1f5166768STheodore Ts'o // SPDX-License-Identifier: GPL-2.0 2ac27a0ecSDave Kleikamp /* 3617ba13bSMingming Cao * linux/fs/ext4/super.c 4ac27a0ecSDave Kleikamp * 5ac27a0ecSDave Kleikamp * Copyright (C) 1992, 1993, 1994, 1995 6ac27a0ecSDave Kleikamp * Remy Card (card@masi.ibp.fr) 7ac27a0ecSDave Kleikamp * Laboratoire MASI - Institut Blaise Pascal 8ac27a0ecSDave Kleikamp * Universite Pierre et Marie Curie (Paris VI) 9ac27a0ecSDave Kleikamp * 10ac27a0ecSDave Kleikamp * from 11ac27a0ecSDave Kleikamp * 12ac27a0ecSDave Kleikamp * linux/fs/minix/inode.c 13ac27a0ecSDave Kleikamp * 14ac27a0ecSDave Kleikamp * Copyright (C) 1991, 1992 Linus Torvalds 15ac27a0ecSDave Kleikamp * 16ac27a0ecSDave Kleikamp * Big-endian to little-endian byte-swapping/bitmaps by 17ac27a0ecSDave Kleikamp * David S. Miller (davem@caip.rutgers.edu), 1995 18ac27a0ecSDave Kleikamp */ 19ac27a0ecSDave Kleikamp 20ac27a0ecSDave Kleikamp #include <linux/module.h> 21ac27a0ecSDave Kleikamp #include <linux/string.h> 22ac27a0ecSDave Kleikamp #include <linux/fs.h> 23ac27a0ecSDave Kleikamp #include <linux/time.h> 24c5ca7c76STheodore Ts'o #include <linux/vmalloc.h> 25ac27a0ecSDave Kleikamp #include <linux/slab.h> 26ac27a0ecSDave Kleikamp #include <linux/init.h> 27ac27a0ecSDave Kleikamp #include <linux/blkdev.h> 2866114cadSTejun Heo #include <linux/backing-dev.h> 29ac27a0ecSDave Kleikamp #include <linux/parser.h> 30ac27a0ecSDave Kleikamp #include <linux/buffer_head.h> 31a5694255SChristoph Hellwig #include <linux/exportfs.h> 32ac27a0ecSDave Kleikamp #include <linux/vfs.h> 33ac27a0ecSDave Kleikamp #include <linux/random.h> 34ac27a0ecSDave Kleikamp #include <linux/mount.h> 35ac27a0ecSDave Kleikamp #include <linux/namei.h> 36ac27a0ecSDave Kleikamp #include <linux/quotaops.h> 37ac27a0ecSDave Kleikamp #include <linux/seq_file.h> 383197ebdbSTheodore Ts'o #include <linux/ctype.h> 391330593eSVignesh Babu #include <linux/log2.h> 40717d50e4SAndreas Dilger #include <linux/crc16.h> 41ef510424SDan Williams #include <linux/dax.h> 427abc52c2SDan Magenheimer #include <linux/cleancache.h> 437c0f6ba6SLinus Torvalds #include <linux/uaccess.h> 44ee73f9a5SJeff Layton #include <linux/iversion.h> 45c83ad55eSGabriel Krisman Bertazi #include <linux/unicode.h> 46c6a564ffSChristoph Hellwig #include <linux/part_stat.h> 47bfff6873SLukas Czerner #include <linux/kthread.h> 48bfff6873SLukas Czerner #include <linux/freezer.h> 49bfff6873SLukas Czerner 503dcf5451SChristoph Hellwig #include "ext4.h" 514a092d73STheodore Ts'o #include "ext4_extents.h" /* Needed for trace points definition */ 523dcf5451SChristoph Hellwig #include "ext4_jbd2.h" 53ac27a0ecSDave Kleikamp #include "xattr.h" 54ac27a0ecSDave Kleikamp #include "acl.h" 553661d286STheodore Ts'o #include "mballoc.h" 560c9ec4beSDarrick J. Wong #include "fsmap.h" 57ac27a0ecSDave Kleikamp 589bffad1eSTheodore Ts'o #define CREATE_TRACE_POINTS 599bffad1eSTheodore Ts'o #include <trace/events/ext4.h> 609bffad1eSTheodore Ts'o 610b75a840SLukas Czerner static struct ext4_lazy_init *ext4_li_info; 6259ebc7fdSZheng Yongjun static DEFINE_MUTEX(ext4_li_mtx); 63e294a537STheodore Ts'o static struct ratelimit_state ext4_mount_msg_ratelimit; 649f6200bbSTheodore Ts'o 65617ba13bSMingming Cao static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 66ac27a0ecSDave Kleikamp unsigned long journal_devnum); 672adf6da8STheodore Ts'o static int ext4_show_options(struct seq_file *seq, struct dentry *root); 682d01ddc8SJan Kara static void ext4_update_super(struct super_block *sb); 694392fbc4SJan Kara static int ext4_commit_super(struct super_block *sb); 7011215630SJan Kara static int ext4_mark_recovery_complete(struct super_block *sb, 71617ba13bSMingming Cao struct ext4_super_block *es); 7211215630SJan Kara static int ext4_clear_journal_err(struct super_block *sb, 73617ba13bSMingming Cao struct ext4_super_block *es); 74617ba13bSMingming Cao static int ext4_sync_fs(struct super_block *sb, int wait); 75617ba13bSMingming Cao static int ext4_remount(struct super_block *sb, int *flags, char *data); 76617ba13bSMingming Cao static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 77c4be0c1dSTakashi Sato static int ext4_unfreeze(struct super_block *sb); 78c4be0c1dSTakashi Sato static int ext4_freeze(struct super_block *sb); 79152a0836SAl Viro static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 80152a0836SAl Viro const char *dev_name, void *data); 812035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb); 822035e776STheodore Ts'o static inline int ext3_feature_set_ok(struct super_block *sb); 83d39195c3SAmir Goldstein static int ext4_feature_set_ok(struct super_block *sb, int readonly); 84bfff6873SLukas Czerner static void ext4_destroy_lazyinit_thread(void); 85bfff6873SLukas Czerner static void ext4_unregister_li_request(struct super_block *sb); 868f1f7453SEric Sandeen static void ext4_clear_request_list(void); 87c6cb7e77SEric Whitney static struct inode *ext4_get_journal_inode(struct super_block *sb, 88c6cb7e77SEric Whitney unsigned int journal_inum); 89ac27a0ecSDave Kleikamp 90e74031fdSJan Kara /* 91e74031fdSJan Kara * Lock ordering 92e74031fdSJan Kara * 93e74031fdSJan Kara * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and 94e74031fdSJan Kara * i_mmap_rwsem (inode->i_mmap_rwsem)! 95e74031fdSJan Kara * 96e74031fdSJan Kara * page fault path: 97c1e8d7c6SMichel Lespinasse * mmap_lock -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start -> 98e74031fdSJan Kara * page lock -> i_data_sem (rw) 99e74031fdSJan Kara * 100e74031fdSJan Kara * buffered write path: 101c1e8d7c6SMichel Lespinasse * sb_start_write -> i_mutex -> mmap_lock 102e74031fdSJan Kara * sb_start_write -> i_mutex -> transaction start -> page lock -> 103e74031fdSJan Kara * i_data_sem (rw) 104e74031fdSJan Kara * 105e74031fdSJan Kara * truncate: 1061d39834fSNikolay Borisov * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock 1071d39834fSNikolay Borisov * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start -> 1081d39834fSNikolay Borisov * i_data_sem (rw) 109e74031fdSJan Kara * 110e74031fdSJan Kara * direct IO: 111c1e8d7c6SMichel Lespinasse * sb_start_write -> i_mutex -> mmap_lock 1121d39834fSNikolay Borisov * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw) 113e74031fdSJan Kara * 114e74031fdSJan Kara * writepages: 115e74031fdSJan Kara * transaction start -> page lock(s) -> i_data_sem (rw) 116e74031fdSJan Kara */ 117e74031fdSJan Kara 118c290ea01SJan Kara #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 1192035e776STheodore Ts'o static struct file_system_type ext2_fs_type = { 1202035e776STheodore Ts'o .owner = THIS_MODULE, 1212035e776STheodore Ts'o .name = "ext2", 1222035e776STheodore Ts'o .mount = ext4_mount, 1232035e776STheodore Ts'o .kill_sb = kill_block_super, 1242035e776STheodore Ts'o .fs_flags = FS_REQUIRES_DEV, 1252035e776STheodore Ts'o }; 1267f78e035SEric W. Biederman MODULE_ALIAS_FS("ext2"); 127fa7614ddSEric W. Biederman MODULE_ALIAS("ext2"); 1282035e776STheodore Ts'o #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 1292035e776STheodore Ts'o #else 1302035e776STheodore Ts'o #define IS_EXT2_SB(sb) (0) 1312035e776STheodore Ts'o #endif 1322035e776STheodore Ts'o 1332035e776STheodore Ts'o 134ba69f9abSJan Kara static struct file_system_type ext3_fs_type = { 135ba69f9abSJan Kara .owner = THIS_MODULE, 136ba69f9abSJan Kara .name = "ext3", 137152a0836SAl Viro .mount = ext4_mount, 138ba69f9abSJan Kara .kill_sb = kill_block_super, 139ba69f9abSJan Kara .fs_flags = FS_REQUIRES_DEV, 140ba69f9abSJan Kara }; 1417f78e035SEric W. Biederman MODULE_ALIAS_FS("ext3"); 142fa7614ddSEric W. Biederman MODULE_ALIAS("ext3"); 143ba69f9abSJan Kara #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 144bd81d8eeSLaurent Vivier 145fa491b14Szhangyi (F) 146fa491b14Szhangyi (F) static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags, 147fa491b14Szhangyi (F) bh_end_io_t *end_io) 148fa491b14Szhangyi (F) { 149fb265c9cSTheodore Ts'o /* 150fa491b14Szhangyi (F) * buffer's verified bit is no longer valid after reading from 151fa491b14Szhangyi (F) * disk again due to write out error, clear it to make sure we 152fa491b14Szhangyi (F) * recheck the buffer contents. 153fa491b14Szhangyi (F) */ 154fa491b14Szhangyi (F) clear_buffer_verified(bh); 155fa491b14Szhangyi (F) 156fa491b14Szhangyi (F) bh->b_end_io = end_io ? end_io : end_buffer_read_sync; 157fa491b14Szhangyi (F) get_bh(bh); 158fa491b14Szhangyi (F) submit_bh(REQ_OP_READ, op_flags, bh); 159fa491b14Szhangyi (F) } 160fa491b14Szhangyi (F) 161fa491b14Szhangyi (F) void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags, 162fa491b14Szhangyi (F) bh_end_io_t *end_io) 163fa491b14Szhangyi (F) { 164fa491b14Szhangyi (F) BUG_ON(!buffer_locked(bh)); 165fa491b14Szhangyi (F) 166fa491b14Szhangyi (F) if (ext4_buffer_uptodate(bh)) { 167fa491b14Szhangyi (F) unlock_buffer(bh); 168fa491b14Szhangyi (F) return; 169fa491b14Szhangyi (F) } 170fa491b14Szhangyi (F) __ext4_read_bh(bh, op_flags, end_io); 171fa491b14Szhangyi (F) } 172fa491b14Szhangyi (F) 173fa491b14Szhangyi (F) int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io) 174fa491b14Szhangyi (F) { 175fa491b14Szhangyi (F) BUG_ON(!buffer_locked(bh)); 176fa491b14Szhangyi (F) 177fa491b14Szhangyi (F) if (ext4_buffer_uptodate(bh)) { 178fa491b14Szhangyi (F) unlock_buffer(bh); 179fa491b14Szhangyi (F) return 0; 180fa491b14Szhangyi (F) } 181fa491b14Szhangyi (F) 182fa491b14Szhangyi (F) __ext4_read_bh(bh, op_flags, end_io); 183fa491b14Szhangyi (F) 184fa491b14Szhangyi (F) wait_on_buffer(bh); 185fa491b14Szhangyi (F) if (buffer_uptodate(bh)) 186fa491b14Szhangyi (F) return 0; 187fa491b14Szhangyi (F) return -EIO; 188fa491b14Szhangyi (F) } 189fa491b14Szhangyi (F) 190fa491b14Szhangyi (F) int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait) 191fa491b14Szhangyi (F) { 192fa491b14Szhangyi (F) if (trylock_buffer(bh)) { 193fa491b14Szhangyi (F) if (wait) 194fa491b14Szhangyi (F) return ext4_read_bh(bh, op_flags, NULL); 195fa491b14Szhangyi (F) ext4_read_bh_nowait(bh, op_flags, NULL); 196fa491b14Szhangyi (F) return 0; 197fa491b14Szhangyi (F) } 198fa491b14Szhangyi (F) if (wait) { 199fa491b14Szhangyi (F) wait_on_buffer(bh); 200fa491b14Szhangyi (F) if (buffer_uptodate(bh)) 201fa491b14Szhangyi (F) return 0; 202fa491b14Szhangyi (F) return -EIO; 203fa491b14Szhangyi (F) } 204fa491b14Szhangyi (F) return 0; 205fa491b14Szhangyi (F) } 206fa491b14Szhangyi (F) 207fb265c9cSTheodore Ts'o /* 2088394a6abSzhangyi (F) * This works like __bread_gfp() except it uses ERR_PTR for error 209fb265c9cSTheodore Ts'o * returns. Currently with sb_bread it's impossible to distinguish 210fb265c9cSTheodore Ts'o * between ENOMEM and EIO situations (since both result in a NULL 211fb265c9cSTheodore Ts'o * return. 212fb265c9cSTheodore Ts'o */ 2138394a6abSzhangyi (F) static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb, 2148394a6abSzhangyi (F) sector_t block, int op_flags, 2158394a6abSzhangyi (F) gfp_t gfp) 216fb265c9cSTheodore Ts'o { 2172d069c08Szhangyi (F) struct buffer_head *bh; 2182d069c08Szhangyi (F) int ret; 219fb265c9cSTheodore Ts'o 2208394a6abSzhangyi (F) bh = sb_getblk_gfp(sb, block, gfp); 221fb265c9cSTheodore Ts'o if (bh == NULL) 222fb265c9cSTheodore Ts'o return ERR_PTR(-ENOMEM); 223cf2834a5STheodore Ts'o if (ext4_buffer_uptodate(bh)) 224fb265c9cSTheodore Ts'o return bh; 2252d069c08Szhangyi (F) 2262d069c08Szhangyi (F) ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true); 2272d069c08Szhangyi (F) if (ret) { 228fb265c9cSTheodore Ts'o put_bh(bh); 2292d069c08Szhangyi (F) return ERR_PTR(ret); 2302d069c08Szhangyi (F) } 2312d069c08Szhangyi (F) return bh; 232fb265c9cSTheodore Ts'o } 233fb265c9cSTheodore Ts'o 2348394a6abSzhangyi (F) struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block, 2358394a6abSzhangyi (F) int op_flags) 2368394a6abSzhangyi (F) { 2378394a6abSzhangyi (F) return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE); 2388394a6abSzhangyi (F) } 2398394a6abSzhangyi (F) 2408394a6abSzhangyi (F) struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb, 2418394a6abSzhangyi (F) sector_t block) 2428394a6abSzhangyi (F) { 2438394a6abSzhangyi (F) return __ext4_sb_bread_gfp(sb, block, 0, 0); 2448394a6abSzhangyi (F) } 2458394a6abSzhangyi (F) 2465df1d412Szhangyi (F) void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block) 2475df1d412Szhangyi (F) { 2485df1d412Szhangyi (F) struct buffer_head *bh = sb_getblk_gfp(sb, block, 0); 2495df1d412Szhangyi (F) 2505df1d412Szhangyi (F) if (likely(bh)) { 2515df1d412Szhangyi (F) ext4_read_bh_lock(bh, REQ_RAHEAD, false); 2525df1d412Szhangyi (F) brelse(bh); 2535df1d412Szhangyi (F) } 254c197855eSStephen Hemminger } 255a9c47317SDarrick J. Wong 256a9c47317SDarrick J. Wong static int ext4_verify_csum_type(struct super_block *sb, 2579aa5d32bSDmitry Monakhov struct ext4_super_block *es) 258a9c47317SDarrick J. Wong { 259a9c47317SDarrick J. Wong if (!ext4_has_feature_metadata_csum(sb)) 260a9c47317SDarrick J. Wong return 1; 261a9c47317SDarrick J. Wong 262a9c47317SDarrick J. Wong return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 26306db49e6STheodore Ts'o } 264a9c47317SDarrick J. Wong 26506db49e6STheodore Ts'o static __le32 ext4_superblock_csum(struct super_block *sb, 26606db49e6STheodore Ts'o struct ext4_super_block *es) 2679aa5d32bSDmitry Monakhov { 268a9c47317SDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 269a9c47317SDarrick J. Wong int offset = offsetof(struct ext4_super_block, s_checksum); 270a9c47317SDarrick J. Wong __u32 csum; 271a9c47317SDarrick J. Wong 272a9c47317SDarrick J. Wong csum = ext4_chksum(sbi, ~0, (char *)es, offset); 2739933fc0aSTheodore Ts'o 2749933fc0aSTheodore Ts'o return cpu_to_le32(csum); 2759933fc0aSTheodore Ts'o } 2769933fc0aSTheodore Ts'o 2778be04b93SJoe Perches static int ext4_superblock_csum_verify(struct super_block *sb, 2789933fc0aSTheodore Ts'o struct ext4_super_block *es) 2799933fc0aSTheodore Ts'o { 2809933fc0aSTheodore Ts'o if (!ext4_has_metadata_csum(sb)) 2819933fc0aSTheodore Ts'o return 1; 2829933fc0aSTheodore Ts'o 2839933fc0aSTheodore Ts'o return es->s_checksum == ext4_superblock_csum(sb, es); 2849933fc0aSTheodore Ts'o } 2859933fc0aSTheodore Ts'o 2869933fc0aSTheodore Ts'o void ext4_superblock_csum_set(struct super_block *sb) 2878be04b93SJoe Perches { 2889933fc0aSTheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 2899933fc0aSTheodore Ts'o 2909933fc0aSTheodore Ts'o if (!ext4_has_metadata_csum(sb)) 2919933fc0aSTheodore Ts'o return; 2929933fc0aSTheodore Ts'o 2938fadc143SAlexandre Ratchov es->s_checksum = ext4_superblock_csum(sb, es); 2948fadc143SAlexandre Ratchov } 295bd81d8eeSLaurent Vivier 296bd81d8eeSLaurent Vivier ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 297bd81d8eeSLaurent Vivier struct ext4_group_desc *bg) 298bd81d8eeSLaurent Vivier { 2993a14589cSAneesh Kumar K.V return le32_to_cpu(bg->bg_block_bitmap_lo) | 3008fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 3018fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 302bd81d8eeSLaurent Vivier } 303bd81d8eeSLaurent Vivier 3048fadc143SAlexandre Ratchov ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 3058fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 306bd81d8eeSLaurent Vivier { 3075272f837SAneesh Kumar K.V return le32_to_cpu(bg->bg_inode_bitmap_lo) | 3088fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 3098fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 310bd81d8eeSLaurent Vivier } 311bd81d8eeSLaurent Vivier 3128fadc143SAlexandre Ratchov ext4_fsblk_t ext4_inode_table(struct super_block *sb, 3138fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 314bd81d8eeSLaurent Vivier { 3155272f837SAneesh Kumar K.V return le32_to_cpu(bg->bg_inode_table_lo) | 3168fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 3178fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 318bd81d8eeSLaurent Vivier } 319bd81d8eeSLaurent Vivier 320021b65bbSTheodore Ts'o __u32 ext4_free_group_clusters(struct super_block *sb, 321560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 322560671a0SAneesh Kumar K.V { 323560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_free_blocks_count_lo) | 324560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 325560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 326560671a0SAneesh Kumar K.V } 327560671a0SAneesh Kumar K.V 328560671a0SAneesh Kumar K.V __u32 ext4_free_inodes_count(struct super_block *sb, 329560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 330560671a0SAneesh Kumar K.V { 331560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_free_inodes_count_lo) | 332560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 333560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 334560671a0SAneesh Kumar K.V } 335560671a0SAneesh Kumar K.V 336560671a0SAneesh Kumar K.V __u32 ext4_used_dirs_count(struct super_block *sb, 337560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 338560671a0SAneesh Kumar K.V { 339560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_used_dirs_count_lo) | 340560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 341560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 342560671a0SAneesh