan Kara * modifications. 4618ac27a0ecSDave Kleikamp */ 4619c4be0c1dSTakashi Sato static int ext4_freeze(struct super_block *sb) 4620ac27a0ecSDave Kleikamp { 4621c4be0c1dSTakashi Sato int error = 0; 4622c4be0c1dSTakashi Sato journal_t *journal; 4623ac27a0ecSDave Kleikamp 46249ca92389STheodore Ts'o if (sb->s_flags & MS_RDONLY) 46259ca92389STheodore Ts'o return 0; 46269ca92389STheodore Ts'o 4627c4be0c1dSTakashi Sato journal = EXT4_SB(sb)->s_journal; 4628ac27a0ecSDave Kleikamp 4629ac27a0ecSDave Kleikamp /* Now we set up the journal barrier. */ 4630dab291afSMingming Cao jbd2_journal_lock_updates(journal); 46317ffe1ea8SHidehiro Kawai 46327ffe1ea8SHidehiro Kawai /* 46339ca92389STheodore Ts'o * Don't clear the needs_recovery flag if we failed to flush 46349ca92389STheodore Ts'o * the journal. 46357ffe1ea8SHidehiro Kawai */ 4636c4be0c1dSTakashi Sato error = jbd2_journal_flush(journal); 46376b0310fbSEric Sandeen if (error < 0) 46386b0310fbSEric Sandeen goto out; 4639ac27a0ecSDave Kleikamp 4640ac27a0ecSDave Kleikamp /* Journal blocked and flushed, clear needs_recovery flag. */ 4641617ba13bSMingming Cao EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4642e2d67052STheodore Ts'o error = ext4_commit_super(sb, 1); 46436b0310fbSEric Sandeen out: 46448e8ad8a5SJan Kara /* we rely on upper layer to stop further updates */ 46456b0310fbSEric Sandeen jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 46466b0310fbSEric Sandeen return error; 4647ac27a0ecSDave Kleikamp } 4648ac27a0ecSDave Kleikamp 4649ac27a0ecSDave Kleikamp /* 4650ac27a0ecSDave Kleikamp * Called by LVM after the snapshot is done. We need to reset the RECOVER 4651ac27a0ecSDave Kleikamp * flag here, even though the filesystem is not technically dirty yet. 4652ac27a0ecSDave Kleikamp */ 4653c4be0c1dSTakashi Sato static int ext4_unfreeze(struct super_block *sb) 4654ac27a0ecSDave Kleikamp { 46559ca92389STheodore Ts'o if (sb->s_flags & MS_RDONLY) 46569ca92389STheodore Ts'o return 0; 46579ca92389STheodore Ts'o 46589ca92389STheodore Ts'o /* Reset the needs_recovery flag before the fs is unlocked. */ 4659617ba13bSMingming Cao EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 4660e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 4661c4be0c1dSTakashi Sato return 0; 4662ac27a0ecSDave Kleikamp } 4663ac27a0ecSDave Kleikamp 4664673c6100STheodore Ts'o /* 4665673c6100STheodore Ts'o * Structure to save mount options for ext4_remount's benefit 4666673c6100STheodore Ts'o */ 4667673c6100STheodore Ts'o struct ext4_mount_options { 4668673c6100STheodore Ts'o unsigned long s_mount_opt; 4669a2595b8aSTheodore Ts'o unsigned long s_mount_opt2; 467008cefc7aSEric W. Biederman kuid_t s_resuid; 467108cefc7aSEric W. Biederman kgid_t s_resgid; 4672673c6100STheodore Ts'o unsigned long s_commit_interval; 4673673c6100STheodore Ts'o u32 s_min_batch_time, s_max_batch_time; 4674673c6100STheodore Ts'o #ifdef CONFIG_QUOTA 4675673c6100STheodore Ts'o int s_jquota_fmt; 4676673c6100STheodore Ts'o char *s_qf_names[MAXQUOTAS]; 4677673c6100STheodore Ts'o #endif 4678673c6100STheodore Ts'o }; 4679673c6100STheodore Ts'o 4680617ba13bSMingming Cao static int ext4_remount(struct super_block *sb, int *flags, char *data) 4681ac27a0ecSDave Kleikamp { 4682617ba13bSMingming Cao struct ext4_super_block *es; 4683617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 4684ac27a0ecSDave Kleikamp unsigned long old_sb_flags; 4685617ba13bSMingming Cao struct ext4_mount_options old_opts; 4686c79d967dSChristoph Hellwig int enable_quota = 0; 46878a266467STheodore Ts'o ext4_group_t g; 4688b3881f74STheodore Ts'o unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 4689c5e06d10SJohann Lombardi int err = 0; 4690ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 469103dafb5fSChen Gang int i, j; 4692ac27a0ecSDave Kleikamp #endif 4693d4c402d9SCurt Wohlgemuth char *orig_data = kstrdup(data, GFP_KERNEL); 4694ac27a0ecSDave Kleikamp 4695ac27a0ecSDave Kleikamp /* Store the original options */ 4696ac27a0ecSDave Kleikamp old_sb_flags = sb->s_flags; 4697ac27a0ecSDave Kleikamp old_opts.s_mount_opt = sbi->s_mount_opt; 4698a2595b8aSTheodore Ts'o old_opts.s_mount_opt2 = sbi->s_mount_opt2; 4699ac27a0ecSDave Kleikamp old_opts.s_resuid = sbi->s_resuid; 4700ac27a0ecSDave Kleikamp old_opts.s_resgid = sbi->s_resgid; 4701ac27a0ecSDave Kleikamp old_opts.s_commit_interval = sbi->s_commit_interval; 470230773840STheodore Ts'o old_opts.s_min_batch_time = sbi->s_min_batch_time; 470330773840STheodore Ts'o old_opts.s_max_batch_time = sbi->s_max_batch_time; 4704ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4705ac27a0ecSDave Kleikamp old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 4706ac27a0ecSDave Kleikamp for (i = 0; i < MAXQUOTAS; i++) 470703dafb5fSChen Gang if (sbi->s_qf_names[i]) { 470803dafb5fSChen Gang old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i], 470903dafb5fSChen Gang GFP_KERNEL); 471003dafb5fSChen Gang if (!old_opts.s_qf_names[i]) { 471103dafb5fSChen Gang for (j = 0; j < i; j++) 471203dafb5fSChen Gang kfree(old_opts.s_qf_names[j]); 47133e36a163SWei Yongjun kfree(orig_data); 471403dafb5fSChen Gang return -ENOMEM; 471503dafb5fSChen Gang } 471603dafb5fSChen Gang } else 471703dafb5fSChen Gang old_opts.s_qf_names[i] = NULL; 4718ac27a0ecSDave Kleikamp #endif 4719b3881f74STheodore Ts'o if (sbi->s_journal && sbi->s_journal->j_task->io_context) 4720b3881f74STheodore Ts'o journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 4721ac27a0ecSDave Kleikamp 4722ac27a0ecSDave Kleikamp /* 4723ac27a0ecSDave Kleikamp * Allow the "check" option to be passed as a remount option. 4724ac27a0ecSDave Kleikamp */ 4725661aa520SEric Sandeen if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 4726ac27a0ecSDave Kleikamp err = -EINVAL; 4727ac27a0ecSDave Kleikamp goto restore_opts; 4728ac27a0ecSDave Kleikamp } 4729ac27a0ecSDave Kleikamp 47304ab2f15bSTheodore Ts'o if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 4731c67d859eSTheodore Ts'o ext4_abort(sb, "Abort forced by user"); 4732ac27a0ecSDave Kleikamp 4733ac27a0ecSDave Kleikamp sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 4734482a7425SDmitry Monakhov (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 4735ac27a0ecSDave Kleikamp 4736ac27a0ecSDave Kleikamp es = sbi->s_es; 4737ac27a0ecSDave Kleikamp 4738b3881f74STheodore Ts'o if (sbi->s_journal) { 4739617ba13bSMingming Cao ext4_init_journal_params(sb, sbi->s_journal); 4740b3881f74STheodore Ts'o set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4741b3881f74STheodore Ts'o } 4742ac27a0ecSDave Kleikamp 4743661aa520SEric Sandeen if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) { 47444ab2f15bSTheodore Ts'o if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 4745ac27a0ecSDave Kleikamp err = -EROFS; 4746ac27a0ecSDave Kleikamp goto restore_opts; 4747ac27a0ecSDave Kleikamp } 4748ac27a0ecSDave Kleikamp 4749ac27a0ecSDave Kleikamp if (*flags & MS_RDONLY) { 47500f0dd62fSChristoph Hellwig err = dquot_suspend(sb, -1); 47510f0dd62fSChristoph Hellwig if (err < 0) 4752c79d967dSChristoph Hellwig goto restore_opts; 4753c79d967dSChristoph Hellwig 4754ac27a0ecSDave Kleikamp /* 4755ac27a0ecSDave Kleikamp * First of all, the unconditional stuff we have to do 4756ac27a0ecSDave Kleikamp * to disable replay of the journal when we next remount 4757ac27a0ecSDave Kleikamp */ 4758ac27a0ecSDave Kleikamp sb->s_flags |= MS_RDONLY; 4759ac27a0ecSDave Kleikamp 4760ac27a0ecSDave Kleikamp /* 4761ac27a0ecSDave Kleikamp * OK, test if we are remounting a valid rw partition 4762ac27a0ecSDave Kleikamp * readonly, and if so set the rdonly flag and then 4763ac27a0ecSDave Kleikamp * mark the partition as valid again. 4764ac27a0ecSDave Kleikamp */ 4765617ba13bSMingming Cao if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 4766617ba13bSMingming Cao (sbi->s_mount_state & EXT4_VALID_FS)) 4767ac27a0ecSDave Kleikamp es->s_state = cpu_to_le16(sbi->s_mount_state); 4768ac27a0ecSDave Kleikamp 4769a63c9eb2STheodore Ts'o if (sbi->s_journal) 4770617ba13bSMingming Cao ext4_mark_recovery_complete(sb, es); 4771ac27a0ecSDave Kleikamp } else { 4772a13fb1a4SEric Sandeen /* Make sure we can mount this feature set readwrite */ 4773a13fb1a4SEric Sandeen if (!ext4_feature_set_ok(sb, 0)) { 4774ac27a0ecSDave Kleikamp err = -EROFS; 4775ac27a0ecSDave Kleikamp goto restore_opts; 4776ac27a0ecSDave Kleikamp } 4777ead6596bSEric Sandeen /* 47788a266467STheodore Ts'o * Make sure the group descriptor checksums 47790b8e58a1SAndreas Dilger * are sane. If they aren't, refuse to remount r/w. 47808a266467STheodore Ts'o */ 47818a266467STheodore Ts'o for (g = 0; g < sbi->s_groups_count; g++) { 47828a266467STheodore Ts'o struct ext4_group_desc *gdp = 47838a266467STheodore Ts'o ext4_get_group_desc(sb, g, NULL); 47848a266467STheodore Ts'o 4785feb0ab32SDarrick J. Wong if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 4786b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4787b31e1552SEric Sandeen "ext4_remount: Checksum for group %u failed (%u!=%u)", 47888a266467STheodore Ts'o g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)), 47898a266467STheodore Ts'o le16_to_cpu(gdp->bg_checksum)); 47908a266467STheodore Ts'o err = -EINVAL; 47918a266467STheodore Ts'o goto restore_opts; 47928a266467STheodore Ts'o } 47938a266467STheodore Ts'o } 47948a266467STheodore Ts'o 47958a266467STheodore Ts'o /* 4796ead6596bSEric Sandeen * If we have an unprocessed orphan list hanging 4797ead6596bSEric Sandeen * around from a previously readonly bdev mount, 4798ead6596bSEric Sandeen * require a full umount/remount for now. 4799ead6596bSEric Sandeen */ 4800ead6596bSEric Sandeen if (es->s_last_orphan) { 4801b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "Couldn't " 4802ead6596bSEric Sandeen "remount RDWR because of unprocessed " 4803ead6596bSEric Sandeen "orphan inode list. Please " 4804b31e1552SEric Sandeen "umount/remount instead"); 4805ead6596bSEric Sandeen err = -EINVAL; 4806ead6596bSEric Sandeen goto restore_opts; 4807ead6596bSEric Sandeen } 4808ead6596bSEric Sandeen 4809ac27a0ecSDave Kleikamp /* 4810ac27a0ecSDave Kleikamp * Mounting a RDONLY partition read-write, so reread 4811ac27a0ecSDave Kleikamp * and store the current valid flag. (It may have 4812ac27a0ecSDave Kleikamp * been changed by e2fsck since we originally mounted 4813ac27a0ecSDave Kleikamp * the partition.) 