Kumar K.V } 343560671a0SAneesh Kumar K.V 344560671a0SAneesh Kumar K.V __u32 ext4_itable_unused_count(struct super_block *sb, 345560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 346560671a0SAneesh Kumar K.V { 347560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_itable_unused_lo) | 348560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 349560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 350560671a0SAneesh Kumar K.V } 351560671a0SAneesh Kumar K.V 3528fadc143SAlexandre Ratchov void ext4_block_bitmap_set(struct super_block *sb, 3538fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 354bd81d8eeSLaurent Vivier { 3553a14589cSAneesh Kumar K.V bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 3568fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 3578fadc143SAlexandre Ratchov bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 358bd81d8eeSLaurent Vivier } 359bd81d8eeSLaurent Vivier 3608fadc143SAlexandre Ratchov void ext4_inode_bitmap_set(struct super_block *sb, 3618fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 362bd81d8eeSLaurent Vivier { 3635272f837SAneesh Kumar K.V bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 3648fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 3658fadc143SAlexandre Ratchov bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 366bd81d8eeSLaurent Vivier } 367bd81d8eeSLaurent Vivier 3688fadc143SAlexandre Ratchov void ext4_inode_table_set(struct super_block *sb, 3698fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 370bd81d8eeSLaurent Vivier { 3715272f837SAneesh Kumar K.V bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 3728fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 3738fadc143SAlexandre Ratchov bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 374bd81d8eeSLaurent Vivier } 375bd81d8eeSLaurent Vivier 376021b65bbSTheodore Ts'o void ext4_free_group_clusters_set(struct super_block *sb, 377560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 378560671a0SAneesh Kumar K.V { 379560671a0SAneesh Kumar K.V bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 380560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 381560671a0SAneesh Kumar K.V bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 382560671a0SAneesh Kumar K.V } 383560671a0SAneesh Kumar K.V 384560671a0SAneesh Kumar K.V void ext4_free_inodes_set(struct super_block *sb, 385560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 386560671a0SAneesh Kumar K.V { 387560671a0SAneesh Kumar K.V bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 388560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 389560671a0SAneesh Kumar K.V bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 390560671a0SAneesh Kumar K.V } 391560671a0SAneesh Kumar K.V 392560671a0SAneesh Kumar K.V void ext4_used_dirs_set(struct super_block *sb, 393560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 394560671a0SAneesh Kumar K.V { 395560671a0SAneesh Kumar K.V bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 396560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 397560671a0SAneesh Kumar K.V bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 398560671a0SAneesh Kumar K.V } 399560671a0SAneesh Kumar K.V 400560671a0SAneesh Kumar K.V void ext4_itable_unused_set(struct super_block *sb, 401560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 402560671a0SAneesh Kumar K.V { 403560671a0SAneesh Kumar K.V bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 404560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 405560671a0SAneesh Kumar K.V bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 406560671a0SAneesh Kumar K.V } 407560671a0SAneesh Kumar K.V 408c92dc856SJan Kara static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now) 4096a0678a7SArnd Bergmann { 4106a0678a7SArnd Bergmann now = clamp_val(now, 0, (1ull << 40) - 1); 4116a0678a7SArnd Bergmann 4126a0678a7SArnd Bergmann *lo = cpu_to_le32(lower_32_bits(now)); 4136a0678a7SArnd Bergmann *hi = upper_32_bits(now); 4146a0678a7SArnd Bergmann } 4156a0678a7SArnd Bergmann 4166a0678a7SArnd Bergmann static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi) 4176a0678a7SArnd Bergmann { 4186a0678a7SArnd Bergmann return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo); 4196a0678a7SArnd Bergmann } 4206a0678a7SArnd Bergmann #define ext4_update_tstamp(es, tstamp) \ 421c92dc856SJan Kara __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \ 422c92dc856SJan Kara ktime_get_real_seconds()) 4236a0678a7SArnd Bergmann #define ext4_get_tstamp(es, tstamp) \ 4246a0678a7SArnd Bergmann __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 425d3d1faf6SCurt Wohlgemuth 426bdfe0cbdSTheodore Ts'o /* 427bdfe0cbdSTheodore Ts'o * The del_gendisk() function uninitializes the disk-specific data 428bdfe0cbdSTheodore Ts'o * structures, including the bdi structure, without telling anyone 429bdfe0cbdSTheodore Ts'o * else. Once this happens, any attempt to call mark_buffer_dirty() 430bdfe0cbdSTheodore Ts'o * (for example, by ext4_commit_super), will cause a kernel OOPS. 431bdfe0cbdSTheodore Ts'o * This is a kludge to prevent these oops until we can put in a proper 432bdfe0cbdSTheodore Ts'o * hook in del_gendisk() to inform the VFS and file system layers. 433bdfe0cbdSTheodore Ts'o */ 434bdfe0cbdSTheodore Ts'o static int block_device_ejected(struct super_block *sb) 435bdfe0cbdSTheodore Ts'o { 436bdfe0cbdSTheodore Ts'o struct inode *bd_inode = sb->s_bdev->bd_inode; 437bdfe0cbdSTheodore Ts'o struct backing_dev_info *bdi = inode_to_bdi(bd_inode); 438bdfe0cbdSTheodore Ts'o 439bdfe0cbdSTheodore Ts'o return bdi->dev == NULL; 440bdfe0cbdSTheodore Ts'o } 441bdfe0cbdSTheodore Ts'o 44218aadd47SBobi Jam static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 44318aadd47SBobi Jam { 44418aadd47SBobi Jam struct super_block *sb = journal->j_private; 44518aadd47SBobi Jam struct ext4_sb_info *sbi = EXT4_SB(sb); 44618aadd47SBobi Jam int error = is_journal_aborted(journal); 4475d3ee208SDmitry Monakhov struct ext4_journal_cb_entry *jce; 44818aadd47SBobi Jam 4495d3ee208SDmitry Monakhov BUG_ON(txn->t_state == T_FINISHED); 450a0154344SDaeho Jeong 451a0154344SDaeho Jeong ext4_process_freed_data(sb, txn->t_tid); 452a0154344SDaeho Jeong 45318aadd47SBobi Jam spin_lock(&sbi->s_md_lock); 4545d3ee208SDmitry Monakhov while (!list_empty(&txn->t_private_list)) { 4555d3ee208SDmitry Monakhov jce = list_entry(txn->t_private_list.next, 4565d3ee208SDmitry Monakhov struct ext4_journal_cb_entry, jce_list); 45718aadd47SBobi Jam list_del_init(&jce->jce_list); 45818aadd47SBobi Jam spin_unlock(&sbi->s_md_lock); 45918aadd47SBobi Jam jce->jce_func(sb, jce, error); 46018aadd47SBobi Jam spin_lock(&sbi->s_md_lock); 46118aadd47SBobi Jam } 46218aadd47SBobi Jam spin_unlock(&sbi->s_md_lock); 46318aadd47SBobi Jam } 4641c13d5c0STheodore Ts'o 465afb585a9SMauricio Faria de Oliveira /* 466afb585a9SMauricio Faria de Oliveira * This writepage callback for write_cache_pages() 467afb585a9SMauricio Faria de Oliveira * takes care of a few cases after page cleaning. 468afb585a9SMauricio Faria de Oliveira * 469afb585a9SMauricio Faria de Oliveira * write_cache_pages() already checks for dirty pages 470afb585a9SMauricio Faria de Oliveira * and calls clear_page_dirty_for_io(), which we want, 471afb585a9SMauricio Faria de Oliveira * to write protect the pages. 472afb585a9SMauricio Faria de Oliveira * 473afb585a9SMauricio Faria de Oliveira * However, we may have to redirty a page (see below.) 474afb585a9SMauricio Faria de Oliveira */ 475afb585a9SMauricio Faria de Oliveira static int ext4_journalled_writepage_callback(struct page *page, 476afb585a9SMauricio Faria de Oliveira struct writeback_control *wbc, 477afb585a9SMauricio Faria de Oliveira void *data) 478afb585a9SMauricio Faria de Oliveira { 479afb585a9SMauricio Faria de Oliveira transaction_t *transaction = (transaction_t *) data; 480afb585a9SMauricio Faria de Oliveira struct buffer_head *bh, *head; 481afb585a9SMauricio Faria de Oliveira struct journal_head *jh; 482afb585a9SMauricio Faria de Oliveira 483afb585a9SMauricio Faria de Oliveira bh = head = page_buffers(page); 484afb585a9SMauricio Faria de Oliveira do { 485afb585a9SMauricio Faria de Oliveira /* 486afb585a9SMauricio Faria de Oliveira * We have to redirty a page in these cases: 487afb585a9SMauricio Faria de Oliveira * 1) If buffer is dirty, it means the page was dirty because it 488afb585a9SMauricio Faria de Oliveira * contains a buffer that needs checkpointing. So the dirty bit 489afb585a9SMauricio Faria de Oliveira * needs to be preserved so that checkpointing writes the buffer 490afb585a9SMauricio Faria de Oliveira * properly. 491afb585a9SMauricio Faria de Oliveira * 2) If buffer is not part of the committing transaction 492afb585a9SMauricio Faria de Oliveira * (we may have just accidentally come across this buffer because 493afb585a9SMauricio Faria de Oliveira * inode range tracking is not exact) or if the currently running 494afb585a9SMauricio Faria de Oliveira * transaction already contains this buffer as well, dirty bit 495afb585a9SMauricio Faria de Oliveira * needs to be preserved so that the buffer gets writeprotected 496afb585a9SMauricio Faria de Oliveira * properly on running transaction's commit. 497afb585a9SMauricio Faria de Oliveira */ 498afb585a9SMauricio Faria de Oliveira jh = bh2jh(bh); 499afb585a9SMauricio Faria de Oliveira if (buffer_dirty(bh) || 500afb585a9SMauricio Faria de Oliveira (jh && (jh->b_transaction != transaction || 501afb585a9SMauricio Faria de Oliveira jh->b_next_transaction))) { 502afb585a9SMauricio Faria de Oliveira redirty_page_for_writepage(wbc, page); 503afb585a9SMauricio Faria de Oliveira goto out; 504afb585a9SMauricio Faria de Oliveira } 505afb585a9SMauricio Faria de Oliveira } while ((bh = bh->b_this_page) != head); 506afb585a9SMauricio Faria de Oliveira 507afb585a9SMauricio Faria de Oliveira out: 508afb585a9SMauricio Faria de Oliveira return AOP_WRITEPAGE_ACTIVATE; 509afb585a9SMauricio Faria de Oliveira } 510afb585a9SMauricio Faria de Oliveira 511afb585a9SMauricio Faria de Oliveira static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode) 512afb585a9SMauricio Faria de Oliveira { 513afb585a9SMauricio Faria de Oliveira struct address_space *mapping = jinode->i_vfs_inode->i_mapping; 514afb585a9SMauricio Faria de Oliveira struct writeback_control wbc = { 515afb585a9SMauricio Faria de Oliveira .sync_mode = WB_SYNC_ALL, 516afb585a9SMauricio Faria de Oliveira .nr_to_write = LONG_MAX, 517afb585a9SMauricio Faria de Oliveira .range_start = jinode->i_dirty_start, 518afb585a9SMauricio Faria de Oliveira .range_end = jinode->i_dirty_end, 519afb585a9SMauricio Faria de Oliveira }; 520afb585a9SMauricio Faria de Oliveira 521afb585a9SMauricio Faria de Oliveira return write_cache_pages(mapping, &wbc, 522afb585a9SMauricio Faria de Oliveira ext4_journalled_writepage_callback, 523afb585a9SMauricio Faria de Oliveira jinode->i_transaction); 524afb585a9SMauricio Faria de Oliveira } 525afb585a9SMauricio Faria de Oliveira 526afb585a9SMauricio Faria de Oliveira static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode) 527afb585a9SMauricio Faria de Oliveira { 528afb585a9SMauricio Faria de Oliveira int ret; 529afb585a9SMauricio Faria de Oliveira 530afb585a9SMauricio Faria de Oliveira if (ext4_should_journal_data(jinode->i_vfs_inode)) 531afb585a9SMauricio Faria de Oliveira ret = ext4_journalled_submit_inode_data_buffers(jinode); 532afb585a9SMauricio Faria de Oliveira else 533afb585a9SMauricio Faria de Oliveira ret = jbd2_journal_submit_inode_data_buffers(jinode); 534afb585a9SMauricio Faria de Oliveira 535afb585a9SMauricio Faria de Oliveira return ret; 536afb585a9SMauricio Faria de Oliveira } 537afb585a9SMauricio Faria de Oliveira 538afb585a9SMauricio Faria de Oliveira static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode) 539afb585a9SMauricio Faria de Oliveira { 540afb585a9SMauricio Faria de Oliveira int ret = 0; 541afb585a9SMauricio Faria de Oliveira 542afb585a9SMauricio Faria de Oliveira if (!ext4_should_journal_data(jinode->i_vfs_inode)) 543afb585a9SMauricio Faria de Oliveira ret = jbd2_journal_finish_inode_data_buffers(jinode); 544afb585a9SMauricio Faria de Oliveira 545afb585a9SMauricio Faria de Oliveira return ret; 546afb585a9SMauricio Faria de Oliveira } 547afb585a9SMauricio Faria de Oliveira 5481dc1097fSJan Kara static bool system_going_down(void) 5491dc1097fSJan Kara { 5501dc1097fSJan Kara return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 5511dc1097fSJan Kara || system_state == SYSTEM_RESTART; 5521dc1097fSJan Kara } 5531dc1097fSJan Kara 55402a7780eSJan Kara struct ext4_err_translation { 55502a7780eSJan Kara int code; 55602a7780eSJan Kara int errno; 55702a7780eSJan Kara }; 55802a7780eSJan Kara 55902a7780eSJan Kara #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err } 56002a7780eSJan Kara 56102a7780eSJan Kara static struct ext4_err_translation err_translation[] = { 56202a7780eSJan Kara EXT4_ERR_TRANSLATE(EIO), 56302a7780eSJan Kara EXT4_ERR_TRANSLATE(ENOMEM), 56402a7780eSJan Kara EXT4_ERR_TRANSLATE(EFSBADCRC), 56502a7780eSJan Kara EXT4_ERR_TRANSLATE(EFSCORRUPTED), 56602a7780eSJan Kara EXT4_ERR_TRANSLATE(ENOSPC), 56702a7780eSJan Kara EXT4_ERR_TRANSLATE(ENOKEY), 56802a7780eSJan Kara EXT4_ERR_TRANSLATE(EROFS), 56902a7780eSJan Kara EXT4_ERR_TRANSLATE(EFBIG), 57002a7780eSJan Kara EXT4_ERR_TRANSLATE(EEXIST), 57102a7780eSJan Kara EXT4_ERR_TRANSLATE(ERANGE), 57202a7780eSJan Kara EXT4_ERR_TRANSLATE(EOVERFLOW), 57302a7780eSJan Kara EXT4_ERR_TRANSLATE(EBUSY), 57402a7780eSJan Kara EXT4_ERR_TRANSLATE(ENOTDIR), 57502a7780eSJan Kara EXT4_ERR_TRANSLATE(ENOTEMPTY), 57602a7780eSJan Kara EXT4_ERR_TRANSLATE(ESHUTDOWN), 57702a7780eSJan Kara EXT4_ERR_TRANSLATE(EFAULT), 57802a7780eSJan Kara }; 57902a7780eSJan Kara 58002a7780eSJan Kara static int ext4_errno_to_code(int errno) 58102a7780eSJan Kara { 58202a7780eSJan Kara int i; 58302a7780eSJan Kara 58402a7780eSJan Kara for (i = 0; i < ARRAY_SIZE(err_translation); i++) 58502a7780eSJan Kara if (err_translation[i].errno == errno) 58602a7780eSJan Kara return err_translation[i].code; 58702a7780eSJan Kara return EXT4_ERR_UNKNOWN; 58802a7780eSJan Kara } 58902a7780eSJan Kara 5902d01ddc8SJan Kara static void save_error_info(struct super_block *sb, int error, 59140676623SJan Kara __u32 ino, __u64 block, 59240676623SJan Kara const char *func, unsigned int line) 59340676623SJan Kara { 594c92dc856SJan Kara struct ext4_sb_info *sbi = EXT4_SB(sb); 59540676623SJan Kara 59602a7780eSJan Kara /* We default to EFSCORRUPTED error... */ 59702a7780eSJan Kara if (error == 0) 59802a7780eSJan Kara error = EFSCORRUPTED; 599c92dc856SJan Kara 600c92dc856SJan Kara spin_lock(&sbi->s_error_lock); 601c92dc856SJan Kara sbi->s_add_error_count++; 602c92dc856SJan Kara sbi->s_last_error_code = error; 603c92dc856SJan Kara sbi->s_last_error_line = line; 604c92dc856SJan Kara sbi->s_last_error_ino = ino; 605c92dc856SJan Kara sbi->s_last_error_block = block; 606c92dc856SJan Kara sbi->s_last_error_func = func; 607c92dc856SJan Kara sbi->s_last_error_time = ktime_get_real_seconds(); 608c92dc856SJan Kara if (!sbi->s_first_error_time) { 609c92dc856SJan Kara sbi->s_first_error_code = error; 610c92dc856SJan Kara sbi->s_first_error_line = line; 611c92dc856SJan Kara sbi->s_first_error_ino = ino; 612c92dc856SJan Kara sbi->s_first_error_block = block; 613c92dc856SJan Kara sbi->s_first_error_func = func; 614c92dc856SJan Kara sbi->s_first_error_time = sbi->s_last_error_time; 61540676623SJan Kara } 616c92dc856SJan Kara spin_unlock(&sbi->s_error_lock); 61740676623SJan Kara } 61840676623SJan Kara 619ac27a0ecSDave Kleikamp /* Deal with the reporting of failure conditions on a filesystem such as 620ac27a0ecSDave Kleikamp * inconsistencies detected or read IO failures. 621ac27a0ecSDave Kleikamp * 622ac27a0ecSDave Kleikamp * On ext2, we can store the error state of the filesystem in the 623617ba13bSMingming Cao * superblock. That is not possible on ext4, because we may have other 624ac27a0ecSDave Kleikamp * write ordering constraints on the superblock which prevent us from 625ac27a0ecSDave Kleikamp * writing it out straight away; and given that the journal is about to 626ac27a0ecSDave Kleikamp * be aborted, we can't rely on the current, or future, transactions to 627ac27a0ecSDave Kleikamp * write out the superblock safely. 628ac27a0ecSDave Kleikamp * 629dab291afSMingming Cao * We'll just use the jbd2_journal_abort() error code to record an error in 630d6b198bcSThadeu Lima de Souza Cascardo * the journal instead. On recovery, the journal will complain about 631ac27a0ecSDave Kleikamp * that error until we've noted it down and cleared it. 632014c9caaSJan Kara * 633014c9caaSJan Kara * If force_ro is set, we unconditionally force the filesystem into an 634014c9caaSJan Kara * ABORT|READONLY state, unless the error response on the fs has been set to 635014c9caaSJan Kara * panic in which case we take the easy way out and panic immediately. This is 636014c9caaSJan Kara * used to deal with unrecoverable failures such as journal IO errors or ENOMEM 637014c9caaSJan Kara * at a critical moment in log management. 638ac27a0ecSDave Kleikamp */ 639e789ca0cSJan Kara static void ext4_handle_error(struct super_block *sb, bool force_ro, int error, 640e789ca0cSJan Kara __u32 ino, __u64 block, 641e789ca0cSJan Kara const char *func, unsigned int line) 642ac27a0ecSDave Kleikamp { 643b08070ecSJan Kara journal_t *journal = EXT4_SB(sb)->s_journal; 6442d01ddc8SJan Kara bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT); 645b08070ecSJan Kara 646e789ca0cSJan Kara EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 647327eaf73STheodore Ts'o if (test_opt(sb, WARN_ON_ERROR)) 648327eaf73STheodore Ts'o WARN_ON_ONCE(1); 649327eaf73STheodore Ts'o 6502d01ddc8SJan Kara if (!continue_fs && !sb_rdonly(sb)) { 6519b5f6c9bSHarshad Shirwadkar ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED); 652ac27a0ecSDave Kleikamp if (journal) 653dab291afSMingming Cao jbd2_journal_abort(journal, -EIO); 6542d01ddc8SJan Kara } 6552d01ddc8SJan Kara 6562d01ddc8SJan Kara if (!bdev_read_only(sb->s_bdev)) { 6572d01ddc8SJan Kara save_error_info(sb, error, ino, block, func, line); 6582d01ddc8SJan Kara /* 6592d01ddc8SJan Kara * In case the fs should keep running, we need to writeout 6602d01ddc8SJan Kara * superblock through the journal. Due to lock ordering 6612d01ddc8SJan Kara * constraints, it may not be safe to do it right here so we 6622d01ddc8SJan Kara * defer superblock flushing to a workqueue. 6632d01ddc8SJan Kara */ 6642d01ddc8SJan Kara if (continue_fs) 6652d01ddc8SJan Kara schedule_work(&EXT4_SB(sb)->s_error_work); 6662d01ddc8SJan Kara else 6672d01ddc8SJan Kara ext4_commit_super(sb); 6682d01ddc8SJan Kara } 6692d01ddc8SJan Kara 6702d01ddc8SJan Kara if (sb_rdonly(sb) || continue_fs) 6712d01ddc8SJan Kara return; 6722d01ddc8SJan Kara 6731dc1097fSJan Kara /* 6741dc1097fSJan Kara * We force ERRORS_RO behavior when system is rebooting. Otherwise we 6751dc1097fSJan Kara * could panic during 'reboot -f' as the underlying device got already 6761dc1097fSJan Kara * disabled. 6771dc1097fSJan Kara */ 678014c9caaSJan Kara if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) { 679617ba13bSMingming Cao panic("EXT4-fs (device %s): panic forced after error\n", 680ac27a0ecSDave Kleikamp sb->s_id); 681ac27a0ecSDave Kleikamp } 682014c9caaSJan Kara ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 683014c9caaSJan Kara /* 684014c9caaSJan Kara * Make sure updated value of ->s_mount_flags will be visible before 685014c9caaSJan Kara * ->s_flags update 686014c9caaSJan Kara */ 687014c9caaSJan Kara smp_wmb(); 688014c9caaSJan Kara sb->s_flags |= SB_RDONLY; 6894327ba52SDaeho Jeong } 690ac27a0ecSDave Kleikamp 691c92dc856SJan Kara static void flush_stashed_error_work(struct work_struct *work) 692c92dc856SJan Kara { 693c92dc856SJan Kara struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info, 694c92dc856SJan Kara s_error_work); 6952d01ddc8SJan Kara journal_t *journal = sbi->s_journal; 6962d01ddc8SJan Kara handle_t *handle; 697c92dc856SJan Kara 6982d01ddc8SJan Kara /* 6992d01ddc8SJan Kara * If the journal is still running, we have to write out superblock 7002d01ddc8SJan Kara * through the journal to avoid collisions of other journalled sb 7012d01ddc8SJan Kara * updates. 7022d01ddc8SJan Kara * 7032d01ddc8SJan Kara * We use directly jbd2 functions here to avoid recursing back into 7042d01ddc8SJan Kara * ext4 error handling code during handling of previous errors. 7052d01ddc8SJan Kara */ 7062d01ddc8SJan Kara if (!sb_rdonly(sbi->s_sb) && journal) { 7072d01ddc8SJan Kara handle = jbd2_journal_start(journal, 1); 7082d01ddc8SJan Kara if (IS_ERR(handle)) 7092d01ddc8SJan Kara goto write_directly; 7102d01ddc8SJan Kara if (jbd2_journal_get_write_access(handle, sbi->s_sbh)) { 7112d01ddc8SJan Kara jbd2_journal_stop(handle); 7122d01ddc8SJan Kara goto write_directly; 7132d01ddc8SJan Kara } 7142d01ddc8SJan Kara ext4_update_super(sbi->s_sb); 7152d01ddc8SJan Kara if (jbd2_journal_dirty_metadata(handle, sbi->s_sbh)) { 7162d01ddc8SJan Kara jbd2_journal_stop(handle); 7172d01ddc8SJan Kara goto write_directly; 7182d01ddc8SJan Kara } 7192d01ddc8SJan Kara jbd2_journal_stop(handle); 7202d01ddc8SJan Kara return; 7212d01ddc8SJan Kara } 7222d01ddc8SJan Kara write_directly: 7232d01ddc8SJan Kara /* 7242d01ddc8SJan Kara * Write through journal failed. Write sb directly to get error info 7252d01ddc8SJan Kara * out and hope for the best. 7262d01ddc8SJan Kara */ 7274392fbc4SJan Kara ext4_commit_super(sbi->s_sb); 728ac27a0ecSDave Kleikamp } 729ac27a0ecSDave Kleikamp 730efbed4dcSTheodore Ts'o #define ext4_error_ratelimit(sb) \ 731efbed4dcSTheodore Ts'o ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 732efbed4dcSTheodore Ts'o "EXT4-fs error") 733efbed4dcSTheodore Ts'o 73412062dddSEric Sandeen void __ext4_error(struct super_block *sb, const char *function, 735014c9caaSJan Kara unsigned int line, bool force_ro, int error, __u64 block, 73654d3adbcSTheodore Ts'o const char *fmt, ...) 737ac27a0ecSDave Kleikamp { 7380ff2ea7dSJoe Perches struct va_format vaf; 739ac27a0ecSDave Kleikamp va_list args; 740ac27a0ecSDave Kleikamp 7410db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 7420db1ff22STheodore Ts'o return; 7430db1ff22STheodore Ts'o 744ccf0f32aSTheodore Ts'o trace_ext4_error(sb, function, line); 745efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 746ac27a0ecSDave Kleikamp va_start(args, fmt); 7470ff2ea7dSJoe Perches vaf.fmt = fmt; 7480ff2ea7dSJoe Perches vaf.va = &args; 749efbed4dcSTheodore Ts'o printk(KERN_CRIT 750efbed4dcSTheodore Ts'o "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 7510ff2ea7dSJoe Perches sb->s_id, function, line, current->comm, &vaf); 752ac27a0ecSDave Kleikamp va_end(args); 753efbed4dcSTheodore Ts'o } 754e789ca0cSJan Kara ext4_handle_error(sb, force_ro, error, 0, block, function, line); 755ac27a0ecSDave Kleikamp } 756ac27a0ecSDave Kleikamp 757e7c96e8eSJoe Perches void __ext4_error_inode(struct inode *inode, const char *function, 75854d3adbcSTheodore Ts'o unsigned int line, ext4_fsblk_t block, int error, 759273df556SFrank Mayhar const char *fmt, ...) 760273df556SFrank Mayhar { 761273df556SFrank Mayhar va_list args; 762f7c21177STheodore Ts'o struct va_format vaf; 763273df556SFrank Mayhar 7640db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 7650db1ff22STheodore Ts'o return; 7660db1ff22STheodore Ts'o 767ccf0f32aSTheodore Ts'o trace_ext4_error(inode->i_sb, function, line); 768efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(inode->i_sb)) { 769273df556SFrank Mayhar va_start(args, fmt); 770f7c21177STheodore Ts'o vaf.fmt = fmt; 771f7c21177STheodore Ts'o vaf.va = &args; 772c398eda0STheodore Ts'o if (block) 773d9ee81daSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 774d9ee81daSJoe Perches "inode #%lu: block %llu: comm %s: %pV\n", 775d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 776d9ee81daSJoe Perches block, current->comm, &vaf); 777d9ee81daSJoe Perches else 778d9ee81daSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 779d9ee81daSJoe Perches "inode #%lu: comm %s: %pV\n", 780d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 781d9ee81daSJoe Perches current->comm, &vaf); 782273df556SFrank Mayhar va_end(args); 783efbed4dcSTheodore Ts'o } 784e789ca0cSJan Kara ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block, 78554d3adbcSTheodore Ts'o function, line); 786273df556SFrank Mayhar } 787273df556SFrank Mayhar 788e7c96e8eSJoe Perches void __ext4_error_file(struct file *file, const char *function, 789f7c21177STheodore Ts'o unsigned int line, ext4_fsblk_t block, 790f7c21177STheodore Ts'o const char *fmt, ...) 791273df556SFrank Mayhar { 792273df556SFrank Mayhar va_list args; 793f7c21177STheodore Ts'o struct va_format vaf; 794496ad9aaSAl Viro struct inode *inode = file_inode(file); 795273df556SFrank Mayhar char pathname[80], *path; 796273df556SFrank Mayhar 7970db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 7980db1ff22STheodore Ts'o return; 7990db1ff22STheodore Ts'o 800ccf0f32aSTheodore Ts'o trace_ext4_error(inode->i_sb, function, line); 801efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(inode->i_sb)) { 8029bf39ab2SMiklos Szeredi path = file_path(file, pathname, sizeof(pathname)); 803f9a62d09SDan Carpenter if (IS_ERR(path)) 804273df556SFrank Mayhar path = "(unknown)"; 805f7c21177STheodore Ts'o va_start(args, fmt); 806f7c21177STheodore Ts'o vaf.fmt = fmt; 807f7c21177STheodore Ts'o vaf.va = &args; 808d9ee81daSJoe Perches if (block) 809d9ee81daSJoe Perches printk(KERN_CRIT 810d9ee81daSJoe Perches "EXT4-fs error (device %s): %s:%d: inode #%lu: " 811d9ee81daSJoe Perches "block %llu: comm %s: path %s: %pV\n", 812d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 813d9ee81daSJoe Perches block, current->comm, path, &vaf); 814d9ee81daSJoe Perches else 815d9ee81daSJoe Perches printk(KERN_CRIT 816d9ee81daSJoe Perches "EXT4-fs error (device %s): %s:%d: inode #%lu: " 817d9ee81daSJoe Perches "comm %s: path %s: %pV\n", 818d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 819d9ee81daSJoe Perches current->comm, path, &vaf); 820273df556SFrank Mayhar va_end(args); 821efbed4dcSTheodore Ts'o } 822e789ca0cSJan Kara ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block, 82354d3adbcSTheodore Ts'o function, line); 824273df556SFrank Mayhar } 825273df556SFrank Mayhar 826722887ddSTheodore Ts'o const char *ext4_decode_error(struct super_block *sb, int errno, 827ac27a0ecSDave Kleikamp char nbuf[16]) 828ac27a0ecSDave Kleikamp { 829ac27a0ecSDave Kleikamp char *errstr = NULL; 830ac27a0ecSDave Kleikamp 831ac27a0ecSDave Kleikamp switch (errno) { 8326a797d27SDarrick J. Wong case -EFSCORRUPTED: 8336a797d27SDarrick J. Wong errstr = "Corrupt filesystem"; 8346a797d27SDarrick J. Wong break; 8356a797d27SDarrick J. Wong case -EFSBADCRC: 8366a797d27SDarrick J. Wong errstr = "Filesystem failed CRC"; 8376a797d27SDarrick J. Wong break; 838ac27a0ecSDave Kleikamp case -EIO: 839ac27a0ecSDave Kleikamp errstr = "IO failure"; 840ac27a0ecSDave Kleikamp break; 841ac27a0ecSDave Kleikamp case -ENOMEM: 842ac27a0ecSDave Kleikamp errstr = "Out of memory"; 843ac27a0ecSDave Kleikamp break; 844ac27a0ecSDave Kleikamp case -EROFS: 84578f1ddbbSTheodore Ts'o if (!sb || (EXT4_SB(sb)->s_journal && 84678f1ddbbSTheodore Ts'o EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 847ac27a0ecSDave Kleikamp errstr = "Journal has aborted"; 848ac27a0ecSDave Kleikamp else 849ac27a0ecSDave Kleikamp errstr = "Readonly filesystem"; 850ac27a0ecSDave Kleikamp break; 851ac27a0ecSDave Kleikamp default: 852ac27a0ecSDave Kleikamp /* If the caller passed in an extra buffer for unknown 853ac27a0ecSDave Kleikamp * errors, textualise them now. Else we just return 854ac27a0ecSDave Kleikamp * NULL. */ 855ac27a0ecSDave Kleikamp if (nbuf) { 856ac27a0ecSDave Kleikamp /* Check for truncated error codes... */ 857ac27a0ecSDave Kleikamp if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 858ac27a0ecSDave Kleikamp errstr = nbuf; 859ac27a0ecSDave Kleikamp } 860ac27a0ecSDave Kleikamp break; 861ac27a0ecSDave Kleikamp } 862ac27a0ecSDave Kleikamp 863ac27a0ecSDave Kleikamp return errstr; 864ac27a0ecSDave Kleikamp } 865ac27a0ecSDave Kleikamp 866617ba13bSMingming Cao /* __ext4_std_error decodes expected errors from journaling functions 867ac27a0ecSDave Kleikamp * automatically and invokes the appropriate error response. */ 868ac27a0ecSDave Kleikamp 869c398eda0STheodore Ts'o void __ext4_std_error(struct super_block *sb, const char *function, 870c398eda0STheodore Ts'o unsigned int line, int errno) 871ac27a0ecSDave Kleikamp { 872ac27a0ecSDave Kleikamp char nbuf[16]; 873ac27a0ecSDave Kleikamp const char *errstr; 874ac27a0ecSDave Kleikamp 8750db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 8760db1ff22STheodore Ts'o return; 8770db1ff22STheodore Ts'o 878ac27a0ecSDave Kleikamp /* Special case: if the error is EROFS, and we're not already 879ac27a0ecSDave Kleikamp * inside a transaction, then there's really no point in logging 880ac27a0ecSDave Kleikamp * an error. */ 881bc98a42cSDavid Howells if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb)) 882ac27a0ecSDave Kleikamp return; 883ac27a0ecSDave Kleikamp 884efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 885617ba13bSMingming Cao errstr = ext4_decode_error(sb, errno, nbuf); 886c398eda0STheodore Ts'o printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 887c398eda0STheodore Ts'o sb->s_id, function, line, errstr); 888efbed4dcSTheodore Ts'o } 889ac27a0ecSDave Kleikamp 890e789ca0cSJan Kara ext4_handle_error(sb, false, -errno, 0, 0, function, line); 891ac27a0ecSDave Kleikamp } 892ac27a0ecSDave Kleikamp 893e7c96e8eSJoe Perches void __ext4_msg(struct super_block *sb, 894e7c96e8eSJoe Perches const char *prefix, const char *fmt, ...) 895b31e1552SEric Sandeen { 8960ff2ea7dSJoe Perches struct va_format vaf; 897b31e1552SEric Sandeen va_list args; 898b31e1552SEric Sandeen 8991cf006edSDmitry Monakhov atomic_inc(&EXT4_SB(sb)->s_msg_count); 900efbed4dcSTheodore Ts'o if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs")) 901efbed4dcSTheodore Ts'o return; 902efbed4dcSTheodore Ts'o 903b31e1552SEric Sandeen va_start(args, fmt); 9040ff2ea7dSJoe Perches vaf.fmt = fmt; 9050ff2ea7dSJoe Perches vaf.va = &args; 9060ff2ea7dSJoe Perches printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 907b31e1552SEric Sandeen va_end(args); 908b31e1552SEric Sandeen } 909b31e1552SEric Sandeen 9101cf006edSDmitry Monakhov static int ext4_warning_ratelimit(struct super_block *sb) 9111cf006edSDmitry Monakhov { 9121cf006edSDmitry Monakhov atomic_inc(&EXT4_SB(sb)->s_warning_count); 9131cf006edSDmitry Monakhov return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), 9141cf006edSDmitry Monakhov "EXT4-fs warning"); 9151cf006edSDmitry Monakhov } 916b03a2f7eSAndreas Dilger 91712062dddSEric Sandeen void __ext4_warning(struct super_block *sb, const char *function, 918c398eda0STheodore Ts'o unsigned int line, const char *fmt, ...) 919ac27a0ecSDave Kleikamp { 9200ff2ea7dSJoe Perches struct va_format vaf; 921ac27a0ecSDave Kleikamp va_list args; 922ac27a0ecSDave Kleikamp 923b03a2f7eSAndreas Dilger if (!ext4_warning_ratelimit(sb)) 924efbed4dcSTheodore Ts'o return; 925efbed4dcSTheodore Ts'o 926ac27a0ecSDave Kleikamp va_start(args, fmt); 9270ff2ea7dSJoe Perches vaf.fmt = fmt; 9280ff2ea7dSJoe Perches vaf.va = &args; 9290ff2ea7dSJoe Perches printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 9300ff2ea7dSJoe Perches sb->s_id, function, line, &vaf); 931ac27a0ecSDave Kleikamp va_end(args); 932ac27a0ecSDave Kleikamp } 933ac27a0ecSDave Kleikamp 934b03a2f7eSAndreas Dilger void __ext4_warning_inode(const struct inode *inode, const char *function, 935b03a2f7eSAndreas Dilger unsigned int line, const char *fmt, ...) 936b03a2f7eSAndreas Dilger { 937b03a2f7eSAndreas Dilger struct va_format vaf; 938b03a2f7eSAndreas Dilger va_list args; 939b03a2f7eSAndreas Dilger 940b03a2f7eSAndreas Dilger if (!ext4_warning_ratelimit(inode->i_sb)) 941b03a2f7eSAndreas Dilger return; 942b03a2f7eSAndreas Dilger 943b03a2f7eSAndreas Dilger va_start(args, fmt); 944b03a2f7eSAndreas Dilger vaf.fmt = fmt; 945b03a2f7eSAndreas Dilger vaf.va = &args; 946b03a2f7eSAndreas Dilger printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " 947b03a2f7eSAndreas Dilger "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, 948b03a2f7eSAndreas Dilger function, line, inode->i_ino, current->comm, &vaf); 949b03a2f7eSAndreas Dilger va_end(args); 950b03a2f7eSAndreas Dilger } 951b03a2f7eSAndreas Dilger 952e29136f8STheodore Ts'o void __ext4_grp_locked_error(const char *function, unsigned int line, 953e29136f8STheodore Ts'o struct super_block *sb, ext4_group_t grp, 954e29136f8STheodore Ts'o unsigned long ino, ext4_fsblk_t block, 955e29136f8STheodore Ts'o const char *fmt, ...) 