4814ac27a0ecSDave Kleikamp */ 48150390131bSFrank Mayhar if (sbi->s_journal) 4816617ba13bSMingming Cao ext4_clear_journal_err(sb, es); 4817ac27a0ecSDave Kleikamp sbi->s_mount_state = le16_to_cpu(es->s_state); 4818617ba13bSMingming Cao if (!ext4_setup_super(sb, es, 0)) 4819ac27a0ecSDave Kleikamp sb->s_flags &= ~MS_RDONLY; 4820c5e06d10SJohann Lombardi if (EXT4_HAS_INCOMPAT_FEATURE(sb, 4821c5e06d10SJohann Lombardi EXT4_FEATURE_INCOMPAT_MMP)) 4822c5e06d10SJohann Lombardi if (ext4_multi_mount_protect(sb, 4823c5e06d10SJohann Lombardi le64_to_cpu(es->s_mmp_block))) { 4824c5e06d10SJohann Lombardi err = -EROFS; 4825c5e06d10SJohann Lombardi goto restore_opts; 4826c5e06d10SJohann Lombardi } 4827c79d967dSChristoph Hellwig enable_quota = 1; 4828ac27a0ecSDave Kleikamp } 4829ac27a0ecSDave Kleikamp } 4830bfff6873SLukas Czerner 4831bfff6873SLukas Czerner /* 4832bfff6873SLukas Czerner * Reinitialize lazy itable initialization thread based on 4833bfff6873SLukas Czerner * current settings 4834bfff6873SLukas Czerner */ 4835bfff6873SLukas Czerner if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE)) 4836bfff6873SLukas Czerner ext4_unregister_li_request(sb); 4837bfff6873SLukas Czerner else { 4838bfff6873SLukas Czerner ext4_group_t first_not_zeroed; 4839bfff6873SLukas Czerner first_not_zeroed = ext4_has_uninit_itable(sb); 4840bfff6873SLukas Czerner ext4_register_li_request(sb, first_not_zeroed); 4841bfff6873SLukas Czerner } 4842bfff6873SLukas Czerner 48436fd058f7STheodore Ts'o ext4_setup_system_zone(sb); 4844d096ad0fSMichael Tokarev if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY)) 4845e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 48460390131bSFrank Mayhar 4847ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4848ac27a0ecSDave Kleikamp /* Release old quota file names */ 4849ac27a0ecSDave Kleikamp for (i = 0; i < MAXQUOTAS; i++) 4850ac27a0ecSDave Kleikamp kfree(old_opts.s_qf_names[i]); 48517c319d32SAditya Kali if (enable_quota) { 48527c319d32SAditya Kali if (sb_any_quota_suspended(sb)) 48530f0dd62fSChristoph Hellwig dquot_resume(sb, -1); 48547c319d32SAditya Kali else if (EXT4_HAS_RO_COMPAT_FEATURE(sb, 48557c319d32SAditya Kali EXT4_FEATURE_RO_COMPAT_QUOTA)) { 48567c319d32SAditya Kali err = ext4_enable_quotas(sb); 485707724f98STheodore Ts'o if (err) 48587c319d32SAditya Kali goto restore_opts; 48597c319d32SAditya Kali } 48607c319d32SAditya Kali } 48617c319d32SAditya Kali #endif 4862d4c402d9SCurt Wohlgemuth 4863d4c402d9SCurt Wohlgemuth ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 4864d4c402d9SCurt Wohlgemuth kfree(orig_data); 4865ac27a0ecSDave Kleikamp return 0; 48660b8e58a1SAndreas Dilger 4867ac27a0ecSDave Kleikamp restore_opts: 4868ac27a0ecSDave Kleikamp sb->s_flags = old_sb_flags; 4869ac27a0ecSDave Kleikamp sbi->s_mount_opt = old_opts.s_mount_opt; 4870a2595b8aSTheodore Ts'o sbi->s_mount_opt2 = old_opts.s_mount_opt2; 4871ac27a0ecSDave Kleikamp sbi->s_resuid = old_opts.s_resuid; 4872ac27a0ecSDave Kleikamp sbi->s_resgid = old_opts.s_resgid; 4873ac27a0ecSDave Kleikamp sbi->s_commit_interval = old_opts.s_commit_interval; 487430773840STheodore Ts'o sbi->s_min_batch_time = old_opts.s_min_batch_time; 487530773840STheodore Ts'o sbi->s_max_batch_time = old_opts.s_max_batch_time; 4876ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4877ac27a0ecSDave Kleikamp sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 4878ac27a0ecSDave Kleikamp for (i = 0; i < MAXQUOTAS; i++) { 4879ac27a0ecSDave Kleikamp kfree(sbi->s_qf_names[i]); 4880ac27a0ecSDave Kleikamp sbi->s_qf_names[i] = old_opts.s_qf_names[i]; 4881ac27a0ecSDave Kleikamp } 4882ac27a0ecSDave Kleikamp #endif 4883d4c402d9SCurt Wohlgemuth kfree(orig_data); 4884ac27a0ecSDave Kleikamp return err; 4885ac27a0ecSDave Kleikamp } 4886ac27a0ecSDave Kleikamp 4887617ba13bSMingming Cao static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 4888ac27a0ecSDave Kleikamp { 4889ac27a0ecSDave Kleikamp struct super_block *sb = dentry->d_sb; 4890617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 4891617ba13bSMingming Cao struct ext4_super_block *es = sbi->s_es; 489227dd4385SLukas Czerner ext4_fsblk_t overhead = 0, resv_blocks; 4893960cc398SPekka Enberg u64 fsid; 4894d02a9391SKazuya Mio s64 bfree; 489527dd4385SLukas Czerner resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 4896ac27a0ecSDave Kleikamp 4897952fc18eSTheodore Ts'o if (!test_opt(sb, MINIX_DF)) 4898952fc18eSTheodore Ts'o overhead = sbi->s_overhead; 4899ac27a0ecSDave Kleikamp 4900617ba13bSMingming Cao buf->f_type = EXT4_SUPER_MAGIC; 4901ac27a0ecSDave Kleikamp buf->f_bsize = sb->s_blocksize; 4902b72f78cbSEric Sandeen buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 490357042651STheodore Ts'o bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 490457042651STheodore Ts'o percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 4905d02a9391SKazuya Mio /* prevent underflow in case that few free space is available */ 490657042651STheodore Ts'o buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 490727dd4385SLukas Czerner buf->f_bavail = buf->f_bfree - 490827dd4385SLukas Czerner (ext4_r_blocks_count(es) + resv_blocks); 490927dd4385SLukas Czerner if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 4910ac27a0ecSDave Kleikamp buf->f_bavail = 0; 4911ac27a0ecSDave Kleikamp buf->f_files = le32_to_cpu(es->s_inodes_count); 491252d9f3b4SPeter Zijlstra buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 4913617ba13bSMingming Cao buf->f_namelen = EXT4_NAME_LEN; 4914960cc398SPekka Enberg fsid = le64_to_cpup((void *)es->s_uuid) ^ 4915960cc398SPekka Enberg le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 4916960cc398SPekka Enberg buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 4917960cc398SPekka Enberg buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 49180b8e58a1SAndreas Dilger 4919ac27a0ecSDave Kleikamp return 0; 4920ac27a0ecSDave Kleikamp } 4921ac27a0ecSDave Kleikamp 49220b8e58a1SAndreas Dilger /* Helper function for writing quotas on sync - we need to start transaction 49230b8e58a1SAndreas Dilger * before quota file is locked for write. Otherwise the are possible deadlocks: 4924ac27a0ecSDave Kleikamp * Process 1 Process 2 4925617ba13bSMingming Cao * ext4_create() quota_sync() 4926dab291afSMingming Cao * jbd2_journal_start() write_dquot() 4927871a2931SChristoph