9565d1b1b3fSAneesh Kumar K.V __releases(bitlock) 9575d1b1b3fSAneesh Kumar K.V __acquires(bitlock) 9585d1b1b3fSAneesh Kumar K.V { 9590ff2ea7dSJoe Perches struct va_format vaf; 9605d1b1b3fSAneesh Kumar K.V va_list args; 9615d1b1b3fSAneesh Kumar K.V 9620db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 9630db1ff22STheodore Ts'o return; 9640db1ff22STheodore Ts'o 965ccf0f32aSTheodore Ts'o trace_ext4_error(sb, function, line); 966efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 9675d1b1b3fSAneesh Kumar K.V va_start(args, fmt); 9680ff2ea7dSJoe Perches vaf.fmt = fmt; 9690ff2ea7dSJoe Perches vaf.va = &args; 97021149d61SRobin Dong printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 971e29136f8STheodore Ts'o sb->s_id, function, line, grp); 972e29136f8STheodore Ts'o if (ino) 9730ff2ea7dSJoe Perches printk(KERN_CONT "inode %lu: ", ino); 974e29136f8STheodore Ts'o if (block) 975efbed4dcSTheodore Ts'o printk(KERN_CONT "block %llu:", 976efbed4dcSTheodore Ts'o (unsigned long long) block); 9770ff2ea7dSJoe Perches printk(KERN_CONT "%pV\n", &vaf); 9785d1b1b3fSAneesh Kumar K.V va_end(args); 979efbed4dcSTheodore Ts'o } 9805d1b1b3fSAneesh Kumar K.V 981c92dc856SJan Kara if (test_opt(sb, ERRORS_CONT)) { 982327eaf73STheodore Ts'o if (test_opt(sb, WARN_ON_ERROR)) 983327eaf73STheodore Ts'o WARN_ON_ONCE(1); 984e789ca0cSJan Kara EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 9852d01ddc8SJan Kara if (!bdev_read_only(sb->s_bdev)) { 9862d01ddc8SJan Kara save_error_info(sb, EFSCORRUPTED, ino, block, function, 9872d01ddc8SJan Kara line); 988c92dc856SJan Kara schedule_work(&EXT4_SB(sb)->s_error_work); 9892d01ddc8SJan Kara } 9905d1b1b3fSAneesh Kumar K.V return; 9915d1b1b3fSAneesh Kumar K.V } 9925d1b1b3fSAneesh Kumar K.V ext4_unlock_group(sb, grp); 993e789ca0cSJan Kara ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line); 9945d1b1b3fSAneesh Kumar K.V /* 9955d1b1b3fSAneesh Kumar K.V * We only get here in the ERRORS_RO case; relocking the group 9965d1b1b3fSAneesh Kumar K.V * may be dangerous, but nothing bad will happen since the 9975d1b1b3fSAneesh Kumar K.V * filesystem will have already been marked read/only and the 9985d1b1b3fSAneesh Kumar K.V * journal has been aborted. We return 1 as a hint to callers 9995d1b1b3fSAneesh Kumar K.V * who might what to use the return value from 100025985edcSLucas De Marchi * ext4_grp_locked_error() to distinguish between the 10015d1b1b3fSAneesh Kumar K.V * ERRORS_CONT and ERRORS_RO case, and perhaps return more 10025d1b1b3fSAneesh Kumar K.V * aggressively from the ext4 function in question, with a 10035d1b1b3fSAneesh Kumar K.V * more appropriate error code. 10045d1b1b3fSAneesh Kumar K.V */ 10055d1b1b3fSAneesh Kumar K.V ext4_lock_group(sb, grp); 10065d1b1b3fSAneesh Kumar K.V return; 10075d1b1b3fSAneesh Kumar K.V } 10085d1b1b3fSAneesh Kumar K.V 1009db79e6d1SWang Shilong void ext4_mark_group_bitmap_corrupted(struct super_block *sb, 1010db79e6d1SWang Shilong ext4_group_t group, 1011db79e6d1SWang Shilong unsigned int flags) 1012db79e6d1SWang Shilong { 1013db79e6d1SWang Shilong struct ext4_sb_info *sbi = EXT4_SB(sb); 1014db79e6d1SWang Shilong struct ext4_group_info *grp = ext4_get_group_info(sb, group); 1015db79e6d1SWang Shilong struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); 10169af0b3d1SWang Shilong int ret; 1017db79e6d1SWang Shilong 10189af0b3d1SWang Shilong if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) { 10199af0b3d1SWang Shilong ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, 10209af0b3d1SWang Shilong &grp->bb_state); 10219af0b3d1SWang Shilong if (!ret) 1022db79e6d1SWang Shilong percpu_counter_sub(&sbi->s_freeclusters_counter, 1023db79e6d1SWang Shilong grp->bb_free); 1024db79e6d1SWang Shilong } 1025db79e6d1SWang Shilong 10269af0b3d1SWang Shilong if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) { 10279af0b3d1SWang Shilong ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, 10289af0b3d1SWang Shilong &grp->bb_state); 10299af0b3d1SWang Shilong if (!ret && gdp) { 1030db79e6d1SWang Shilong int count; 1031db79e6d1SWang Shilong 1032db79e6d1SWang Shilong count = ext4_free_inodes_count(sb, gdp); 1033db79e6d1SWang Shilong percpu_counter_sub(&sbi->s_freeinodes_counter, 1034db79e6d1SWang Shilong count); 1035db79e6d1SWang Shilong } 1036db79e6d1SWang Shilong } 1037db79e6d1SWang Shilong } 1038db79e6d1SWang Shilong 1039617ba13bSMingming Cao void ext4_update_dynamic_rev(struct super_block *sb) 1040ac27a0ecSDave Kleikamp { 1041617ba13bSMingming Cao struct ext4_super_block *es = EXT4_SB(sb)->s_es; 1042ac27a0ecSDave Kleikamp 1043617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 1044ac27a0ecSDave Kleikamp return; 1045ac27a0ecSDave Kleikamp 104612062dddSEric Sandeen ext4_warning(sb, 1047ac27a0ecSDave Kleikamp "updating to rev %d because of new feature flag, " 1048ac27a0ecSDave Kleikamp "running e2fsck is recommended", 1049617ba13bSMingming Cao EXT4_DYNAMIC_REV); 1050ac27a0ecSDave Kleikamp 1051617ba13bSMingming Cao es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 1052617ba13bSMingming Cao es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 1053617ba13bSMingming Cao es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 1054ac27a0ecSDave Kleikamp /* leave es->s_feature_*compat flags alone */ 1055ac27a0ecSDave Kleikamp /* es->s_uuid will be set by e2fsck if empty */ 1056ac27a0ecSDave Kleikamp 1057ac27a0ecSDave Kleikamp /* 1058ac27a0ecSDave Kleikamp * The rest of the superblock fields should be zero, and if not it 1059ac27a0ecSDave Kleikamp * means they are likely already in use, so leave them alone. We 1060ac27a0ecSDave Kleikamp * can leave it up to e2fsck to clean up any inconsistencies there. 1061ac27a0ecSDave Kleikamp */ 1062ac27a0ecSDave Kleikamp } 1063ac27a0ecSDave Kleikamp 1064ac27a0ecSDave Kleikamp /* 1065ac27a0ecSDave Kleikamp * Open the external journal device 1066ac27a0ecSDave Kleikamp */ 1067b31e1552SEric Sandeen static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 1068ac27a0ecSDave Kleikamp { 1069ac27a0ecSDave Kleikamp struct block_device *bdev; 1070ac27a0ecSDave Kleikamp 1071d4d77629STejun Heo bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 1072ac27a0ecSDave Kleikamp if (IS_ERR(bdev)) 1073ac27a0ecSDave Kleikamp goto fail; 1074ac27a0ecSDave Kleikamp return bdev; 1075ac27a0ecSDave Kleikamp 1076ac27a0ecSDave Kleikamp fail: 1077ea3edd4dSChristoph Hellwig ext4_msg(sb, KERN_ERR, 1078ea3edd4dSChristoph Hellwig "failed to open journal device unknown-block(%u,%u) %ld", 1079ea3edd4dSChristoph Hellwig MAJOR(dev), MINOR(dev), PTR_ERR(bdev)); 1080ac27a0ecSDave Kleikamp return NULL; 1081ac27a0ecSDave Kleikamp } 1082ac27a0ecSDave Kleikamp 1083ac27a0ecSDave Kleikamp /* 1084ac27a0ecSDave Kleikamp * Release the journal device 1085ac27a0ecSDave Kleikamp */ 10864385bab1SAl Viro static void ext4_blkdev_put(struct block_device *bdev) 1087ac27a0ecSDave Kleikamp { 10884385bab1SAl Viro blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 1089ac27a0ecSDave Kleikamp } 1090ac27a0ecSDave Kleikamp 10914385bab1SAl Viro static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 1092ac27a0ecSDave Kleikamp { 1093ac27a0ecSDave Kleikamp struct block_device *bdev; 1094ee7ed3aaSChunguang Xu bdev = sbi->s_journal_bdev; 1095ac27a0ecSDave Kleikamp if (bdev) { 10964385bab1SAl Viro ext4_blkdev_put(bdev); 1097ee7ed3aaSChunguang Xu sbi->s_journal_bdev = NULL; 1098ac27a0ecSDave Kleikamp } 1099ac27a0ecSDave Kleikamp } 1100ac27a0ecSDave Kleikamp 1101ac27a0ecSDave Kleikamp static inline struct inode *orphan_list_entry(struct list_head *l) 1102ac27a0ecSDave Kleikamp { 1103617ba13bSMingming Cao return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 1104ac27a0ecSDave Kleikamp } 1105ac27a0ecSDave Kleikamp 1106617ba13bSMingming Cao static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 1107ac27a0ecSDave Kleikamp { 1108ac27a0ecSDave Kleikamp struct list_head *l; 1109ac27a0ecSDave Kleikamp 1110b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 1111ac27a0ecSDave Kleikamp le32_to_cpu(sbi->s_es->s_last_orphan)); 1112ac27a0ecSDave Kleikamp 1113ac27a0ecSDave Kleikamp printk(KERN_ERR "sb_info orphan list:\n"); 1114ac27a0ecSDave Kleikamp list_for_each(l, &sbi->s_orphan) { 1115ac27a0ecSDave Kleikamp struct inode *inode = orphan_list_entry(l); 1116ac27a0ecSDave Kleikamp printk(KERN_ERR " " 1117ac27a0ecSDave Kleikamp "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 1118ac27a0ecSDave Kleikamp inode->i_sb->s_id, inode->i_ino, inode, 1119ac27a0ecSDave Kleikamp inode->i_mode, inode->i_nlink, 1120ac27a0ecSDave Kleikamp NEXT_ORPHAN(inode)); 1121ac27a0ecSDave Kleikamp } 1122ac27a0ecSDave Kleikamp } 1123ac27a0ecSDave Kleikamp 1124957153fcSJan Kara #ifdef CONFIG_QUOTA 1125957153fcSJan Kara static int ext4_quota_off(struct super_block *sb, int type); 1126957153fcSJan Kara 1127957153fcSJan Kara static inline void ext4_quota_off_umount(struct super_block *sb) 1128957153fcSJan Kara { 1129957153fcSJan Kara int type; 1130957153fcSJan Kara 1131957153fcSJan Kara /* Use our quota_off function to clear inode flags etc. */ 1132957153fcSJan Kara for (type = 0; type < EXT4_MAXQUOTAS; type++) 1133957153fcSJan Kara ext4_quota_off(sb, type); 1134957153fcSJan Kara } 113533458eabSTheodore Ts'o 113633458eabSTheodore Ts'o /* 113733458eabSTheodore Ts'o * This is a helper function which is used in the mount/remount 113833458eabSTheodore Ts'o * codepaths (which holds s_umount) to fetch the quota file name. 113933458eabSTheodore Ts'o */ 114033458eabSTheodore Ts'o static inline char *get_qf_name(struct super_block *sb, 114133458eabSTheodore Ts'o struct ext4_sb_info *sbi, 114233458eabSTheodore Ts'o int type) 114333458eabSTheodore Ts'o { 114433458eabSTheodore Ts'o return rcu_dereference_protected(sbi->s_qf_names[type], 114533458eabSTheodore Ts'o lockdep_is_held(&sb->s_umount)); 114633458eabSTheodore Ts'o } 1147957153fcSJan Kara #else 1148957153fcSJan Kara static inline void ext4_quota_off_umount(struct super_block *sb) 1149957153fcSJan Kara { 1150957153fcSJan Kara } 1151957153fcSJan Kara #endif 1152957153fcSJan Kara 1153617ba13bSMingming Cao static void ext4_put_super(struct super_block *sb) 1154ac27a0ecSDave Kleikamp { 1155617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 1156617ba13bSMingming Cao struct ext4_super_block *es = sbi->s_es; 11571d0c3924STheodore Ts'o struct buffer_head **group_desc; 11587c990728SSuraj Jitindar Singh struct flex_groups **flex_groups; 115997abd7d4STheodore Ts'o int aborted = 0; 1160ef2cabf7SHidehiro Kawai int i, err; 1161ac27a0ecSDave Kleikamp 1162857ac889SLukas Czerner ext4_unregister_li_request(sb); 1163957153fcSJan Kara ext4_quota_off_umount(sb); 1164e0ccfd95SChristoph Hellwig 1165c92dc856SJan Kara flush_work(&sbi->s_error_work); 11662e8fa54eSJan Kara destroy_workqueue(sbi->rsv_conversion_wq); 11674c0425ffSMingming Cao 11685e47868fSRitesh Harjani /* 11695e47868fSRitesh Harjani * Unregister sysfs before destroying jbd2 journal. 11705e47868fSRitesh Harjani * Since we could still access attr_journal_task attribute via sysfs 11715e47868fSRitesh Harjani * path which could have sbi->s_journal->j_task as NULL 11725e47868fSRitesh Harjani */ 11735e47868fSRitesh Harjani ext4_unregister_sysfs(sb); 11745e47868fSRitesh Harjani 11750390131bSFrank Mayhar if (sbi->s_journal) { 117697abd7d4STheodore Ts'o aborted = is_journal_aborted(sbi->s_journal); 1177ef2cabf7SHidehiro Kawai err = jbd2_journal_destroy(sbi->s_journal); 117847b4a50bSJan Kara sbi->s_journal = NULL; 1179878520acSTheodore Ts'o if ((err < 0) && !aborted) { 118054d3adbcSTheodore Ts'o ext4_abort(sb, -err, "Couldn't clean up the journal"); 11810390131bSFrank Mayhar } 1182878520acSTheodore Ts'o } 1183d4edac31SJosef Bacik 1184d3922a77SZheng Liu ext4_es_unregister_shrinker(sbi); 11859105bb14SAl Viro del_timer_sync(&sbi->s_err_report); 1186d4edac31SJosef Bacik ext4_release_system_zone(sb); 1187d4edac31SJosef Bacik ext4_mb_release(sb); 1188d4edac31SJosef Bacik ext4_ext_release(sb); 1189d4edac31SJosef Bacik 1190bc98a42cSDavid Howells if (!sb_rdonly(sb) && !aborted) { 1191e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 1192ac27a0ecSDave Kleikamp es->s_state = cpu_to_le16(sbi->s_mount_state); 1193ac27a0ecSDave Kleikamp } 1194bc98a42cSDavid Howells if (!sb_rdonly(sb)) 11954392fbc4SJan Kara ext4_commit_super(sb); 1196a8e25a83SArtem Bityutskiy 11971d0c3924STheodore Ts'o rcu_read_lock(); 11981d0c3924STheodore Ts'o group_desc = rcu_dereference(sbi->s_group_desc); 1199ac27a0ecSDave Kleikamp for (i = 0; i < sbi->s_gdb_count; i++) 12001d0c3924STheodore Ts'o brelse(group_desc[i]); 12011d0c3924STheodore Ts'o kvfree(group_desc); 12027c990728SSuraj Jitindar Singh flex_groups = rcu_dereference(sbi->s_flex_groups); 12037c990728SSuraj Jitindar Singh if (flex_groups) { 12047c990728SSuraj Jitindar Singh for (i = 0; i < sbi->s_flex_groups_allocated; i++) 12057c990728SSuraj Jitindar Singh kvfree(flex_groups[i]); 12067c990728SSuraj Jitindar Singh kvfree(flex_groups); 12077c990728SSuraj Jitindar Singh } 12081d0c3924STheodore Ts'o rcu_read_unlock(); 120957042651STheodore Ts'o percpu_counter_destroy(&sbi->s_freeclusters_counter); 1210ac27a0ecSDave Kleikamp percpu_counter_destroy(&sbi->s_freeinodes_counter); 1211ac27a0ecSDave Kleikamp percpu_counter_destroy(&sbi->s_dirs_counter); 121257042651STheodore Ts'o percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 1213bbd55937SEric Biggers percpu_free_rwsem(&sbi->s_writepages_rwsem); 1214ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1215a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 121633458eabSTheodore Ts'o kfree(get_qf_name(sb, sbi, i)); 1217ac27a0ecSDave Kleikamp #endif 1218ac27a0ecSDave Kleikamp 1219ac27a0ecSDave Kleikamp /* Debugging code just in case the in-memory inode orphan list 1220ac27a0ecSDave Kleikamp * isn't empty. The on-disk one can be non-empty if we've 1221ac27a0ecSDave Kleikamp * detected an error and taken the fs readonly, but the 1222ac27a0ecSDave Kleikamp * in-memory list had better be clean by this point. */ 1223ac27a0ecSDave Kleikamp if (!list_empty(&sbi->s_orphan)) 1224ac27a0ecSDave Kleikamp dump_orphan_list(sb, sbi); 1225837c23fbSChunguang Xu ASSERT(list_empty(&sbi->s_orphan)); 1226ac27a0ecSDave Kleikamp 122789d96a6fSTheodore Ts'o sync_blockdev(sb->s_bdev); 1228f98393a6SPeter Zijlstra invalidate_bdev(sb->s_bdev); 1229ee7ed3aaSChunguang Xu if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) { 1230ac27a0ecSDave Kleikamp /* 1231ac27a0ecSDave Kleikamp * Invalidate the journal device's buffers. We don't want them 1232ac27a0ecSDave Kleikamp * floating about in memory - the physical journal device may 1233ac27a0ecSDave Kleikamp * hotswapped, and it breaks the `ro-after' testing code. 1234ac27a0ecSDave Kleikamp */ 1235ee7ed3aaSChunguang Xu sync_blockdev(sbi->s_journal_bdev); 1236ee7ed3aaSChunguang Xu invalidate_bdev(sbi->s_journal_bdev); 1237617ba13bSMingming Cao ext4_blkdev_remove(sbi); 1238ac27a0ecSDave Kleikamp } 123950c15df6SChengguang Xu 1240dec214d0STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 1241dec214d0STahsin Erdogan sbi->s_ea_inode_cache = NULL; 124250c15df6SChengguang Xu 124347387409STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 124447387409STahsin Erdogan sbi->s_ea_block_cache = NULL; 124550c15df6SChengguang Xu 1246c5e06d10SJohann Lombardi if (sbi->s_mmp_tsk) 1247c5e06d10SJohann Lombardi kthread_stop(sbi->s_mmp_tsk); 12489060dd2cSEric Sandeen brelse(sbi->s_sbh); 1249ac27a0ecSDave Kleikamp sb->s_fs_info = NULL; 12503197ebdbSTheodore Ts'o /* 12513197ebdbSTheodore Ts'o * Now that we are completely done shutting down the 12523197ebdbSTheodore Ts'o * superblock, we need to actually destroy the kobject. 12533197ebdbSTheodore Ts'o */ 12543197ebdbSTheodore Ts'o kobject_put(&sbi->s_kobj); 12553197ebdbSTheodore Ts'o wait_for_completion(&sbi->s_kobj_unregister); 12560441984aSDarrick J. Wong if (sbi->s_chksum_driver) 12570441984aSDarrick J. Wong crypto_free_shash(sbi->s_chksum_driver); 1258705895b6SPekka Enberg kfree(sbi->s_blockgroup_lock); 12595e405595SDan Williams fs_put_dax(sbi->s_daxdev); 1260ac4acb1fSEric Biggers fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); 1261c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 1262f8f4acb6SDaniel Rosenberg utf8_unload(sb->s_encoding); 1263c83ad55eSGabriel Krisman Bertazi #endif 1264ac27a0ecSDave Kleikamp kfree(sbi); 1265ac27a0ecSDave Kleikamp } 1266ac27a0ecSDave Kleikamp 1267e18b890bSChristoph Lameter static struct kmem_cache *ext4_inode_cachep; 1268ac27a0ecSDave Kleikamp 1269ac27a0ecSDave Kleikamp /* 1270ac27a0ecSDave Kleikamp * Called inside transaction, so use GFP_NOFS 1271ac27a0ecSDave Kleikamp */ 1272617ba13bSMingming Cao static struct inode *ext4_alloc_inode(struct super_block *sb) 1273ac27a0ecSDave Kleikamp { 1274617ba13bSMingming Cao struct ext4_inode_info *ei; 1275ac27a0ecSDave Kleikamp 1276e6b4f8daSChristoph Lameter ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 1277ac27a0ecSDave Kleikamp if (!ei) 1278ac27a0ecSDave Kleikamp return NULL; 12790b8e58a1SAndreas Dilger 1280ee73f9a5SJeff Layton inode_set_iversion(&ei->vfs_inode, 1); 1281202ee5dfSTheodore Ts'o spin_lock_init(&ei->i_raw_lock); 1282c9de560dSAlex Tomas INIT_LIST_HEAD(&ei->i_prealloc_list); 128327bc446eSbrookxu atomic_set(&ei->i_prealloc_active, 0); 1284c9de560dSAlex Tomas spin_lock_init(&ei->i_prealloc_lock); 12859a26b661SZheng Liu ext4_es_init_tree(&ei->i_es_tree); 12869a26b661SZheng Liu rwlock_init(&ei->i_es_lock); 1287edaa53caSZheng Liu INIT_LIST_HEAD(&ei->i_es_list); 1288eb68d0e2SZheng Liu ei->i_es_all_nr = 0; 1289edaa53caSZheng Liu ei->i_es_shk_nr = 0; 1290dd475925SJan Kara ei->i_es_shrink_lblk = 0; 1291d2a17637SMingming Cao ei->i_reserved_data_blocks = 0; 1292d2a17637SMingming Cao spin_lock_init(&(ei->i_block_reservation_lock)); 12931dc0aa46SEric Whitney ext4_init_pending_tree(&ei->i_pending_tree); 1294a9e7f447SDmitry Monakhov #ifdef CONFIG_QUOTA 1295a9e7f447SDmitry Monakhov ei->i_reserved_quota = 0; 129696c7e0d9SJan Kara memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 1297a9e7f447SDmitry Monakhov #endif 12988aefcd55STheodore Ts'o ei->jinode = NULL; 12992e8fa54eSJan Kara INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 1300744692dcSJiaying Zhang spin_lock_init(&ei->i_completed_io_lock); 1301b436b9beSJan Kara ei->i_sync_tid = 0; 1302b436b9beSJan Kara ei->i_datasync_tid = 0; 1303e27f41e1SDmitry Monakhov atomic_set(&ei->i_unwritten, 0); 13042e8fa54eSJan Kara INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 