Hellwig * dquot_initialize() down(dqio_mutex) 4928dab291afSMingming Cao * down(dqio_mutex) jbd2_journal_start() 4929ac27a0ecSDave Kleikamp * 4930ac27a0ecSDave Kleikamp */ 4931ac27a0ecSDave Kleikamp 4932ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4933ac27a0ecSDave Kleikamp 4934ac27a0ecSDave Kleikamp static inline struct inode *dquot_to_inode(struct dquot *dquot) 4935ac27a0ecSDave Kleikamp { 49364c376dcaSEric W. Biederman return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 4937ac27a0ecSDave Kleikamp } 4938ac27a0ecSDave Kleikamp 4939617ba13bSMingming Cao static int ext4_write_dquot(struct dquot *dquot) 4940ac27a0ecSDave Kleikamp { 4941ac27a0ecSDave Kleikamp int ret, err; 4942ac27a0ecSDave Kleikamp handle_t *handle; 4943ac27a0ecSDave Kleikamp struct inode *inode; 4944ac27a0ecSDave Kleikamp 4945ac27a0ecSDave Kleikamp inode = dquot_to_inode(dquot); 49469924a92aSTheodore Ts'o handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 4947617ba13bSMingming Cao EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 4948ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 4949ac27a0ecSDave Kleikamp return PTR_ERR(handle); 4950ac27a0ecSDave Kleikamp ret = dquot_commit(dquot); 4951617ba13bSMingming Cao err = ext4_journal_stop(handle); 4952ac27a0ecSDave Kleikamp if (!ret) 4953ac27a0ecSDave Kleikamp ret = err; 4954ac27a0ecSDave Kleikamp return ret; 4955ac27a0ecSDave Kleikamp } 4956ac27a0ecSDave Kleikamp 4957617ba13bSMingming Cao static int ext4_acquire_dquot(struct dquot *dquot) 4958ac27a0ecSDave Kleikamp { 4959ac27a0ecSDave Kleikamp int ret, err; 4960ac27a0ecSDave Kleikamp handle_t *handle; 4961ac27a0ecSDave Kleikamp 49629924a92aSTheodore Ts'o handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 4963617ba13bSMingming Cao EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 4964ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 4965ac27a0ecSDave Kleikamp return PTR_ERR(handle); 4966ac27a0ecSDave Kleikamp ret = dquot_acquire(dquot); 4967617ba13bSMingming Cao err = ext4_journal_stop(handle); 4968ac27a0ecSDave Kleikamp if (!ret) 4969ac27a0ecSDave Kleikamp ret = err; 4970ac27a0ecSDave Kleikamp return ret; 4971ac27a0ecSDave Kleikamp } 4972ac27a0ecSDave Kleikamp 4973617ba13bSMingming Cao static int ext4_release_dquot(struct dquot *dquot) 4974ac27a0ecSDave Kleikamp { 4975ac27a0ecSDave Kleikamp int ret, err; 4976ac27a0ecSDave Kleikamp handle_t *handle; 4977ac27a0ecSDave Kleikamp 49789924a92aSTheodore Ts'o handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 4979617ba13bSMingming Cao EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 49809c3013e9SJan Kara if (IS_ERR(handle)) { 49819c3013e9SJan Kara /* Release dquot anyway to avoid endless cycle in dqput() */ 49829c3013e9SJan Kara dquot_release(dquot); 4983ac27a0ecSDave Kleikamp return PTR_ERR(handle); 49849c3013e9SJan Kara } 4985ac27a0ecSDave Kleikamp ret = dquot_release(dquot); 4986617ba13bSMingming Cao err = ext4_journal_stop(handle); 4987ac27a0ecSDave Kleikamp if (!ret) 4988ac27a0ecSDave Kleikamp ret = err; 4989ac27a0ecSDave Kleikamp return ret; 4990ac27a0ecSDave Kleikamp } 4991ac27a0ecSDave Kleikamp 4992617ba13bSMingming Cao static int ext4_mark_dquot_dirty(struct dquot *dquot) 4993ac27a0ecSDave Kleikamp { 4994262b4662SJan Kara struct super_block *sb = dquot->dq_sb; 4995262b4662SJan Kara struct ext4_sb_info *sbi = EXT4_SB(sb); 4996262b4662SJan Kara 49972c8be6b2SJan Kara /* Are we journaling quotas? */ 4998262b4662SJan Kara if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) || 4999262b4662SJan Kara sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 5000ac27a0ecSDave Kleikamp dquot_mark_dquot_dirty(dquot); 5001617ba13bSMingming Cao return ext4_write_dquot(dquot); 5002ac27a0ecSDave Kleikamp } else { 5003ac27a0ecSDave Kleikamp return dquot_mark_dquot_dirty(dquot); 5004ac27a0ecSDave Kleikamp } 5005ac27a0ecSDave Kleikamp } 5006ac27a0ecSDave Kleikamp 5007617ba13bSMingming Cao static int ext4_write_info(struct super_block *sb, int type) 5008ac27a0ecSDave Kleikamp { 5009ac27a0ecSDave Kleikamp int ret, err; 5010ac27a0ecSDave Kleikamp handle_t *handle; 5011ac27a0ecSDave Kleikamp 5012ac27a0ecSDave Kleikamp /* Data block + inode block */ 50139924a92aSTheodore Ts'o handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2); 5014ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 5015ac27a0ecSDave Kleikamp return PTR_ERR(handle); 5016ac27a0ecSDave Kleikamp ret = dquot_commit_info(sb, type); 5017617ba13bSMingming Cao err = ext4_journal_stop(handle); 5018ac27a0ecSDave Kleikamp if (!ret) 5019ac27a0ecSDave Kleikamp ret = err; 5020ac27a0ecSDave Kleikamp return ret; 5021ac27a0ecSDave Kleikamp } 5022ac27a0ecSDave Kleikamp 5023ac27a0ecSDave Kleikamp /* 5024ac27a0ecSDave Kleikamp * Turn on quotas during mount time - we need to find 5025ac27a0ecSDave Kleikamp * the quota file and such... 5026ac27a0ecSDave Kleikamp */ 5027617ba13bSMingming Cao static int ext4_quota_on_mount(struct super_block *sb, int type) 5028ac27a0ecSDave Kleikamp { 5029287a8095SChristoph Hellwig return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type], 5030617ba13bSMingming Cao EXT4_SB(sb)->s_jquota_fmt, type); 5031ac27a0ecSDave Kleikamp } 5032ac27a0ecSDave Kleikamp 5033ac27a0ecSDave Kleikamp /* 5034ac27a0ecSDave Kleikamp * Standard function to be called on quota_on 5035ac27a0ecSDave Kleikamp */ 5036617ba13bSMingming Cao static int ext4_quota_on(struct super_block *sb, int type, int format_id, 5037f00c9e44SJan Kara struct path *path) 5038ac27a0ecSDave Kleikamp { 5039ac27a0ecSDave Kleikamp int err; 