1305aa75f4d3SHarshad Shirwadkar ext4_fc_init_inode(&ei->vfs_inode); 1306aa75f4d3SHarshad Shirwadkar mutex_init(&ei->i_fc_lock); 1307ac27a0ecSDave Kleikamp return &ei->vfs_inode; 1308ac27a0ecSDave Kleikamp } 1309ac27a0ecSDave Kleikamp 13107ff9c073STheodore Ts'o static int ext4_drop_inode(struct inode *inode) 13117ff9c073STheodore Ts'o { 13127ff9c073STheodore Ts'o int drop = generic_drop_inode(inode); 13137ff9c073STheodore Ts'o 131429b3692eSEric Biggers if (!drop) 131529b3692eSEric Biggers drop = fscrypt_drop_inode(inode); 131629b3692eSEric Biggers 13177ff9c073STheodore Ts'o trace_ext4_drop_inode(inode, drop); 13187ff9c073STheodore Ts'o return drop; 13197ff9c073STheodore Ts'o } 13207ff9c073STheodore Ts'o 132194053139SAl Viro static void ext4_free_in_core_inode(struct inode *inode) 1322fa0d7e3dSNick Piggin { 13232c58d548SEric Biggers fscrypt_free_inode(inode); 1324aa75f4d3SHarshad Shirwadkar if (!list_empty(&(EXT4_I(inode)->i_fc_list))) { 1325aa75f4d3SHarshad Shirwadkar pr_warn("%s: inode %ld still in fc list", 1326aa75f4d3SHarshad Shirwadkar __func__, inode->i_ino); 1327aa75f4d3SHarshad Shirwadkar } 1328fa0d7e3dSNick Piggin kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 1329fa0d7e3dSNick Piggin } 1330fa0d7e3dSNick Piggin 1331617ba13bSMingming Cao static void ext4_destroy_inode(struct inode *inode) 1332ac27a0ecSDave Kleikamp { 13339f7dd93dSVasily Averin if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 1334b31e1552SEric Sandeen ext4_msg(inode->i_sb, KERN_ERR, 1335b31e1552SEric Sandeen "Inode %lu (%p): orphan list check failed!", 1336b31e1552SEric Sandeen inode->i_ino, EXT4_I(inode)); 13379f7dd93dSVasily Averin print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 13389f7dd93dSVasily Averin EXT4_I(inode), sizeof(struct ext4_inode_info), 13399f7dd93dSVasily Averin true); 13409f7dd93dSVasily Averin dump_stack(); 13419f7dd93dSVasily Averin } 1342ac27a0ecSDave Kleikamp } 1343ac27a0ecSDave Kleikamp 134451cc5068SAlexey Dobriyan static void init_once(void *foo) 1345ac27a0ecSDave Kleikamp { 1346617ba13bSMingming Cao struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 1347ac27a0ecSDave Kleikamp 1348ac27a0ecSDave Kleikamp INIT_LIST_HEAD(&ei->i_orphan); 1349ac27a0ecSDave Kleikamp init_rwsem(&ei->xattr_sem); 13500e855ac8SAneesh Kumar K.V init_rwsem(&ei->i_data_sem); 1351ea3d7209SJan Kara init_rwsem(&ei->i_mmap_sem); 1352ac27a0ecSDave Kleikamp inode_init_once(&ei->vfs_inode); 1353aa75f4d3SHarshad Shirwadkar ext4_fc_init_inode(&ei->vfs_inode); 1354ac27a0ecSDave Kleikamp } 1355ac27a0ecSDave Kleikamp 1356e67bc2b3SFabian Frederick static int __init init_inodecache(void) 1357ac27a0ecSDave Kleikamp { 1358f8dd7c70SDavid Windsor ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache", 1359f8dd7c70SDavid Windsor sizeof(struct ext4_inode_info), 0, 1360f8dd7c70SDavid Windsor (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD| 1361f8dd7c70SDavid Windsor SLAB_ACCOUNT), 1362f8dd7c70SDavid Windsor offsetof(struct ext4_inode_info, i_data), 1363f8dd7c70SDavid Windsor sizeof_field(struct ext4_inode_info, i_data), 136420c2df83SPaul Mundt init_once); 1365617ba13bSMingming Cao if (ext4_inode_cachep == NULL) 1366ac27a0ecSDave Kleikamp return -ENOMEM; 1367ac27a0ecSDave Kleikamp return 0; 1368ac27a0ecSDave Kleikamp } 1369ac27a0ecSDave Kleikamp 1370ac27a0ecSDave Kleikamp static void destroy_inodecache(void) 1371ac27a0ecSDave Kleikamp { 13728c0a8537SKirill A. Shutemov /* 13738c0a8537SKirill A. Shutemov * Make sure all delayed rcu free inodes are flushed before we 13748c0a8537SKirill A. Shutemov * destroy cache. 13758c0a8537SKirill A. Shutemov */ 13768c0a8537SKirill A. Shutemov rcu_barrier(); 1377617ba13bSMingming Cao kmem_cache_destroy(ext4_inode_cachep); 1378ac27a0ecSDave Kleikamp } 1379ac27a0ecSDave Kleikamp 13800930fcc1SAl Viro void ext4_clear_inode(struct inode *inode) 1381ac27a0ecSDave Kleikamp { 1382aa75f4d3SHarshad Shirwadkar ext4_fc_del(inode); 13830930fcc1SAl Viro invalidate_inode_buffers(inode); 1384dbd5768fSJan Kara clear_inode(inode); 138527bc446eSbrookxu ext4_discard_preallocations(inode, 0); 138651865fdaSZheng Liu ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1387f4c2d372SJan Kara dquot_drop(inode); 13888aefcd55STheodore Ts'o if (EXT4_I(inode)->jinode) { 13898aefcd55STheodore Ts'o jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 13908aefcd55STheodore Ts'o EXT4_I(inode)->jinode); 13918aefcd55STheodore Ts'o jbd2_free_inode(EXT4_I(inode)->jinode); 13928aefcd55STheodore Ts'o EXT4_I(inode)->jinode = NULL; 13938aefcd55STheodore Ts'o } 13943d204e24SEric Biggers fscrypt_put_encryption_info(inode); 1395c93d8f88SEric Biggers fsverity_cleanup_inode(inode); 1396ac27a0ecSDave Kleikamp } 1397ac27a0ecSDave Kleikamp 13981b961ac0SChristoph Hellwig static struct inode *ext4_nfs_get_inode(struct super_block *sb, 13991b961ac0SChristoph Hellwig u64 ino, u32 generation) 1400ac27a0ecSDave Kleikamp { 1401ac27a0ecSDave Kleikamp struct inode *inode; 1402ac27a0ecSDave Kleikamp 14038a363970STheodore Ts'o /* 1404ac27a0ecSDave Kleikamp * Currently we don't know the generation for parent directory, so 1405ac27a0ecSDave Kleikamp * a generation of 0 means "accept any" 1406ac27a0ecSDave Kleikamp */ 14078a363970STheodore Ts'o inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); 14081d1fe1eeSDavid Howells if (IS_ERR(inode)) 14091d1fe1eeSDavid Howells return ERR_CAST(inode); 14101d1fe1eeSDavid Howells if (generation && inode->i_generation != generation) { 1411ac27a0ecSDave Kleikamp iput(inode); 1412ac27a0ecSDave Kleikamp return ERR_PTR(-ESTALE); 1413ac27a0ecSDave Kleikamp } 14141b961ac0SChristoph Hellwig 14151b961ac0SChristoph Hellwig return inode; 1416ac27a0ecSDave Kleikamp } 14171b961ac0SChristoph Hellwig 14181b961ac0SChristoph Hellwig static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 14191b961ac0SChristoph Hellwig int fh_len, int fh_type) 14201b961ac0SChristoph Hellwig { 14211b961ac0SChristoph Hellwig return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 14221b961ac0SChristoph Hellwig ext4_nfs_get_inode); 14231b961ac0SChristoph Hellwig } 14241b961ac0SChristoph Hellwig 14251b961ac0SChristoph Hellwig static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 14261b961ac0SChristoph Hellwig int fh_len, int fh_type) 14271b961ac0SChristoph Hellwig { 14281b961ac0SChristoph Hellwig return generic_fh_to_parent(sb, fid, fh_len, fh_type, 14291b961ac0SChristoph Hellwig ext4_nfs_get_inode); 1430ac27a0ecSDave Kleikamp } 1431ac27a0ecSDave Kleikamp 1432fde87268STheodore Ts'o static int ext4_nfs_commit_metadata(struct inode *inode) 1433fde87268STheodore Ts'o { 1434fde87268STheodore Ts'o struct writeback_control wbc = { 1435fde87268STheodore Ts'o .sync_mode = WB_SYNC_ALL 1436fde87268STheodore Ts'o }; 1437fde87268STheodore Ts'o 1438fde87268STheodore Ts'o trace_ext4_nfs_commit_metadata(inode); 1439fde87268STheodore Ts'o return ext4_write_inode(inode, &wbc); 1440fde87268STheodore Ts'o } 1441fde87268STheodore Ts'o 1442c39a7f84SToshiyuki Okajima /* 1443c39a7f84SToshiyuki Okajima * Try to release metadata pages (indirect blocks, directories) which are 1444c39a7f84SToshiyuki Okajima * mapped via the block device. Since these pages could have journal heads 1445c39a7f84SToshiyuki Okajima * which would prevent try_to_free_buffers() from freeing them, we must use 1446c39a7f84SToshiyuki Okajima * jbd2 layer's try_to_free_buffers() function to release them. 1447c39a7f84SToshiyuki Okajima */ 14480b8e58a1SAndreas Dilger static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 14490b8e58a1SAndreas Dilger gfp_t wait) 1450c39a7f84SToshiyuki Okajima { 1451c39a7f84SToshiyuki Okajima journal_t *journal = EXT4_SB(sb)->s_journal; 1452c39a7f84SToshiyuki Okajima 1453c39a7f84SToshiyuki Okajima WARN_ON(PageChecked(page)); 1454c39a7f84SToshiyuki Okajima if (!page_has_buffers(page)) 1455c39a7f84SToshiyuki Okajima return 0; 1456c39a7f84SToshiyuki Okajima if (journal) 1457529a781eSzhangyi (F) return jbd2_journal_try_to_free_buffers(journal, page); 1458529a781eSzhangyi (F) 1459c39a7f84SToshiyuki Okajima return try_to_free_buffers(page); 1460c39a7f84SToshiyuki Okajima } 1461c39a7f84SToshiyuki Okajima 1462643fa961SChandan Rajendra #ifdef CONFIG_FS_ENCRYPTION 1463a7550b30SJaegeuk Kim static int ext4_get_context(struct inode *inode, void *ctx, size_t len) 1464a7550b30SJaegeuk Kim { 1465a7550b30SJaegeuk Kim return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION, 1466a7550b30SJaegeuk Kim EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len); 1467a7550b30SJaegeuk Kim } 1468a7550b30SJaegeuk Kim 1469a7550b30SJaegeuk Kim static int ext4_set_context(struct inode *inode, const void *ctx, size_t len, 1470a7550b30SJaegeuk Kim void *fs_data) 1471a7550b30SJaegeuk Kim { 14722f8f5e76SEric Biggers handle_t *handle = fs_data; 1473c1a5d5f6STahsin Erdogan int res, res2, credits, retries = 0; 1474a7550b30SJaegeuk Kim 14759ce0151aSEric Biggers /* 14769ce0151aSEric Biggers * Encrypting the root directory is not allowed because e2fsck expects 14779ce0151aSEric Biggers * lost+found to exist and be unencrypted, and encrypting the root 14789ce0151aSEric Biggers * directory would imply encrypting the lost+found directory as well as 14799ce0151aSEric Biggers * the filename "lost+found" itself. 14809ce0151aSEric Biggers */ 14819ce0151aSEric Biggers if (inode->i_ino == EXT4_ROOT_INO) 14829ce0151aSEric Biggers return -EPERM; 1483a7550b30SJaegeuk Kim 14847d3e06a8SRoss Zwisler if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode))) 14857d3e06a8SRoss Zwisler return -EINVAL; 14867d3e06a8SRoss Zwisler 1487b383a73fSIra Weiny if (ext4_test_inode_flag(inode, EXT4_INODE_DAX)) 1488b383a73fSIra Weiny return -EOPNOTSUPP; 1489b383a73fSIra Weiny 149094840e3cSEric Biggers res = ext4_convert_inline_data(inode); 149194840e3cSEric Biggers if (res) 149294840e3cSEric Biggers return res; 149394840e3cSEric Biggers 14942f8f5e76SEric Biggers /* 14952f8f5e76SEric Biggers * If a journal handle was specified, then the encryption context is 14962f8f5e76SEric Biggers * being set on a new inode via inheritance and is part of a larger 14972f8f5e76SEric Biggers * transaction to create the inode. Otherwise the encryption context is 14982f8f5e76SEric Biggers * being set on an existing inode in its own transaction. Only in the 14992f8f5e76SEric Biggers * latter case should the "retry on ENOSPC" logic be used. 15002f8f5e76SEric Biggers */ 15012f8f5e76SEric Biggers 15022f8f5e76SEric Biggers if (handle) { 15032f8f5e76SEric Biggers res = ext4_xattr_set_handle(handle, inode, 15042f8f5e76SEric Biggers EXT4_XATTR_INDEX_ENCRYPTION, 15052f8f5e76SEric Biggers EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 15062f8f5e76SEric Biggers ctx, len, 0); 1507a7550b30SJaegeuk Kim if (!res) { 1508a7550b30SJaegeuk Kim ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1509a7550b30SJaegeuk Kim ext4_clear_inode_state(inode, 1510a7550b30SJaegeuk Kim EXT4_STATE_MAY_INLINE_DATA); 1511a3caa24bSJan Kara /* 15122ee6a576SEric Biggers * Update inode->i_flags - S_ENCRYPTED will be enabled, 15132ee6a576SEric Biggers * S_DAX may be disabled 1514a3caa24bSJan Kara */ 1515043546e4SIra Weiny ext4_set_inode_flags(inode, false); 1516a7550b30SJaegeuk Kim } 1517a7550b30SJaegeuk Kim return res; 1518a7550b30SJaegeuk Kim } 1519a7550b30SJaegeuk Kim 1520b8cb5a54STahsin Erdogan res = dquot_initialize(inode); 1521b8cb5a54STahsin Erdogan if (res) 1522b8cb5a54STahsin Erdogan return res; 15232f8f5e76SEric Biggers retry: 1524af65207cSTahsin Erdogan res = ext4_xattr_set_credits(inode, len, false /* is_create */, 1525af65207cSTahsin Erdogan &credits); 1526dec214d0STahsin Erdogan if (res) 1527dec214d0STahsin Erdogan return res; 1528dec214d0STahsin Erdogan 1529c1a5d5f6STahsin Erdogan handle = ext4_journal_start(inode, EXT4_HT_MISC, credits); 1530a7550b30SJaegeuk Kim if (IS_ERR(handle)) 1531a7550b30SJaegeuk Kim return PTR_ERR(handle); 1532a7550b30SJaegeuk Kim 15332f8f5e76SEric Biggers res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION, 15342f8f5e76SEric Biggers EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 15352f8f5e76SEric Biggers ctx, len, 0); 1536a7550b30SJaegeuk Kim if (!res) { 1537a7550b30SJaegeuk Kim ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 15382ee6a576SEric Biggers /* 15392ee6a576SEric Biggers * Update inode->i_flags - S_ENCRYPTED will be enabled, 15402ee6a576SEric Biggers * S_DAX may be disabled 15412ee6a576SEric Biggers */ 1542043546e4SIra Weiny ext4_set_inode_flags(inode, false); 1543a7550b30SJaegeuk Kim res = ext4_mark_inode_dirty(handle, inode); 1544a7550b30SJaegeuk Kim if (res) 1545a7550b30SJaegeuk Kim EXT4_ERROR_INODE(inode, "Failed to mark inode dirty"); 1546a7550b30SJaegeuk Kim } 1547a7550b30SJaegeuk Kim res2 = ext4_journal_stop(handle); 15482f8f5e76SEric Biggers 15492f8f5e76SEric Biggers if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) 15502f8f5e76SEric Biggers goto retry; 1551a7550b30SJaegeuk Kim if (!res) 1552a7550b30SJaegeuk Kim res = res2; 1553a7550b30SJaegeuk Kim return res; 1554a7550b30SJaegeuk Kim } 1555a7550b30SJaegeuk Kim 1556ac4acb1fSEric Biggers static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb) 1557a7550b30SJaegeuk Kim { 1558ac4acb1fSEric Biggers return EXT4_SB(sb)->s_dummy_enc_policy.policy; 1559a7550b30SJaegeuk Kim } 1560a7550b30SJaegeuk Kim 1561b925acb8SEric Biggers static bool ext4_has_stable_inodes(struct super_block *sb) 1562b925acb8SEric Biggers { 1563b925acb8SEric Biggers return ext4_has_feature_stable_inodes(sb); 1564b925acb8SEric Biggers } 1565b925acb8SEric Biggers 1566b925acb8SEric Biggers static void ext4_get_ino_and_lblk_bits(struct super_block *sb, 1567b925acb8SEric Biggers int *ino_bits_ret, int *lblk_bits_ret) 1568b925acb8SEric Biggers { 1569b925acb8SEric Biggers *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count); 1570b925acb8SEric Biggers *lblk_bits_ret = 8 * sizeof(ext4_lblk_t); 1571b925acb8SEric Biggers } 1572b925acb8SEric Biggers 15736f69f0edSEric Biggers static const struct fscrypt_operations ext4_cryptops = { 1574a5d431efSEric Biggers .key_prefix = "ext4:", 1575a7550b30SJaegeuk Kim .get_context = ext4_get_context, 1576a7550b30SJaegeuk Kim .set_context = ext4_set_context, 1577ac4acb1fSEric Biggers .get_dummy_policy = ext4_get_dummy_policy, 1578a7550b30SJaegeuk Kim .empty_dir = ext4_empty_dir, 1579e12ee683SEric Biggers .max_namelen = EXT4_NAME_LEN, 1580b925acb8SEric Biggers .has_stable_inodes = ext4_has_stable_inodes, 1581b925acb8SEric Biggers .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits, 1582a7550b30SJaegeuk Kim }; 1583a7550b30SJaegeuk Kim #endif 1584a7550b30SJaegeuk Kim 1585ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1586d6006186SEric Biggers static const char * const quotatypes[] = INITQFNAMES; 1587689c958cSLi Xi #define QTYPE2NAME(t) (quotatypes[t]) 1588ac27a0ecSDave Kleikamp 1589617ba13bSMingming Cao static int ext4_write_dquot(struct dquot *dquot); 1590617ba13bSMingming Cao static int ext4_acquire_dquot(struct dquot *dquot); 1591617ba13bSMingming Cao static int ext4_release_dquot(struct dquot *dquot); 1592617ba13bSMingming Cao static int ext4_mark_dquot_dirty(struct dquot *dquot); 1593617ba13bSMingming Cao static int ext4_write_info(struct super_block *sb, int type); 15946f28e087SJan Kara static int ext4_quota_on(struct super_block *sb, int type, int format_id, 15958c54ca9cSAl Viro const struct path *path); 1596617ba13bSMingming Cao static int ext4_quota_on_mount(struct super_block *sb, int type); 1597617ba13bSMingming Cao static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1598ac27a0ecSDave Kleikamp size_t len, loff_t off); 1599617ba13bSMingming Cao static ssize_t ext4_quota_write(struct super_block *sb, int type, 1600ac27a0ecSDave Kleikamp const char *data, size_t len, loff_t off); 16017c319d32SAditya Kali static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 16027c319d32SAditya Kali unsigned int flags); 16037c319d32SAditya Kali static int ext4_enable_quotas(struct super_block *sb); 1604ac27a0ecSDave Kleikamp 160596c7e0d9SJan Kara static struct dquot **ext4_get_dquots(struct inode *inode) 160696c7e0d9SJan Kara { 160796c7e0d9SJan Kara return EXT4_I(inode)->i_dquot; 160896c7e0d9SJan Kara } 160996c7e0d9SJan Kara 161061e225dcSAlexey Dobriyan static const struct dquot_operations ext4_quota_operations = { 161160e58e0fSMingming Cao .get_reserved_space = ext4_get_reserved_space, 1612617ba13bSMingming Cao .write_dquot = ext4_write_dquot, 1613617ba13bSMingming Cao .acquire_dquot = ext4_acquire_dquot, 1614617ba13bSMingming Cao .release_dquot = ext4_release_dquot, 1615617ba13bSMingming Cao .mark_dirty = ext4_mark_dquot_dirty, 1616a5b5ee32SJan Kara .write_info = ext4_write_info, 1617a5b5ee32SJan Kara .alloc_dquot = dquot_alloc, 1618a5b5ee32SJan Kara .destroy_dquot = dquot_destroy, 1619040cb378SLi Xi .get_projid = ext4_get_projid, 16207a9ca53aSTahsin Erdogan .get_inode_usage = ext4_get_inode_usage, 1621ebc11f7bSChengguang Xu .get_next_id = dquot_get_next_id, 1622ac27a0ecSDave Kleikamp }; 1623ac27a0ecSDave Kleikamp 16240d54b217SAlexey Dobriyan static const struct quotactl_ops ext4_qctl_operations = { 1625617ba13bSMingming Cao .quota_on = ext4_quota_on, 1626ca0e05e4SDmitry Monakhov .quota_off = ext4_quota_off, 1627287a8095SChristoph Hellwig .quota_sync = dquot_quota_sync, 16280a240339SJan Kara .get_state = dquot_get_state, 1629287a8095SChristoph Hellwig .set_info = dquot_set_dqinfo, 1630287a8095SChristoph Hellwig .get_dqblk = dquot_get_dqblk, 16316332b9b5SEric Sandeen .set_dqblk = dquot_set_dqblk, 16326332b9b5SEric Sandeen .get_nextdqblk = dquot_get_next_dqblk, 1633ac27a0ecSDave Kleikamp }; 1634ac27a0ecSDave Kleikamp #endif 1635ac27a0ecSDave Kleikamp 1636ee9b6d61SJosef 'Jeff' Sipek static const struct super_operations ext4_sops = { 1637617ba13bSMingming Cao .alloc_inode = ext4_alloc_inode, 163894053139SAl Viro .free_inode = ext4_free_in_core_inode, 1639617ba13bSMingming Cao .destroy_inode = ext4_destroy_inode, 1640617ba13bSMingming Cao .write_inode = ext4_write_inode, 1641617ba13bSMingming Cao .dirty_inode = ext4_dirty_inode, 16427ff9c073STheodore Ts'o .drop_inode = ext4_drop_inode, 16430930fcc1SAl Viro .evict_inode = ext4_evict_inode, 1644617ba13bSMingming Cao .put_super = ext4_put_super, 1645617ba13bSMingming Cao .sync_fs = ext4_sync_fs, 1646c4be0c1dSTakashi Sato .freeze_fs = ext4_freeze, 1647c4be0c1dSTakashi Sato .unfreeze_fs = ext4_unfreeze, 1648617ba13bSMingming Cao .statfs = ext4_statfs, 1649617ba13bSMingming Cao .remount_fs = ext4_remount, 1650617ba13bSMingming Cao .show_options = ext4_show_options, 1651ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1652617ba13bSMingming Cao .quota_read = ext4_quota_read, 1653617ba13bSMingming Cao .quota_write = ext4_quota_write, 165496c7e0d9SJan Kara .get_dquots = ext4_get_dquots, 1655ac27a0ecSDave Kleikamp #endif 1656c39a7f84SToshiyuki Okajima .bdev_try_to_free_page = bdev_try_to_free_page, 1657ac27a0ecSDave Kleikamp }; 1658ac27a0ecSDave Kleikamp 165939655164SChristoph Hellwig static const struct export_operations ext4_export_ops = { 16601b961ac0SChristoph Hellwig .fh_to_dentry = ext4_fh_to_dentry, 16611b961ac0SChristoph Hellwig .fh_to_parent = ext4_fh_to_parent, 1662617ba13bSMingming Cao .get_parent = ext4_get_parent, 1663fde87268STheodore Ts'o .commit_metadata = ext4_nfs_commit_metadata, 1664ac27a0ecSDave Kleikamp }; 1665ac27a0ecSDave Kleikamp 1666ac27a0ecSDave Kleikamp enum { 1667ac27a0ecSDave Kleikamp Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1668ac27a0ecSDave Kleikamp Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 166972578c33STheodore Ts'o Opt_nouid32, Opt_debug, Opt_removed, 1670ac27a0ecSDave Kleikamp Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 167172578c33STheodore Ts'o Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1672ad4eec61SEric Sandeen Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1673ad4eec61SEric Sandeen Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1674ac27a0ecSDave Kleikamp Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 16756ddb2447STheodore Ts'o Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 16764f74d15fSEric Biggers Opt_inlinecrypt, 1677ac27a0ecSDave Kleikamp Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 16785a20bdfcSJan Kara Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1679ee4a3fcdSTheodore Ts'o Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 16809cb20f94SIra Weiny Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, 16819cb20f94SIra Weiny Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never, 1682327eaf73STheodore Ts'o Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1683327eaf73STheodore Ts'o Opt_nowarn_on_error, Opt_mblk_io_submit, 1684670e9875STheodore Ts'o Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize, 16851449032bSTheodore Ts'o Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 16865328e635SEric Sandeen Opt_inode_readahead_blks, Opt_journal_ioprio, 1687744692dcSJiaying Zhang Opt_dioread_nolock, Opt_dioread_lock, 1688fc6cb1cdSTheodore Ts'o Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1689cdb7ee4cSTahsin Erdogan Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 169099c880deSHarshad Shirwadkar Opt_prefetch_block_bitmaps, 16918016e29fSHarshad Shirwadkar #ifdef CONFIG_EXT4_DEBUG 169299c880deSHarshad Shirwadkar Opt_fc_debug_max_replay, Opt_fc_debug_force 16938016e29fSHarshad Shirwadkar #endif 1694ac27a0ecSDave Kleikamp }; 1695ac27a0ecSDave Kleikamp 1696a447c093SSteven Whitehouse static const match_table_t tokens = { 1697ac27a0ecSDave Kleikamp {Opt_bsd_df, "bsddf"}, 1698ac27a0ecSDave Kleikamp {Opt_minix_df, "minixdf"}, 1699ac27a0ecSDave Kleikamp {Opt_grpid, "grpid"}, 1700ac27a0ecSDave Kleikamp {Opt_grpid, "bsdgroups"}, 1701ac27a0ecSDave Kleikamp {Opt_nogrpid, "nogrpid"}, 1702ac27a0ecSDave Kleikamp {Opt_nogrpid, "sysvgroups"}, 1703ac27a0ecSDave Kleikamp {Opt_resgid, "resgid=%u"}, 1704ac27a0ecSDave Kleikamp {Opt_resuid, "resuid=%u"}, 1705ac27a0ecSDave Kleikamp {Opt_sb, "sb=%u"}, 1706ac27a0ecSDave Kleikamp {Opt_err_cont, "errors=continue"}, 1707ac27a0ecSDave Kleikamp {Opt_err_panic, "errors=panic"}, 1708ac27a0ecSDave Kleikamp {Opt_err_ro, "errors=remount-ro"}, 1709ac27a0ecSDave Kleikamp {Opt_nouid32, "nouid32"}, 1710ac27a0ecSDave Kleikamp {Opt_debug, "debug"}, 171172578c33STheodore Ts'o {Opt_removed, "oldalloc"}, 171272578c33STheodore Ts'o {Opt_removed, "orlov"}, 1713ac27a0ecSDave Kleikamp {Opt_user_xattr, "user_xattr"}, 1714ac27a0ecSDave Kleikamp {Opt_nouser_xattr, "nouser_xattr"}, 1715ac27a0ecSDave Kleikamp {Opt_acl, "acl"}, 1716ac27a0ecSDave Kleikamp {Opt_noacl, "noacl"}, 1717e3bb52aeSEric Sandeen {Opt_noload, "norecovery"}, 17185a916be1STheodore Ts'o {Opt_noload, "noload"}, 171972578c33STheodore Ts'o {Opt_removed, "nobh"}, 172072578c33STheodore Ts'o {Opt_removed, "bh"}, 1721ac27a0ecSDave Kleikamp {Opt_commit, "commit=%u"}, 172230773840STheodore Ts'o {Opt_min_batch_time, "min_batch_time=%u"}, 172330773840STheodore Ts'o {Opt_max_batch_time, "max_batch_time=%u"}, 1724ac27a0ecSDave Kleikamp {Opt_journal_dev, "journal_dev=%u"}, 1725ad4eec61SEric Sandeen {Opt_journal_path, "journal_path=%s"}, 1726818d276cSGirish Shilamkar {Opt_journal_checksum, "journal_checksum"}, 1727c6d3d56dSDarrick J. Wong {Opt_nojournal_checksum, "nojournal_checksum"}, 1728818d276cSGirish Shilamkar {Opt_journal_async_commit, "journal_async_commit"}, 1729ac27a0ecSDave Kleikamp {Opt_abort, "abort"}, 1730ac27a0ecSDave Kleikamp {Opt_data_journal, "data=journal"}, 1731ac27a0ecSDave Kleikamp {Opt_data_ordered, "data=ordered"}, 1732ac27a0ecSDave Kleikamp {Opt_data_writeback, "data=writeback"}, 17335bf5683aSHidehiro Kawai {Opt_data_err_abort, "data_err=abort"}, 17345bf5683aSHidehiro Kawai {Opt_data_err_ignore, "data_err=ignore"}, 1735ac27a0ecSDave Kleikamp {Opt_offusrjquota, "usrjquota="}, 1736ac27a0ecSDave Kleikamp {Opt_usrjquota, "usrjquota=%s"}, 1737ac27a0ecSDave Kleikamp {Opt_offgrpjquota, "grpjquota="}, 1738ac27a0ecSDave Kleikamp {Opt_grpjquota, "grpjquota=%s"}, 1739ac27a0ecSDave Kleikamp {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1740ac27a0ecSDave Kleikamp {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 17415a20bdfcSJan Kara {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1742ac27a0ecSDave Kleikamp {Opt_grpquota, "grpquota"}, 1743ac27a0ecSDave Kleikamp {Opt_noquota, "noquota"}, 1744ac27a0ecSDave Kleikamp {Opt_quota, "quota"}, 1745ac27a0ecSDave Kleikamp {Opt_usrquota, "usrquota"}, 174649da9392SJan Kara {Opt_prjquota, "prjquota"}, 1747ac27a0ecSDave Kleikamp {Opt_barrier, "barrier=%u"}, 174806705bffSTheodore Ts'o {Opt_barrier, "barrier"}, 174906705bffSTheodore Ts'o {Opt_nobarrier, "nobarrier"}, 175025ec56b5SJean Noel Cordenner {Opt_i_version, "i_version"}, 1751923ae0ffSRoss Zwisler {Opt_dax, "dax"}, 17529cb20f94SIra Weiny {Opt_dax_always, "dax=always"}, 17539cb20f94SIra Weiny {Opt_dax_inode, "dax=inode"}, 17549cb20f94SIra Weiny {Opt_dax_never, "dax=never"}, 1755c9de560dSAlex Tomas {Opt_stripe, "stripe=%u"}, 175664769240SAlex Tomas {Opt_delalloc, "delalloc"}, 1757327eaf73STheodore Ts'o {Opt_warn_on_error, "warn_on_error"}, 1758327eaf73STheodore Ts'o {Opt_nowarn_on_error, "nowarn_on_error"}, 1759a26f4992STheodore Ts'o {Opt_lazytime, "lazytime"}, 1760a26f4992STheodore Ts'o {Opt_nolazytime, "nolazytime"}, 1761670e9875STheodore Ts'o {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"}, 1762dd919b98SAneesh Kumar K.V {Opt_nodelalloc, "nodelalloc"}, 176336ade451SJan Kara {Opt_removed, "mblk_io_submit"}, 176436ade451SJan Kara {Opt_removed, "nomblk_io_submit"}, 17656fd058f7STheodore Ts'o {Opt_block_validity, "block_validity"}, 17666fd058f7STheodore Ts'o {Opt_noblock_validity, "noblock_validity"}, 1767240799cdSTheodore Ts'o {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1768b3881f74STheodore Ts'o {Opt_journal_ioprio, "journal_ioprio=%u"}, 1769afd4672dSTheodore Ts'o {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 177006705bffSTheodore Ts'o {Opt_auto_da_alloc, "auto_da_alloc"}, 177106705bffSTheodore Ts'o {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1772744692dcSJiaying Zhang {Opt_dioread_nolock, "dioread_nolock"}, 1773244adf64STheodore Ts'o {Opt_dioread_lock, "nodioread_nolock"}, 1774744692dcSJiaying Zhang {Opt_dioread_lock, "dioread_lock"}, 17755328e635SEric Sandeen {Opt_discard, "discard"}, 17765328e635SEric Sandeen {Opt_nodiscard, "nodiscard"}, 1777fc6cb1cdSTheodore Ts'o {Opt_init_itable, "init_itable=%u"}, 1778fc6cb1cdSTheodore Ts'o {Opt_init_itable, "init_itable"}, 1779fc6cb1cdSTheodore Ts'o {Opt_noinit_itable, "noinit_itable"}, 17808016e29fSHarshad Shirwadkar #ifdef CONFIG_EXT4_DEBUG 178199c880deSHarshad Shirwadkar {Opt_fc_debug_force, "fc_debug_force"}, 17828016e29fSHarshad Shirwadkar {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"}, 17838016e29fSHarshad Shirwadkar #endif 1784df981d03STheodore Ts'o {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 1785ed318a6cSEric Biggers {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 17866ddb2447STheodore Ts'o {Opt_test_dummy_encryption, "test_dummy_encryption"}, 17874f74d15fSEric Biggers {Opt_inlinecrypt, "inlinecrypt"}, 1788cdb7ee4cSTahsin Erdogan {Opt_nombcache, "nombcache"}, 1789cdb7ee4cSTahsin Erdogan {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */ 17903d392b26STheodore Ts'o {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"}, 1791c7198b9cSTheodore Ts'o {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1792c7198b9cSTheodore Ts'o {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1793c7198b9cSTheodore Ts'o {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1794c7198b9cSTheodore Ts'o {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1795c7198b9cSTheodore Ts'o {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1796f3f12faaSJosef Bacik {Opt_err, NULL}, 1797ac27a0ecSDave Kleikamp }; 1798ac27a0ecSDave Kleikamp 1799617ba13bSMingming Cao static ext4_fsblk_t get_sb_block(void **data) 1800ac27a0ecSDave Kleikamp { 1801617ba13bSMingming Cao ext4_fsblk_t sb_block; 1802ac27a0ecSDave Kleikamp char *options = (char *) *data; 1803ac27a0ecSDave Kleikamp 1804ac27a0ecSDave Kleikamp if (!options || strncmp(options, "sb=", 3) != 0) 1805ac27a0ecSDave Kleikamp return 1; /* Default location */ 18060b8e58a1SAndreas Dilger 1807ac27a0ecSDave Kleikamp options += 3; 18080b8e58a1SAndreas Dilger /* TODO: use simple_strtoll with >32bit ext4 */ 1809ac27a0ecSDave Kleikamp sb_block = simple_strtoul(options, &options, 0); 1810ac27a0ecSDave Kleikamp if (*options && *options != ',') { 18114776004fSTheodore Ts'o printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1812ac27a0ecSDave Kleikamp (char *) *data); 1813ac27a0ecSDave Kleikamp return 1; 1814ac27a0ecSDave Kleikamp } 1815ac27a0ecSDave Kleikamp if (*options == ',') 1816ac27a0ecSDave Kleikamp options++; 1817ac27a0ecSDave Kleikamp *data = (void *) options; 18180b8e58a1SAndreas Dilger 1819ac27a0ecSDave Kleikamp return sb_block; 1820ac27a0ecSDave Kleikamp } 1821ac27a0ecSDave Kleikamp 1822b3881f74STheodore Ts'o #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1823d6006186SEric Biggers static const char deprecated_msg[] = 1824d6006186SEric Biggers "Mount option \"%s\" will be removed by %s\n" 1825437ca0fdSDmitry Monakhov "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1826b3881f74STheodore Ts'o 182756c50f11SDmitry Monakhov #ifdef CONFIG_QUOTA 182856c50f11SDmitry Monakhov static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 182956c50f11SDmitry Monakhov { 183056c50f11SDmitry Monakhov struct ext4_sb_info *sbi = EXT4_SB(sb); 183133458eabSTheodore Ts'o char *qname, *old_qname = get_qf_name(sb, sbi, qtype); 183203dafb5fSChen Gang int ret = -1; 183356c50f11SDmitry Monakhov 183433458eabSTheodore Ts'o if (sb_any_quota_loaded(sb) && !old_qname) { 183556c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 183656c50f11SDmitry Monakhov "Cannot change journaled " 183756c50f11SDmitry Monakhov "quota options when quota turned on"); 183857f73c2cSTheodore Ts'o return -1; 183956c50f11SDmitry Monakhov } 1840e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) { 1841c325a67cSTheodore Ts'o ext4_msg(sb, KERN_INFO, "Journaled quota options " 1842c325a67cSTheodore Ts'o "ignored when QUOTA feature is enabled"); 1843c325a67cSTheodore Ts'o return 1; 1844262b4662SJan Kara } 184556c50f11SDmitry Monakhov qname = match_strdup(args); 184656c50f11SDmitry Monakhov if (!qname) { 184756c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 184856c50f11SDmitry Monakhov "Not enough memory for storing quotafile name"); 184957f73c2cSTheodore Ts'o return -1; 185056c50f11SDmitry Monakhov } 185133458eabSTheodore Ts'o if (old_qname) { 185233458eabSTheodore Ts'o if (strcmp(old_qname, qname) == 0) 185303dafb5fSChen Gang ret = 1; 185403dafb5fSChen Gang else 185556c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 185603dafb5fSChen Gang "%s quota file already specified", 185703dafb5fSChen Gang QTYPE2NAME(qtype)); 185803dafb5fSChen Gang goto errout; 185956c50f11SDmitry Monakhov } 186003dafb5fSChen Gang if (strchr(qname, '/')) { 186156c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 186256c50f11SDmitry Monakhov "quotafile must be on filesystem root"); 186303dafb5fSChen Gang goto errout; 186456c50f11SDmitry Monakhov } 186533458eabSTheodore Ts'o rcu_assign_pointer(sbi->s_qf_names[qtype], qname); 1866fd8c37ecSTheodore Ts'o set_opt(sb, QUOTA); 186756c50f11SDmitry Monakhov return 1; 186803dafb5fSChen Gang errout: 186903dafb5fSChen Gang kfree(qname); 187003dafb5fSChen Gang return ret; 187156c50f11SDmitry Monakhov } 187256c50f11SDmitry Monakhov 187356c50f11SDmitry Monakhov static int clear_qf_name(struct super_block *sb, int qtype) 187456c50f11SDmitry Monakhov { 187556c50f11SDmitry Monakhov 187656c50f11SDmitry Monakhov struct ext4_sb_info *sbi = EXT4_SB(sb); 187733458eabSTheodore Ts'o char *old_qname = get_qf_name(sb, sbi, qtype); 187856c50f11SDmitry Monakhov 187933458eabSTheodore Ts'o if (sb_any_quota_loaded(sb) && old_qname) { 188056c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 188156c50f11SDmitry Monakhov " when quota turned on"); 188257f73c2cSTheodore Ts'o return -1; 188356c50f11SDmitry Monakhov } 188433458eabSTheodore Ts'o rcu_assign_pointer(sbi->s_qf_names[qtype], NULL); 188533458eabSTheodore Ts'o synchronize_rcu(); 188633458eabSTheodore Ts'o kfree(old_qname); 188756c50f11SDmitry Monakhov return 1; 188856c50f11SDmitry Monakhov } 188956c50f11SDmitry Monakhov #endif 189056c50f11SDmitry Monakhov 189126092bf5STheodore Ts'o #define MOPT_SET 0x0001 189226092bf5STheodore Ts'o #define MOPT_CLEAR 0x0002 189326092bf5STheodore Ts'o #define MOPT_NOSUPPORT 0x0004 189426092bf5STheodore Ts'o #define MOPT_EXPLICIT 0x0008 189526092bf5STheodore Ts'o #define MOPT_CLEAR_ERR 0x0010 189626092bf5STheodore Ts'o #define MOPT_GTE0 0x0020 189726092bf5STheodore Ts'o #ifdef CONFIG_QUOTA 189826092bf5STheodore Ts'o #define MOPT_Q 0 189926092bf5STheodore Ts'o #define MOPT_QFMT 0x0040 190026092bf5STheodore Ts'o #else 190126092bf5STheodore Ts'o #define MOPT_Q MOPT_NOSUPPORT 190226092bf5STheodore Ts'o #define MOPT_QFMT MOPT_NOSUPPORT 190326092bf5STheodore Ts'o #endif 190426092bf5STheodore Ts'o #define MOPT_DATAJ 0x0080 19058dc0aa8cSTheodore Ts'o #define MOPT_NO_EXT2 0x0100 19068dc0aa8cSTheodore Ts'o #define MOPT_NO_EXT3 0x0200 19078dc0aa8cSTheodore Ts'o #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1908ad4eec61SEric Sandeen #define MOPT_STRING 0x0400 19099cb20f94SIra Weiny #define MOPT_SKIP 0x0800 1910995a3ed6SHarshad Shirwadkar #define MOPT_2 0x1000 191126092bf5STheodore Ts'o 191226092bf5STheodore Ts'o static const struct mount_opts { 191326092bf5STheodore Ts'o int token; 191426092bf5STheodore Ts'o int mount_opt; 191526092bf5STheodore Ts'o int flags; 191626092bf5STheodore Ts'o } ext4_mount_opts[] = { 191726092bf5STheodore Ts'o {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 191826092bf5STheodore Ts'o {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 191926092bf5STheodore Ts'o {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 192026092bf5STheodore Ts'o {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 192126092bf5STheodore Ts'o {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 192226092bf5STheodore Ts'o {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 19238dc0aa8cSTheodore Ts'o {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 19248dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_SET}, 19258dc0aa8cSTheodore Ts'o {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 19268dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_CLEAR}, 192726092bf5STheodore Ts'o {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 192826092bf5STheodore Ts'o {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 19298dc0aa8cSTheodore Ts'o {Opt_delalloc, EXT4_MOUNT_DELALLOC, 19308dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 19318dc0aa8cSTheodore Ts'o {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 193259d9fa5cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_CLEAR}, 1933327eaf73STheodore Ts'o {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1934327eaf73STheodore Ts'o {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1935c6d3d56dSDarrick J. Wong {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1936c6d3d56dSDarrick J. Wong MOPT_EXT4_ONLY | MOPT_CLEAR}, 19378dc0aa8cSTheodore Ts'o {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 19381e381f60SDmitry Monakhov MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 193926092bf5STheodore Ts'o {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 19408dc0aa8cSTheodore Ts'o EXT4_MOUNT_JOURNAL_CHECKSUM), 19411e381f60SDmitry Monakhov MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 19428dc0aa8cSTheodore Ts'o {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 194326092bf5STheodore Ts'o {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 194426092bf5STheodore Ts'o {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 194526092bf5STheodore Ts'o {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 19468dc0aa8cSTheodore Ts'o {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 19477915a861SAles Novak MOPT_NO_EXT2}, 19488dc0aa8cSTheodore Ts'o {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 19497915a861SAles Novak MOPT_NO_EXT2}, 195026092bf5STheodore Ts'o {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 195126092bf5STheodore Ts'o {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 195226092bf5STheodore Ts'o {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 195326092bf5STheodore Ts'o {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 195426092bf5STheodore Ts'o {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 195526092bf5STheodore Ts'o {Opt_commit, 0, MOPT_GTE0}, 195626092bf5STheodore Ts'o {Opt_max_batch_time, 0, MOPT_GTE0}, 195726092bf5STheodore Ts'o {Opt_min_batch_time, 0, MOPT_GTE0}, 195826092bf5STheodore Ts'o {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 195926092bf5STheodore Ts'o {Opt_init_itable, 0, MOPT_GTE0}, 19609cb20f94SIra Weiny {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP}, 19619cb20f94SIra Weiny {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS, 19629cb20f94SIra Weiny MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 19639cb20f94SIra Weiny {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE, 19649cb20f94SIra Weiny MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 19659cb20f94SIra Weiny {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER, 19669cb20f94SIra Weiny MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 196726092bf5STheodore Ts'o {Opt_stripe, 0, MOPT_GTE0}, 19680efb3b23SJan Kara {Opt_resuid, 0, MOPT_GTE0}, 19690efb3b23SJan Kara {Opt_resgid, 0, MOPT_GTE0}, 19705ba92bcfSCarlos Maiolino {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 19715ba92bcfSCarlos Maiolino {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING}, 19725ba92bcfSCarlos Maiolino {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 19738dc0aa8cSTheodore Ts'o {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 19748dc0aa8cSTheodore Ts'o {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 19758dc0aa8cSTheodore Ts'o {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 19768dc0aa8cSTheodore Ts'o MOPT_NO_EXT2 | MOPT_DATAJ}, 197726092bf5STheodore Ts'o {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 197826092bf5STheodore Ts'o {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 197926092bf5STheodore Ts'o #ifdef CONFIG_EXT4_FS_POSIX_ACL 198026092bf5STheodore Ts'o {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 198126092bf5STheodore Ts'o {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 198226092bf5STheodore Ts'o #else 198326092bf5STheodore Ts'o {Opt_acl, 0, MOPT_NOSUPPORT}, 198426092bf5STheodore Ts'o {Opt_noacl, 0, MOPT_NOSUPPORT}, 198526092bf5STheodore Ts'o #endif 198626092bf5STheodore Ts'o {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 198726092bf5STheodore Ts'o {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1988670e9875STheodore Ts'o {Opt_debug_want_extra_isize, 0, MOPT_GTE0}, 198926092bf5STheodore Ts'o {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 199026092bf5STheodore Ts'o {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 199126092bf5STheodore Ts'o MOPT_SET | MOPT_Q}, 199226092bf5STheodore Ts'o {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 199326092bf5STheodore Ts'o MOPT_SET | MOPT_Q}, 199449da9392SJan Kara {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 199549da9392SJan Kara MOPT_SET | MOPT_Q}, 199626092bf5STheodore Ts'o {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 199749da9392SJan Kara EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 199849da9392SJan Kara MOPT_CLEAR | MOPT_Q}, 1999174fe5baSKaixu Xia {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING}, 2000174fe5baSKaixu Xia {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING}, 200126092bf5STheodore Ts'o {Opt_offusrjquota, 0, MOPT_Q}, 200226092bf5STheodore Ts'o {Opt_offgrpjquota, 0, MOPT_Q}, 200326092bf5STheodore Ts'o {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 200426092bf5STheodore Ts'o {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 200526092bf5STheodore Ts'o {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 2006df981d03STheodore Ts'o {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 2007ed318a6cSEric Biggers {Opt_test_dummy_encryption, 0, MOPT_STRING}, 2008cdb7ee4cSTahsin Erdogan {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 20093d392b26STheodore Ts'o {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS, 20103d392b26STheodore Ts'o MOPT_SET}, 201199c880deSHarshad Shirwadkar #ifdef CONFIG_EXT4_DEBUG 20120f0672ffSHarshad Shirwadkar {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT, 20130f0672ffSHarshad Shirwadkar MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY}, 20148016e29fSHarshad Shirwadkar {Opt_fc_debug_max_replay, 0, MOPT_GTE0}, 20158016e29fSHarshad Shirwadkar #endif 201626092bf5STheodore Ts'o {Opt_err, 0, 0} 201726092bf5STheodore Ts'o }; 201826092bf5STheodore Ts'o 2019c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 2020c83ad55eSGabriel Krisman Bertazi static const struct ext4_sb_encodings { 2021c83ad55eSGabriel Krisman Bertazi __u16 magic; 2022c83ad55eSGabriel Krisman Bertazi char *name; 2023c83ad55eSGabriel Krisman Bertazi char *version; 2024c83ad55eSGabriel Krisman Bertazi } ext4_sb_encoding_map[] = { 2025c83ad55eSGabriel Krisman Bertazi {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"}, 2026c83ad55eSGabriel Krisman Bertazi }; 2027c83ad55eSGabriel Krisman Bertazi 2028c83ad55eSGabriel Krisman Bertazi static int ext4_sb_read_encoding(const struct ext4_super_block *es, 2029c83ad55eSGabriel Krisman Bertazi const struct ext4_sb_encodings **encoding, 2030c83ad55eSGabriel Krisman Bertazi __u16 *flags) 2031c83ad55eSGabriel Krisman Bertazi { 2032c83ad55eSGabriel Krisman Bertazi __u16 magic = le16_to_cpu(es->s_encoding); 2033c83ad55eSGabriel Krisman Bertazi int i; 2034c83ad55eSGabriel Krisman Bertazi 2035c83ad55eSGabriel Krisman Bertazi for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) 2036c83ad55eSGabriel Krisman Bertazi if (magic == ext4_sb_encoding_map[i].magic) 2037c83ad55eSGabriel Krisman Bertazi break; 2038c83ad55eSGabriel Krisman Bertazi 2039c83ad55eSGabriel Krisman Bertazi if (i >= ARRAY_SIZE(ext4_sb_encoding_map)) 2040c83ad55eSGabriel Krisman Bertazi return -EINVAL; 2041c83ad55eSGabriel Krisman Bertazi 2042c83ad55eSGabriel Krisman Bertazi *encoding = &ext4_sb_encoding_map[i]; 2043c83ad55eSGabriel Krisman Bertazi *flags = le16_to_cpu(es->s_encoding_flags); 2044c83ad55eSGabriel Krisman Bertazi 2045c83ad55eSGabriel Krisman Bertazi return 0; 2046c83ad55eSGabriel Krisman Bertazi } 2047c83ad55eSGabriel Krisman Bertazi #endif 2048c83ad55eSGabriel Krisman Bertazi 2049ed318a6cSEric Biggers static int ext4_set_test_dummy_encryption(struct super_block *sb, 2050ed318a6cSEric Biggers const char *opt, 2051ed318a6cSEric Biggers const substring_t *arg, 2052ed318a6cSEric Biggers bool is_remount) 2053ed318a6cSEric Biggers { 2054ed318a6cSEric Biggers #ifdef CONFIG_FS_ENCRYPTION 2055ed318a6cSEric Biggers struct ext4_sb_info *sbi = EXT4_SB(sb); 2056ed318a6cSEric Biggers int err; 2057ed318a6cSEric Biggers 2058ed318a6cSEric Biggers /* 2059ed318a6cSEric Biggers * This mount option is just for testing, and it's not worthwhile to 2060ed318a6cSEric Biggers * implement the extra complexity (e.g. RCU protection) that would be 2061ed318a6cSEric Biggers * needed to allow it to be set or changed during remount. We do allow 2062ed318a6cSEric Biggers * it to be specified during remount, but only if there is no change. 2063ed318a6cSEric Biggers */ 2064ac4acb1fSEric Biggers if (is_remount && !sbi->s_dummy_enc_policy.policy) { 2065ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 2066ed318a6cSEric Biggers "Can't set test_dummy_encryption on remount"); 2067ed318a6cSEric Biggers return -1; 2068ed318a6cSEric Biggers } 2069c8c868abSEric Biggers err = fscrypt_set_test_dummy_encryption(sb, arg->from, 2070ac4acb1fSEric Biggers &sbi->s_dummy_enc_policy); 2071ed318a6cSEric Biggers if (err) { 2072ed318a6cSEric Biggers if (err == -EEXIST) 2073ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 2074ed318a6cSEric Biggers "Can't change test_dummy_encryption on remount"); 2075ed318a6cSEric Biggers else if (err == -EINVAL) 2076ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 2077ed318a6cSEric Biggers "Value of option \"%s\" is unrecognized", opt); 2078ed318a6cSEric Biggers else 2079ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 2080ed318a6cSEric Biggers "Error processing option \"%s\" [%d]", 2081ed318a6cSEric Biggers opt, err); 2082ed318a6cSEric Biggers return -1; 2083ed318a6cSEric Biggers } 2084ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled"); 2085ed318a6cSEric Biggers #else 2086ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 2087ed318a6cSEric Biggers "Test dummy encryption mount option ignored"); 2088ed318a6cSEric Biggers #endif 2089ed318a6cSEric Biggers return 1; 2090ed318a6cSEric Biggers } 2091ed318a6cSEric Biggers 209226092bf5STheodore Ts'o static int handle_mount_opt(struct super_block *sb, char *opt, int token, 209326092bf5STheodore Ts'o substring_t *args, unsigned long *journal_devnum, 209426092bf5STheodore Ts'o unsigned int *journal_ioprio, int is_remount) 209526092bf5STheodore Ts'o { 209626092bf5STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 209726092bf5STheodore Ts'o const struct mount_opts *m; 209808cefc7aSEric W. Biederman kuid_t uid; 209908cefc7aSEric W. Biederman kgid_t gid; 210026092bf5STheodore Ts'o int arg = 0; 210126092bf5STheodore Ts'o 210257f73c2cSTheodore Ts'o #ifdef CONFIG_QUOTA 210357f73c2cSTheodore Ts'o if (token == Opt_usrjquota) 210457f73c2cSTheodore Ts'o return set_qf_name(sb, USRQUOTA, &args[0]); 210557f73c2cSTheodore Ts'o else if (token == Opt_grpjquota) 210657f73c2cSTheodore Ts'o return set_qf_name(sb, GRPQUOTA, &args[0]); 210757f73c2cSTheodore Ts'o else if (token == Opt_offusrjquota) 210857f73c2cSTheodore Ts'o return clear_qf_name(sb, USRQUOTA); 210957f73c2cSTheodore Ts'o else if (token == Opt_offgrpjquota) 211057f73c2cSTheodore Ts'o return clear_qf_name(sb, GRPQUOTA); 211157f73c2cSTheodore Ts'o #endif 211226092bf5STheodore Ts'o switch (token) { 2113f7048605STheodore Ts'o case Opt_noacl: 2114f7048605STheodore Ts'o case Opt_nouser_xattr: 2115f7048605STheodore Ts'o ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 2116f7048605STheodore Ts'o break; 211726092bf5STheodore Ts'o case Opt_sb: 211826092bf5STheodore Ts'o return 1; /* handled by get_sb_block() */ 211926092bf5STheodore Ts'o case Opt_removed: 21205f3633e3SJan Kara ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 212126092bf5STheodore Ts'o return 1; 212226092bf5STheodore Ts'o case Opt_abort: 21239b5f6c9bSHarshad Shirwadkar ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED); 212426092bf5STheodore Ts'o return 1; 212526092bf5STheodore Ts'o case Opt_i_version: 2126357fdad0SMatthew Garrett sb->s_flags |= SB_I_VERSION; 212726092bf5STheodore Ts'o return 1; 2128a26f4992STheodore Ts'o case Opt_lazytime: 21291751e8a6SLinus Torvalds sb->s_flags |= SB_LAZYTIME; 2130a26f4992STheodore Ts'o return 1; 2131a26f4992STheodore Ts'o case Opt_nolazytime: 21321751e8a6SLinus Torvalds sb->s_flags &= ~SB_LAZYTIME; 2133a26f4992STheodore Ts'o return 1; 21344f74d15fSEric Biggers case Opt_inlinecrypt: 21354f74d15fSEric Biggers #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 21364f74d15fSEric Biggers sb->s_flags |= SB_INLINECRYPT; 21374f74d15fSEric Biggers #else 21384f74d15fSEric Biggers ext4_msg(sb, KERN_ERR, "inline encryption not supported"); 21394f74d15fSEric Biggers #endif 21404f74d15fSEric Biggers return 1; 214126092bf5STheodore Ts'o } 214226092bf5STheodore Ts'o 21435f3633e3SJan Kara for (m = ext4_mount_opts; m->token != Opt_err; m++) 21445f3633e3SJan Kara if (token == m->token) 21455f3633e3SJan Kara break; 21465f3633e3SJan Kara 21475f3633e3SJan Kara if (m->token == Opt_err) { 21485f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 21495f3633e3SJan Kara "or missing value", opt); 21505f3633e3SJan Kara return -1; 21515f3633e3SJan Kara } 21525f3633e3SJan Kara 21538dc0aa8cSTheodore Ts'o if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 21548dc0aa8cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 21558dc0aa8cSTheodore Ts'o "Mount option \"%s\" incompatible with ext2", opt); 21568dc0aa8cSTheodore Ts'o return -1; 21578dc0aa8cSTheodore Ts'o } 21588dc0aa8cSTheodore Ts'o if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 21598dc0aa8cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 21608dc0aa8cSTheodore Ts'o "Mount option \"%s\" incompatible with ext3", opt); 21618dc0aa8cSTheodore Ts'o return -1; 21628dc0aa8cSTheodore Ts'o } 21638dc0aa8cSTheodore Ts'o 2164ad4eec61SEric Sandeen if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) 21650efb3b23SJan Kara return -1; 216626092bf5STheodore Ts'o if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 216726092bf5STheodore Ts'o return -1; 2168c93cf2d7SDmitry Monakhov if (m->flags & MOPT_EXPLICIT) { 2169c93cf2d7SDmitry Monakhov if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 217026092bf5STheodore Ts'o set_opt2(sb, EXPLICIT_DELALLOC); 21711e381f60SDmitry Monakhov } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 21721e381f60SDmitry Monakhov set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM); 2173c93cf2d7SDmitry Monakhov } else 2174c93cf2d7SDmitry Monakhov return -1; 2175c93cf2d7SDmitry Monakhov } 217626092bf5STheodore Ts'o if (m->flags & MOPT_CLEAR_ERR) 217726092bf5STheodore Ts'o clear_opt(sb, ERRORS_MASK); 217826092bf5STheodore Ts'o if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 217926092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "Cannot change quota " 218026092bf5STheodore Ts'o "options when quota turned on"); 218126092bf5STheodore Ts'o return -1; 218226092bf5STheodore Ts'o } 218326092bf5STheodore Ts'o 218426092bf5STheodore Ts'o if (m->flags & MOPT_NOSUPPORT) { 218526092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 218626092bf5STheodore Ts'o } else if (token == Opt_commit) { 218726092bf5STheodore Ts'o if (arg == 0) 218826092bf5STheodore Ts'o arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 21899ba55543Szhangyi (F) else if (arg > INT_MAX / HZ) { 21909ba55543Szhangyi (F) ext4_msg(sb, KERN_ERR, 21919ba55543Szhangyi (F) "Invalid commit interval %d, " 21929ba55543Szhangyi (F) "must be smaller than %d", 21939ba55543Szhangyi (F) arg, INT_MAX / HZ); 21949ba55543Szhangyi (F) return -1; 21959ba55543Szhangyi (F) } 219626092bf5STheodore Ts'o sbi->s_commit_interval = HZ * arg; 2197670e9875STheodore Ts'o } else if (token == Opt_debug_want_extra_isize) { 21989803387cSTheodore Ts'o if ((arg & 1) || 21999803387cSTheodore Ts'o (arg < 4) || 22009803387cSTheodore Ts'o (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) { 22019803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 22029803387cSTheodore Ts'o "Invalid want_extra_isize %d", arg); 22039803387cSTheodore Ts'o return -1; 22049803387cSTheodore Ts'o } 2205670e9875STheodore Ts'o sbi->s_want_extra_isize = arg; 220626092bf5STheodore Ts'o } else if (token == Opt_max_batch_time) { 220726092bf5STheodore Ts'o