5040ac27a0ecSDave Kleikamp 5041ac27a0ecSDave Kleikamp if (!test_opt(sb, QUOTA)) 5042ac27a0ecSDave Kleikamp return -EINVAL; 50430623543bSJan Kara 5044ac27a0ecSDave Kleikamp /* Quotafile not on the same filesystem? */ 5045d8c9584eSAl Viro if (path->dentry->d_sb != sb) 5046ac27a0ecSDave Kleikamp return -EXDEV; 50470623543bSJan Kara /* Journaling quota? */ 50480623543bSJan Kara if (EXT4_SB(sb)->s_qf_names[type]) { 50492b2d6d01STheodore Ts'o /* Quotafile not in fs root? */ 5050f00c9e44SJan Kara if (path->dentry->d_parent != sb->s_root) 5051b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 5052b31e1552SEric Sandeen "Quota file not on filesystem root. " 5053b31e1552SEric Sandeen "Journaled quota will not work"); 50540623543bSJan Kara } 50550623543bSJan Kara 50560623543bSJan Kara /* 50570623543bSJan Kara * When we journal data on quota file, we have to flush journal to see 50580623543bSJan Kara * all updates to the file when we bypass pagecache... 50590623543bSJan Kara */ 50600390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal && 5061f00c9e44SJan Kara ext4_should_journal_data(path->dentry->d_inode)) { 50620623543bSJan Kara /* 50630623543bSJan Kara * We don't need to lock updates but journal_flush() could 50640623543bSJan Kara * otherwise be livelocked... 50650623543bSJan Kara */ 50660623543bSJan Kara jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 50677ffe1ea8SHidehiro Kawai err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 50680623543bSJan Kara jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 5069f00c9e44SJan Kara if (err) 50707ffe1ea8SHidehiro Kawai return err; 50717ffe1ea8SHidehiro Kawai } 50720623543bSJan Kara 5073f00c9e44SJan Kara return dquot_quota_on(sb, type, format_id, path); 5074ac27a0ecSDave Kleikamp } 5075ac27a0ecSDave Kleikamp 50767c319d32SAditya Kali static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 50777c319d32SAditya Kali unsigned int flags) 50787c319d32SAditya Kali { 50797c319d32SAditya Kali int err; 50807c319d32SAditya Kali struct inode *qf_inode; 50817c319d32SAditya Kali unsigned long qf_inums[MAXQUOTAS] = { 50827c319d32SAditya Kali le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 50837c319d32SAditya Kali le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum) 50847c319d32SAditya Kali }; 50857c319d32SAditya Kali 50867c319d32SAditya Kali BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)); 50877c319d32SAditya Kali 50887c319d32SAditya Kali if (!qf_inums[type]) 50897c319d32SAditya Kali return -EPERM; 50907c319d32SAditya Kali 50917c319d32SAditya Kali qf_inode = ext4_iget(sb, qf_inums[type]); 50927c319d32SAditya Kali if (IS_ERR(qf_inode)) { 50937c319d32SAditya Kali ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 50947c319d32SAditya Kali return PTR_ERR(qf_inode); 50957c319d32SAditya Kali } 50967c319d32SAditya Kali 5097bcb13850SJan Kara /* Don't account quota for quota files to avoid recursion */ 5098bcb13850SJan Kara qf_inode->i_flags |= S_NOQUOTA; 50997c319d32SAditya Kali err = dquot_enable(qf_inode, type, format_id, flags); 51007c319d32SAditya Kali iput(qf_inode); 51017c319d32SAditya Kali 51027c319d32SAditya Kali return err; 51037c319d32SAditya Kali } 51047c319d32SAditya Kali 51057c319d32SAditya Kali /* Enable usage tracking for all quota types. */ 51067c319d32SAditya Kali static int ext4_enable_quotas(struct super_block *sb) 51077c319d32SAditya Kali { 51087c319d32SAditya Kali int type, err = 0; 51097c319d32SAditya Kali unsigned long qf_inums[MAXQUOTAS] = { 51107c319d32SAditya Kali le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 51117c319d32SAditya Kali le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum) 51127c319d32SAditya Kali }; 51137c319d32SAditya Kali 51147c319d32SAditya Kali sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 51157c319d32SAditya Kali for (type = 0; type < MAXQUOTAS; type++) { 51167c319d32SAditya Kali if (qf_inums[type]) { 51177c319d32SAditya Kali err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 51187c319d32SAditya Kali DQUOT_USAGE_ENABLED); 51197c319d32SAditya Kali if (err) { 51207c319d32SAditya Kali ext4_warning(sb, 512172ba7450STheodore Ts'o "Failed to enable quota tracking " 512272ba7450STheodore Ts'o "(type=%d, err=%d). Please run " 512372ba7450STheodore Ts'o "e2fsck to fix.", type, err); 51247c319d32SAditya Kali return err; 51257c319d32SAditya Kali } 51267c319d32SAditya Kali } 51277c319d32SAditya Kali } 51287c319d32SAditya Kali return 0; 51297c319d32SAditya Kali } 51307c319d32SAditya Kali 51317c319d32SAditya Kali /* 51327c319d32SAditya Kali * quota_on function that is used when QUOTA feature is set. 51337c319d32SAditya Kali */ 51347c319d32SAditya Kali static int ext4_quota_on_sysfile(struct super_block *sb, int type, 51357c319d32SAditya Kali int format_id) 51367c319d32SAditya Kali { 51377c319d32SAditya Kali if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) 51387c319d32SAditya Kali return -EINVAL; 51397c319d32SAditya Kali 51407c319d32SAditya Kali /* 51417c319d32SAditya Kali * USAGE was enabled at mount time. Only need to enable LIMITS now. 51427c319d32SAditya Kali */ 51437c319d32SAditya Kali return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED); 51447c319d32SAditya Kali } 51457c319d32SAditya Kali 5146ca0e05e4SDmitry Monakhov static int ext4_quota_off(struct super_block *sb, int type) 5147ca0e05e4SDmitry Monakhov { 514821f97697SJan Kara struct inode *inode = sb_dqopt(sb)->files[type]; 514921f97697SJan Kara handle_t *handle; 515021f97697SJan Kara 515187009d86SDmitry Monakhov /* Force all delayed allocation blocks to be allocated. 515287009d86SDmitry Monakhov * Caller already holds s_umount sem */ 515387009d86SDmitry Monakhov if (test_opt(sb, DELALLOC)) 5154ca0e05e4SDmitry Monakhov sync_filesystem(sb); 5155ca0e05e4SDmitry Monakhov 51560b268590SAmir Goldstein if (!inode) 51570b268590SAmir Goldstein goto out; 51580b268590SAmir Goldstein 515921f97697SJan Kara /* Update modification times of quota files when userspace can 516021f97697SJan Kara * start looking at them */ 51619924a92aSTheodore Ts'o handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 516221f97697SJan Kara if (IS_ERR(handle)) 516321f97697SJan Kara goto out; 516421f97697SJan Kara inode->i_mtime = inode->i_ctime = CURRENT_TIME; 516521f97697SJan Kara ext4_mark_inode_dirty(handle, inode); 516621f97697SJan Kara ext4_journal_stop(handle); 516721f97697SJan Kara 516821f97697SJan Kara out: 5169ca0e05e4SDmitry Monakhov return dquot_quota_off(sb, type); 5170ca0e05e4SDmitry Monakhov } 5171ca0e05e4SDmitry Monakhov 51727c319d32SAditya Kali /* 51737c319d32SAditya Kali * quota_off function that is used when QUOTA feature is set. 51747c319d32SAditya Kali */ 51757c319d32SAditya Kali static int ext4_quota_off_sysfile(struct super_block *sb, int type) 51767c319d32SAditya Kali { 51777c319d32SAditya Kali if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) 51787c319d32SAditya Kali return -EINVAL; 51797c319d32SAditya Kali 51807c319d32SAditya Kali /* Disable only the limits. */ 51817c319d32SAditya Kali return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED); 51827c319d32SAditya Kali } 51837c319d32SAditya Kali 5184ac27a0ecSDave Kleikamp /* Read data from quotafile - avoid pagecache and such because we cannot afford 5185ac27a0ecSDave Kleikamp * acquiring the locks... As quota files are never truncated and quota code 5186ac27a0ecSDave Kleikamp * itself serializes the operations (and no one else should touch the files) 5187ac27a0ecSDave Kleikamp * we don't have to be afraid of races */ 5188617ba13bSMingming Cao static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 5189ac27a0ecSDave Kleikamp size_t len, loff_t off) 5190ac27a0ecSDave Kleikamp { 5191ac27a0ecSDave Kleikamp struct inode *inode = sb_dqopt(sb)->files[type]; 5192725d26d3SAneesh Kumar K.V ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5193ac27a0ecSDave Kleikamp int err = 0; 5194ac27a0ecSDave Kleikamp int offset = off & (sb->s_blocksize - 1); 5195ac27a0ecSDave Kleikamp int tocopy; 5196ac27a0ecSDave Kleikamp size_t toread; 5197ac27a0ecSDave Kleikamp struct buffer_head *bh; 5198ac27a0ecSDave Kleikamp loff_t i_size = i_size_read(inode); 5199ac27a0ecSDave Kleikamp 5200ac27a0ecSDave Kleikamp if (off > i_size) 5201ac27a0ecSDave Kleikamp return 0; 5202ac27a0ecSDave Kleikamp if (off+len > i_size) 5203ac27a0ecSDave Kleikamp len = i_size-off; 5204ac27a0ecSDave Kleikamp toread = len; 5205ac27a0ecSDave Kleikamp while (toread > 0) { 5206ac27a0ecSDave Kleikamp tocopy = sb->s_blocksize - offset < toread ? 5207ac27a0ecSDave Kleikamp sb->s_blocksize - offset : toread; 5208617ba13bSMingming Cao bh = ext4_bread(NULL, inode, blk, 0, &err); 5209ac27a0ecSDave Kleikamp if (err) 5210ac27a0ecSDave Kleikamp return err; 5211ac27a0ecSDave Kleikamp if (!bh) /* A hole? */ 5212ac27a0ecSDave Kleikamp memset(data, 0, tocopy); 5213ac27a0ecSDave Kleikamp else 5214ac27a0ecSDave Kleikamp memcpy(data, bh->b_data+offset, tocopy); 5215ac27a0ecSDave Kleikamp brelse(bh); 5216ac27a0ecSDave Kleikamp offset = 0; 5217ac27a0ecSDave Kleikamp toread -= tocopy; 5218ac27a0ecSDave Kleikamp data += tocopy; 5219ac27a0ecSDave Kleikamp blk++; 5220ac27a0ecSDave Kleikamp } 5221ac27a0ecSDave Kleikamp return len; 5222ac27a0ecSDave Kleikamp } 5223ac27a0ecSDave Kleikamp 5224ac27a0ecSDave Kleikamp /* Write to quotafile (we know the transaction is already started and has 5225ac27a0ecSDave Kleikamp * enough credits) */ 5226617ba13bSMingming Cao static ssize_t ext4_quota_write(struct super_block *sb, int type, 5227ac27a0ecSDave Kleikamp const char *data, size_t len, loff_t off) 5228ac27a0ecSDave Kleikamp { 5229ac27a0ecSDave Kleikamp struct inode *inode = sb_dqopt(sb)->files[type]; 5230725d26d3SAneesh Kumar K.V ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 5231ac27a0ecSDave Kleikamp int err = 0; 5232ac27a0ecSDave Kleikamp int offset = off & (sb->s_blocksize - 1); 5233ac27a0ecSDave Kleikamp struct buffer_head *bh; 5234ac27a0ecSDave Kleikamp handle_t *handle = journal_current_handle(); 5235ac27a0ecSDave Kleikamp 52360390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal && !handle) { 5237b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 5238b31e1552SEric Sandeen " cancelled because transaction is not started", 52399c3013e9SJan Kara (unsigned long long)off, (unsigned long long)len); 52409c3013e9SJan Kara return -EIO; 52419c3013e9SJan Kara } 524267eeb568SDmitry Monakhov /* 524367eeb568SDmitry Monakhov * Since we account only one data block in transaction credits, 524467eeb568SDmitry Monakhov * then it is impossible to cross a block boundary. 524567eeb568SDmitry Monakhov */ 524667eeb568SDmitry Monakhov if (sb->s_blocksize - offset < len) { 524767eeb568SDmitry Monakhov ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 524867eeb568SDmitry Monakhov " cancelled because not block aligned", 524967eeb568SDmitry Monakhov (unsigned long long)off, (unsigned long long)len); 525067eeb568SDmitry Monakhov return -EIO; 525167eeb568SDmitry Monakhov } 525267eeb568SDmitry Monakhov 5253617ba13bSMingming Cao bh = ext4_bread(handle, inode, blk, 1, &err); 5254ac27a0ecSDave Kleikamp if (!bh) 5255ac27a0ecSDave Kleikamp goto out; 5256617ba13bSMingming Cao err = ext4_journal_get_write_access(handle, bh); 5257ac27a0ecSDave Kleikamp if (err) { 5258ac27a0ecSDave Kleikamp brelse(bh); 5259ac27a0ecSDave Kleikamp goto out; 5260ac27a0ecSDave Kleikamp } 5261ac27a0ecSDave Kleikamp lock_buffer(bh); 526267eeb568SDmitry Monakhov memcpy(bh->b_data+offset, data, len); 5263ac27a0ecSDave Kleikamp flush_dcache_page(bh->b_page); 5264ac27a0ecSDave Kleikamp unlock_buffer(bh); 52650390131bSFrank Mayhar err = ext4_handle_dirty_metadata(handle, NULL, bh); 5266ac27a0ecSDave Kleikamp brelse(bh); 5267ac27a0ecSDave Kleikamp out: 52680b7f7cefSJan Kara if (err) 5269ac27a0ecSDave Kleikamp return err; 527067eeb568SDmitry Monakhov if (inode->i_size < off + len) { 527167eeb568SDmitry Monakhov i_size_write(inode, off + len); 5272617ba13bSMingming Cao EXT4_I(inode)->i_disksize = inode->i_size; 5273617ba13bSMingming Cao ext4_mark_inode_dirty(handle, inode); 527421f97697SJan Kara } 527567eeb568SDmitry Monakhov return len; 5276ac27a0ecSDave Kleikamp } 5277ac27a0ecSDave Kleikamp 5278ac27a0ecSDave Kleikamp #endif 5279ac27a0ecSDave Kleikamp 5280152a0836SAl Viro static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 5281152a0836SAl Viro const char *dev_name, void *data) 5282ac27a0ecSDave Kleikamp { 5283152a0836SAl Viro return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 5284ac27a0ecSDave Kleikamp } 5285ac27a0ecSDave Kleikamp 528637f328ebSTheodore Ts'o #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 528724b58424STheodore Ts'o static inline void register_as_ext2(void) 528824b58424STheodore Ts'o { 528924b58424STheodore Ts'o int err = register_filesystem(&ext2_fs_type); 529024b58424STheodore Ts'o if (err) 529124b58424STheodore Ts'o printk(KERN_WARNING 529224b58424STheodore Ts'o "EXT4-fs: Unable to register as ext2 (%d)\n", err); 529324b58424STheodore Ts'o } 529424b58424STheodore Ts'o 529524b58424STheodore Ts'o static inline void unregister_as_ext2(void) 529624b58424STheodore Ts'o { 529724b58424STheodore Ts'o unregister_filesystem(&ext2_fs_type); 529824b58424STheodore Ts'o } 52992035e776STheodore Ts'o 53002035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb) 53012035e776STheodore Ts'o { 53022035e776STheodore Ts'o if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP)) 53032035e776STheodore Ts'o return 0; 53042035e776STheodore Ts'o if (sb->s_flags & MS_RDONLY) 53052035e776STheodore Ts'o return 1; 53062035e776STheodore Ts'o if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP)) 53072035e776STheodore Ts'o return 0; 53082035e776STheodore Ts'o return 1; 53092035e776STheodore Ts'o } 531024b58424STheodore Ts'o #else 531124b58424STheodore Ts'o static inline void register_as_ext2(void) { } 531224b58424STheodore Ts'o static inline void unregister_as_ext2(void) { } 53132035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 531424b58424STheodore Ts'o #endif 531524b58424STheodore Ts'o 531637f328ebSTheodore Ts'o #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 531724b58424STheodore Ts'o static inline void register_as_ext3(void) 531824b58424STheodore Ts'o { 531924b58424STheodore Ts'o int err = register_filesystem(&ext3_fs_type); 532024b58424STheodore Ts'o if (err) 532124b58424STheodore Ts'o printk(KERN_WARNING 532224b58424STheodore Ts'o "EXT4-fs: Unable to register as ext3 (%d)\n", err); 532324b58424STheodore Ts'o } 532424b58424STheodore Ts'o 532524b58424STheodore Ts'o static inline void unregister_as_ext3(void) 532624b58424STheodore Ts'o { 532724b58424STheodore Ts'o unregister_filesystem(&ext3_fs_type); 532824b58424STheodore Ts'o } 53292035e776STheodore Ts'o 53302035e776STheodore Ts'o static inline int ext3_feature_set_ok(struct super_block *sb) 53312035e776STheodore Ts'o { 53322035e776STheodore Ts'o if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP)) 53332035e776STheodore Ts'o return 0; 53342035e776STheodore Ts'o if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) 53352035e776STheodore Ts'o return 0; 53362035e776STheodore Ts'o if (sb->s_flags & MS_RDONLY) 53372035e776STheodore Ts'o return 1; 53382035e776STheodore Ts'o if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) 53392035e776STheodore Ts'o return 0; 53402035e776STheodore Ts'o return 1; 53412035e776STheodore Ts'o } 534224b58424STheodore Ts'o #else 534324b58424STheodore Ts'o static inline void register_as_ext3(void) { } 534424b58424STheodore Ts'o static inline void unregister_as_ext3(void) { } 53452035e776STheodore Ts'o static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; } 534624b58424STheodore Ts'o #endif 534724b58424STheodore Ts'o 534803010a33STheodore Ts'o static struct file_system_type ext4_fs_type = { 5349ac27a0ecSDave Kleikamp .owner = THIS_MODULE, 535003010a33STheodore Ts'o .name = "ext4", 5351152a0836SAl Viro .mount = ext4_mount, 5352ac27a0ecSDave Kleikamp .kill_sb = kill_block_super, 5353ac27a0ecSDave Kleikamp .fs_flags = FS_REQUIRES_DEV, 5354ac27a0ecSDave Kleikamp }; 53557f78e035SEric W. Biederman MODULE_ALIAS_FS("ext4"); 5356ac27a0ecSDave Kleikamp 53578f021222SLukas Czerner static int __init ext4_init_feat_adverts(void) 5358857ac889SLukas Czerner { 5359857ac889SLukas Czerner struct ext4_features *ef; 5360857ac889SLukas Czerner int ret = -ENOMEM; 5361857ac889SLukas Czerner 5362857ac889SLukas Czerner ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL); 5363857ac889SLukas Czerner if (!ef) 5364857ac889SLukas Czerner goto out; 5365857ac889SLukas Czerner 5366857ac889SLukas Czerner ef->f_kobj.kset = ext4_kset; 5367857ac889SLukas Czerner init_completion(&ef->f_kobj_unregister); 5368857ac889SLukas Czerner ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL, 5369857ac889SLukas Czerner "features"); 5370857ac889SLukas Czerner if (ret) { 5371857ac889SLukas Czerner kfree(ef); 5372857ac889SLukas Czerner goto out; 5373857ac889SLukas Czerner } 5374857ac889SLukas Czerner 5375857ac889SLukas Czerner ext4_feat = ef; 5376857ac889SLukas Czerner ret = 0; 5377857ac889SLukas Czerner out: 5378857ac889SLukas Czerner return ret; 5379857ac889SLukas Czerner } 5380857ac889SLukas Czerner 53818f021222SLukas Czerner static void ext4_exit_feat_adverts(void) 53828f021222SLukas Czerner { 53838f021222SLukas Czerner kobject_put(&ext4_feat->f_kobj); 53848f021222SLukas Czerner wait_for_completion(&ext4_feat->f_kobj_unregister); 53858f021222SLukas Czerner kfree(ext4_feat); 53868f021222SLukas Czerner } 53878f021222SLukas Czerner 5388e9e3bcecSEric Sandeen /* Shared across all ext4 file systems */ 5389e9e3bcecSEric Sandeen wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 5390e9e3bcecSEric Sandeen struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ]; 5391e9e3bcecSEric Sandeen 53925dabfc78STheodore Ts'o static int __init ext4_init_fs(void) 5393ac27a0ecSDave Kleikamp { 5394e9e3bcecSEric Sandeen int i, err; 5395c9de560dSAlex Tomas 539607c0c5d8SAl Viro ext4_li_info = NULL; 539707c0c5d8SAl Viro mutex_init(&ext4_li_mtx); 539807c0c5d8SAl Viro 53999a4c8019SCarlos Maiolino /* Build-time check for flags consistency */ 540012e9b892SDmitry Monakhov ext4_check_flag_values(); 5401e9e3bcecSEric Sandeen 5402e9e3bcecSEric Sandeen for (i = 0; i < EXT4_WQ_HASH_SZ; i++) { 5403e9e3bcecSEric Sandeen mutex_init(&ext4__aio_mutex[i]); 5404e9e3bcecSEric Sandeen init_waitqueue_head(&ext4__ioend_wq[i]); 5405e9e3bcecSEric Sandeen } 5406e9e3bcecSEric Sandeen 540751865fdaSZheng Liu err = ext4_init_es(); 54086fd058f7STheodore Ts'o if (err) 54096fd058f7STheodore Ts'o return err; 541051865fdaSZheng Liu 541151865fdaSZheng Liu err = ext4_init_pageio(); 541251865fdaSZheng Liu if (err) 541351865fdaSZheng Liu goto out7; 541451865fdaSZheng Liu 54155dabfc78STheodore Ts'o err = ext4_init_system_zone(); 5416bd2d0210STheodore Ts'o if (err) 5417d44651d0SFabrice Jouhaud goto out6; 54183197ebdbSTheodore Ts'o ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj); 54190e376b1eSTheodore Ts'o if (!ext4_kset) { 54200e376b1eSTheodore Ts'o err = -ENOMEM; 5421dd68314cSTheodore Ts'o goto out5; 54220e376b1eSTheodore Ts'o } 5423d44651d0SFabrice Jouhaud ext4_proc_root = proc_mkdir("fs/ext4", NULL); 5424857ac889SLukas Czerner 5425857ac889SLukas Czerner err = ext4_init_feat_adverts(); 5426dd68314cSTheodore Ts'o if (err) 5427dd68314cSTheodore Ts'o goto out4; 5428857ac889SLukas Czerner 54295dabfc78STheodore Ts'o err = ext4_init_mballoc(); 5430ac27a0ecSDave Kleikamp if (err) 54316fd058f7STheodore Ts'o goto out3; 5432c9de560dSAlex Tomas 54335dabfc78STheodore Ts'o err = ext4_init_xattr(); 5434c9de560dSAlex Tomas if (err) 5435c9de560dSAlex Tomas goto out2; 5436ac27a0ecSDave Kleikamp err = init_inodecache(); 5437ac27a0ecSDave Kleikamp if (err) 5438ac27a0ecSDave Kleikamp goto out1; 543924b58424STheodore Ts'o register_as_ext3(); 54402035e776STheodore Ts'o register_as_ext2(); 544103010a33STheodore Ts'o err = register_filesystem(&ext4_fs_type); 5442ac27a0ecSDave Kleikamp if (err) 5443ac27a0ecSDave Kleikamp goto out; 5444bfff6873SLukas Czerner 5445ac27a0ecSDave Kleikamp return 0; 5446ac27a0ecSDave Kleikamp out: 544724b58424STheodore Ts'o unregister_as_ext2(); 544824b58424STheodore Ts'o unregister_as_ext3(); 5449ac27a0ecSDave Kleikamp destroy_inodecache(); 5450ac27a0ecSDave Kleikamp out1: 54515dabfc78STheodore Ts'o ext4_exit_xattr(); 5452c9de560dSAlex Tomas out2: 54535dabfc78STheodore Ts'o ext4_exit_mballoc(); 54546fd058f7STheodore Ts'o out3: 54558f021222SLukas Czerner ext4_exit_feat_adverts(); 54566fd058f7STheodore Ts'o out4: 5457d44651d0SFabrice Jouhaud if (ext4_proc_root) 5458dd68314cSTheodore Ts'o remove_proc_entry("fs/ext4", NULL); 5459dd68314cSTheodore Ts'o kset_unregister(ext4_kset); 5460d44651d0SFabrice Jouhaud out5: 5461dd68314cSTheodore Ts'o ext4_exit_system_zone(); 5462d44651d0SFabrice Jouhaud out6: 54635dabfc78STheodore Ts'o ext4_exit_pageio(); 546451865fdaSZheng Liu out7: 546551865fdaSZheng Liu ext4_exit_es(); 546651865fdaSZheng Liu 5467ac27a0ecSDave Kleikamp return err; 5468ac27a0ecSDave Kleikamp } 5469ac27a0ecSDave Kleikamp 54705dabfc78STheodore Ts'o static void __exit ext4_exit_fs(void) 5471ac27a0ecSDave Kleikamp { 5472bfff6873SLukas Czerner ext4_destroy_lazyinit_thread(); 547324b58424STheodore Ts'o unregister_as_ext2(); 547424b58424STheodore Ts'o unregister_as_ext3(); 547503010a33STheodore Ts'o unregister_filesystem(&ext4_fs_type); 5476ac27a0ecSDave Kleikamp destroy_inodecache(); 54775dabfc78STheodore Ts'o ext4_exit_xattr(); 54785dabfc78STheodore Ts'o ext4_exit_mballoc(); 54798f021222SLukas Czerner ext4_exit_feat_adverts(); 54809f6200bbSTheodore Ts'o remove_proc_entry("fs/ext4", NULL); 54813197ebdbSTheodore Ts'o kset_unregister(ext4_kset); 54825dabfc78STheodore Ts'o ext4_exit_system_zone(); 54835dabfc78STheodore Ts'o ext4_exit_pageio(); 5484dd12ed14SEric Sandeen ext4_exit_es(); 5485ac27a0ecSDave Kleikamp } 5486ac27a0ecSDave Kleikamp 5487ac27a0ecSDave Kleikamp MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 548883982b6fSTheodore Ts'o MODULE_DESCRIPTION("Fourth Extended Filesystem"); 5489ac27a0ecSDave Kleikamp MODULE_LICENSE("GPL"); 54905dabfc78STheodore Ts'o module_init(ext4_init_fs) 54915dabfc78STheodore Ts'o module_exit(ext4_exit_fs) 5492