sbi->s_max_batch_time = arg; 220826092bf5STheodore Ts'o } else if (token == Opt_min_batch_time) { 220926092bf5STheodore Ts'o sbi->s_min_batch_time = arg; 221026092bf5STheodore Ts'o } else if (token == Opt_inode_readahead_blks) { 2211e33e60eaSJan Kara if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 2212e33e60eaSJan Kara ext4_msg(sb, KERN_ERR, 2213e33e60eaSJan Kara "EXT4-fs: inode_readahead_blks must be " 2214e33e60eaSJan Kara "0 or a power of 2 smaller than 2^31"); 221526092bf5STheodore Ts'o return -1; 221626092bf5STheodore Ts'o } 221726092bf5STheodore Ts'o sbi->s_inode_readahead_blks = arg; 221826092bf5STheodore Ts'o } else if (token == Opt_init_itable) { 221926092bf5STheodore Ts'o set_opt(sb, INIT_INODE_TABLE); 222026092bf5STheodore Ts'o if (!args->from) 222126092bf5STheodore Ts'o arg = EXT4_DEF_LI_WAIT_MULT; 222226092bf5STheodore Ts'o sbi->s_li_wait_mult = arg; 2223df981d03STheodore Ts'o } else if (token == Opt_max_dir_size_kb) { 2224df981d03STheodore Ts'o sbi->s_max_dir_size_kb = arg; 22258016e29fSHarshad Shirwadkar #ifdef CONFIG_EXT4_DEBUG 22268016e29fSHarshad Shirwadkar } else if (token == Opt_fc_debug_max_replay) { 22278016e29fSHarshad Shirwadkar sbi->s_fc_debug_max_replay = arg; 22288016e29fSHarshad Shirwadkar #endif 222926092bf5STheodore Ts'o } else if (token == Opt_stripe) { 223026092bf5STheodore Ts'o sbi->s_stripe = arg; 22310efb3b23SJan Kara } else if (token == Opt_resuid) { 22320efb3b23SJan Kara uid = make_kuid(current_user_ns(), arg); 22330efb3b23SJan Kara if (!uid_valid(uid)) { 22345f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 22350efb3b23SJan Kara return -1; 22360efb3b23SJan Kara } 22370efb3b23SJan Kara sbi->s_resuid = uid; 22380efb3b23SJan Kara } else if (token == Opt_resgid) { 22390efb3b23SJan Kara gid = make_kgid(current_user_ns(), arg); 22400efb3b23SJan Kara if (!gid_valid(gid)) { 22415f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 22420efb3b23SJan Kara return -1; 22430efb3b23SJan Kara } 22440efb3b23SJan Kara sbi->s_resgid = gid; 22450efb3b23SJan Kara } else if (token == Opt_journal_dev) { 22460efb3b23SJan Kara if (is_remount) { 22470efb3b23SJan Kara ext4_msg(sb, KERN_ERR, 22480efb3b23SJan Kara "Cannot specify journal on remount"); 22490efb3b23SJan Kara return -1; 22500efb3b23SJan Kara } 22510efb3b23SJan Kara *journal_devnum = arg; 2252ad4eec61SEric Sandeen } else if (token == Opt_journal_path) { 2253ad4eec61SEric Sandeen char *journal_path; 2254ad4eec61SEric Sandeen struct inode *journal_inode; 2255ad4eec61SEric Sandeen struct path path; 2256ad4eec61SEric Sandeen int error; 2257ad4eec61SEric Sandeen 2258ad4eec61SEric Sandeen if (is_remount) { 2259ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, 2260ad4eec61SEric Sandeen "Cannot specify journal on remount"); 2261ad4eec61SEric Sandeen return -1; 2262ad4eec61SEric Sandeen } 2263ad4eec61SEric Sandeen journal_path = match_strdup(&args[0]); 2264ad4eec61SEric Sandeen if (!journal_path) { 2265ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: could not dup " 2266ad4eec61SEric Sandeen "journal device string"); 2267ad4eec61SEric Sandeen return -1; 2268ad4eec61SEric Sandeen } 2269ad4eec61SEric Sandeen 2270ad4eec61SEric Sandeen error = kern_path(journal_path, LOOKUP_FOLLOW, &path); 2271ad4eec61SEric Sandeen if (error) { 2272ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: could not find " 2273ad4eec61SEric Sandeen "journal device path: error %d", error); 2274ad4eec61SEric Sandeen kfree(journal_path); 2275ad4eec61SEric Sandeen return -1; 2276ad4eec61SEric Sandeen } 2277ad4eec61SEric Sandeen 22782b0143b5SDavid Howells journal_inode = d_inode(path.dentry); 2279ad4eec61SEric Sandeen if (!S_ISBLK(journal_inode->i_mode)) { 2280ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: journal path %s " 2281ad4eec61SEric Sandeen "is not a block device", journal_path); 2282ad4eec61SEric Sandeen path_put(&path); 2283ad4eec61SEric Sandeen kfree(journal_path); 2284ad4eec61SEric Sandeen return -1; 2285ad4eec61SEric Sandeen } 2286ad4eec61SEric Sandeen 2287ad4eec61SEric Sandeen *journal_devnum = new_encode_dev(journal_inode->i_rdev); 2288ad4eec61SEric Sandeen path_put(&path); 2289ad4eec61SEric Sandeen kfree(journal_path); 22900efb3b23SJan Kara } else if (token == Opt_journal_ioprio) { 22910efb3b23SJan Kara if (arg > 7) { 22925f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 22930efb3b23SJan Kara " (must be 0-7)"); 22940efb3b23SJan Kara return -1; 22950efb3b23SJan Kara } 22960efb3b23SJan Kara *journal_ioprio = 22970efb3b23SJan Kara IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 22986ddb2447STheodore Ts'o } else if (token == Opt_test_dummy_encryption) { 2299ed318a6cSEric Biggers return ext4_set_test_dummy_encryption(sb, opt, &args[0], 2300ed318a6cSEric Biggers is_remount); 230126092bf5STheodore Ts'o } else if (m->flags & MOPT_DATAJ) { 230226092bf5STheodore Ts'o if (is_remount) { 230326092bf5STheodore Ts'o if (!sbi->s_journal) 230426092bf5STheodore Ts'o ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 23055f3633e3SJan Kara else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 230626092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, 230726092bf5STheodore Ts'o "Cannot change data mode on remount"); 230826092bf5STheodore Ts'o return -1; 230926092bf5STheodore Ts'o } 231026092bf5STheodore Ts'o } else { 231126092bf5STheodore Ts'o clear_opt(sb, DATA_FLAGS); 231226092bf5STheodore Ts'o sbi->s_mount_opt |= m->mount_opt; 231326092bf5STheodore Ts'o } 231426092bf5STheodore Ts'o #ifdef CONFIG_QUOTA 231526092bf5STheodore Ts'o } else if (m->flags & MOPT_QFMT) { 231626092bf5STheodore Ts'o if (sb_any_quota_loaded(sb) && 231726092bf5STheodore Ts'o sbi->s_jquota_fmt != m->mount_opt) { 23185f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Cannot change journaled " 23195f3633e3SJan Kara "quota options when quota turned on"); 232026092bf5STheodore Ts'o return -1; 232126092bf5STheodore Ts'o } 2322e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) { 2323c325a67cSTheodore Ts'o ext4_msg(sb, KERN_INFO, 2324c325a67cSTheodore Ts'o "Quota format mount options ignored " 2325262b4662SJan Kara "when QUOTA feature is enabled"); 2326c325a67cSTheodore Ts'o return 1; 2327262b4662SJan Kara } 232826092bf5STheodore Ts'o sbi->s_jquota_fmt = m->mount_opt; 232926092bf5STheodore Ts'o #endif 23309cb20f94SIra Weiny } else if (token == Opt_dax || token == Opt_dax_always || 23319cb20f94SIra Weiny token == Opt_dax_inode || token == Opt_dax_never) { 2332ef83b6e8SDan Williams #ifdef CONFIG_FS_DAX 23339cb20f94SIra Weiny switch (token) { 23349cb20f94SIra Weiny case Opt_dax: 23359cb20f94SIra Weiny case Opt_dax_always: 2336829b37b8STheodore Ts'o if (is_remount && 2337829b37b8STheodore Ts'o (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2338829b37b8STheodore Ts'o (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) { 2339829b37b8STheodore Ts'o fail_dax_change_remount: 2340829b37b8STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't change " 2341829b37b8STheodore Ts'o "dax mount option while remounting"); 2342829b37b8STheodore Ts'o return -1; 2343829b37b8STheodore Ts'o } 2344829b37b8STheodore Ts'o if (is_remount && 2345829b37b8STheodore Ts'o (test_opt(sb, DATA_FLAGS) == 2346829b37b8STheodore Ts'o EXT4_MOUNT_JOURNAL_DATA)) { 2347829b37b8STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't mount with " 2348829b37b8STheodore Ts'o "both data=journal and dax"); 2349829b37b8STheodore Ts'o return -1; 2350829b37b8STheodore Ts'o } 2351ef83b6e8SDan Williams ext4_msg(sb, KERN_WARNING, 2352ef83b6e8SDan Williams "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); 23539cb20f94SIra Weiny sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS; 23549cb20f94SIra Weiny sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER; 23559cb20f94SIra Weiny break; 23569cb20f94SIra Weiny case Opt_dax_never: 2357829b37b8STheodore Ts'o if (is_remount && 2358829b37b8STheodore Ts'o (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2359829b37b8STheodore Ts'o (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) 2360829b37b8STheodore Ts'o goto fail_dax_change_remount; 23619cb20f94SIra Weiny sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER; 23629cb20f94SIra Weiny sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 23639cb20f94SIra Weiny break; 23649cb20f94SIra Weiny case Opt_dax_inode: 2365829b37b8STheodore Ts'o if (is_remount && 2366829b37b8STheodore Ts'o ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2367829b37b8STheodore Ts'o (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2368829b37b8STheodore Ts'o !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) 2369829b37b8STheodore Ts'o goto fail_dax_change_remount; 23709cb20f94SIra Weiny sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 23719cb20f94SIra Weiny sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER; 23729cb20f94SIra Weiny /* Strictly for printing options */ 23739cb20f94SIra Weiny sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE; 23749cb20f94SIra Weiny break; 23759cb20f94SIra Weiny } 2376ef83b6e8SDan Williams #else 2377923ae0ffSRoss Zwisler ext4_msg(sb, KERN_INFO, "dax option not supported"); 23789cb20f94SIra Weiny sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER; 23799cb20f94SIra Weiny sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 2380923ae0ffSRoss Zwisler return -1; 2381923ae0ffSRoss Zwisler #endif 23827915a861SAles Novak } else if (token == Opt_data_err_abort) { 23837915a861SAles Novak sbi->s_mount_opt |= m->mount_opt; 23847915a861SAles Novak } else if (token == Opt_data_err_ignore) { 23857915a861SAles Novak sbi->s_mount_opt &= ~m->mount_opt; 238626092bf5STheodore Ts'o } else { 238726092bf5STheodore Ts'o if (!args->from) 238826092bf5STheodore Ts'o arg = 1; 238926092bf5STheodore Ts'o if (m->flags & MOPT_CLEAR) 239026092bf5STheodore Ts'o arg = !arg; 239126092bf5STheodore Ts'o else if (unlikely(!(m->flags & MOPT_SET))) { 239226092bf5STheodore Ts'o ext4_msg(sb, KERN_WARNING, 239326092bf5STheodore Ts'o "buggy handling of option %s", opt); 239426092bf5STheodore Ts'o WARN_ON(1); 239526092bf5STheodore Ts'o return -1; 239626092bf5STheodore Ts'o } 2397995a3ed6SHarshad Shirwadkar if (m->flags & MOPT_2) { 2398995a3ed6SHarshad Shirwadkar if (arg != 0) 2399995a3ed6SHarshad Shirwadkar sbi->s_mount_opt2 |= m->mount_opt; 2400995a3ed6SHarshad Shirwadkar else 2401995a3ed6SHarshad Shirwadkar sbi->s_mount_opt2 &= ~m->mount_opt; 2402995a3ed6SHarshad Shirwadkar } else { 240326092bf5STheodore Ts'o if (arg != 0) 240426092bf5STheodore Ts'o sbi->s_mount_opt |= m->mount_opt; 240526092bf5STheodore Ts'o else 240626092bf5STheodore Ts'o sbi->s_mount_opt &= ~m->mount_opt; 240726092bf5STheodore Ts'o } 2408995a3ed6SHarshad Shirwadkar } 240926092bf5STheodore Ts'o return 1; 241026092bf5STheodore Ts'o } 241126092bf5STheodore Ts'o 2412ac27a0ecSDave Kleikamp static int parse_options(char *options, struct super_block *sb, 2413c3191067STheodore Ts'o unsigned long *journal_devnum, 2414b3881f74STheodore Ts'o unsigned int *journal_ioprio, 2415661aa520SEric Sandeen int is_remount) 2416ac27a0ecSDave Kleikamp { 24171e1a76edSOlof Johansson struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb); 241833458eabSTheodore Ts'o char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name; 2419ac27a0ecSDave Kleikamp substring_t args[MAX_OPT_ARGS]; 242026092bf5STheodore Ts'o int token; 2421ac27a0ecSDave Kleikamp 2422ac27a0ecSDave Kleikamp if (!options) 2423ac27a0ecSDave Kleikamp return 1; 2424ac27a0ecSDave Kleikamp 2425ac27a0ecSDave Kleikamp while ((p = strsep(&options, ",")) != NULL) { 2426ac27a0ecSDave Kleikamp if (!*p) 2427ac27a0ecSDave Kleikamp continue; 242815121c18SEric Sandeen /* 242915121c18SEric Sandeen * Initialize args struct so we know whether arg was 243015121c18SEric Sandeen * found; some options take optional arguments. 243115121c18SEric Sandeen */ 2432caecd0afSSachin Kamat args[0].to = args[0].from = NULL; 2433ac27a0ecSDave Kleikamp token = match_token(p, tokens, args); 243426092bf5STheodore Ts'o if (handle_mount_opt(sb, p, token, args, journal_devnum, 243526092bf5STheodore Ts'o journal_ioprio, is_remount) < 0) 2436ac27a0ecSDave Kleikamp return 0; 2437ac27a0ecSDave Kleikamp } 2438ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 243949da9392SJan Kara /* 244049da9392SJan Kara * We do the test below only for project quotas. 'usrquota' and 244149da9392SJan Kara * 'grpquota' mount options are allowed even without quota feature 244249da9392SJan Kara * to support legacy quotas in quota files. 244349da9392SJan Kara */ 244449da9392SJan Kara if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) { 244549da9392SJan Kara ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. " 244649da9392SJan Kara "Cannot enable project quota enforcement."); 244749da9392SJan Kara return 0; 244849da9392SJan Kara } 244933458eabSTheodore Ts'o usr_qf_name = get_qf_name(sb, sbi, USRQUOTA); 245033458eabSTheodore Ts'o grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA); 245133458eabSTheodore Ts'o if (usr_qf_name || grp_qf_name) { 245233458eabSTheodore Ts'o if (test_opt(sb, USRQUOTA) && usr_qf_name) 2453fd8c37ecSTheodore Ts'o clear_opt(sb, USRQUOTA); 2454ac27a0ecSDave Kleikamp 245533458eabSTheodore Ts'o if (test_opt(sb, GRPQUOTA) && grp_qf_name) 2456fd8c37ecSTheodore Ts'o clear_opt(sb, GRPQUOTA); 2457ac27a0ecSDave Kleikamp 245856c50f11SDmitry Monakhov if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 2459b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "old and new quota " 2460b31e1552SEric Sandeen "format mixing"); 2461ac27a0ecSDave Kleikamp return 0; 2462ac27a0ecSDave Kleikamp } 2463ac27a0ecSDave Kleikamp 2464ac27a0ecSDave Kleikamp if (!sbi->s_jquota_fmt) { 2465b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journaled quota format " 2466b31e1552SEric Sandeen "not specified"); 2467ac27a0ecSDave Kleikamp return 0; 2468ac27a0ecSDave Kleikamp } 2469ac27a0ecSDave Kleikamp } 2470ac27a0ecSDave Kleikamp #endif 2471626b035bSRitesh Harjani if (test_opt(sb, DIOREAD_NOLOCK)) { 2472626b035bSRitesh Harjani int blocksize = 2473626b035bSRitesh Harjani BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 2474626b035bSRitesh Harjani if (blocksize < PAGE_SIZE) 2475626b035bSRitesh Harjani ext4_msg(sb, KERN_WARNING, "Warning: mounting with an " 2476626b035bSRitesh Harjani "experimental mount option 'dioread_nolock' " 2477626b035bSRitesh Harjani "for blocksize < PAGE_SIZE"); 2478626b035bSRitesh Harjani } 2479ac27a0ecSDave Kleikamp return 1; 2480ac27a0ecSDave Kleikamp } 2481ac27a0ecSDave Kleikamp 24822adf6da8STheodore Ts'o static inline void ext4_show_quota_options(struct seq_file *seq, 24832adf6da8STheodore Ts'o struct super_block *sb) 24842adf6da8STheodore Ts'o { 24852adf6da8STheodore Ts'o #if defined(CONFIG_QUOTA) 24862adf6da8STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 248733458eabSTheodore Ts'o char *usr_qf_name, *grp_qf_name; 24882adf6da8STheodore Ts'o 24892adf6da8STheodore Ts'o if (sbi->s_jquota_fmt) { 24902adf6da8STheodore Ts'o char *fmtname = ""; 24912adf6da8STheodore Ts'o 24922adf6da8STheodore Ts'o switch (sbi->s_jquota_fmt) { 24932adf6da8STheodore Ts'o case QFMT_VFS_OLD: 24942adf6da8STheodore Ts'o fmtname = "vfsold"; 24952adf6da8STheodore Ts'o break; 24962adf6da8STheodore Ts'o case QFMT_VFS_V0: 24972adf6da8STheodore Ts'o fmtname = "vfsv0"; 24982adf6da8STheodore Ts'o break; 24992adf6da8STheodore Ts'o case QFMT_VFS_V1: 25002adf6da8STheodore Ts'o fmtname = "vfsv1"; 25012adf6da8STheodore Ts'o break; 25022adf6da8STheodore Ts'o } 25032adf6da8STheodore Ts'o seq_printf(seq, ",jqfmt=%s", fmtname); 25042adf6da8STheodore Ts'o } 25052adf6da8STheodore Ts'o 250633458eabSTheodore Ts'o rcu_read_lock(); 250733458eabSTheodore Ts'o usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 250833458eabSTheodore Ts'o grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 250933458eabSTheodore Ts'o if (usr_qf_name) 251033458eabSTheodore Ts'o seq_show_option(seq, "usrjquota", usr_qf_name); 251133458eabSTheodore Ts'o if (grp_qf_name) 251233458eabSTheodore Ts'o seq_show_option(seq, "grpjquota", grp_qf_name); 251333458eabSTheodore Ts'o rcu_read_unlock(); 25142adf6da8STheodore Ts'o #endif 25152adf6da8STheodore Ts'o } 25162adf6da8STheodore Ts'o 25175a916be1STheodore Ts'o static const char *token2str(int token) 25185a916be1STheodore Ts'o { 251950df9fd5SHerton Ronaldo Krzesinski const struct match_token *t; 25205a916be1STheodore Ts'o 25215a916be1STheodore Ts'o for (t = tokens; t->token != Opt_err; t++) 25225a916be1STheodore Ts'o if (t->token == token && !strchr(t->pattern, '=')) 25235a916be1STheodore Ts'o break; 25245a916be1STheodore Ts'o return t->pattern; 25255a916be1STheodore Ts'o } 25265a916be1STheodore Ts'o 25272adf6da8STheodore Ts'o /* 25282adf6da8STheodore Ts'o * Show an option if 25292adf6da8STheodore Ts'o * - it's set to a non-default value OR 25302adf6da8STheodore Ts'o * - if the per-sb default is different from the global default 25312adf6da8STheodore Ts'o */ 253266acdcf4STheodore Ts'o static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 253366acdcf4STheodore Ts'o int nodefs) 25342adf6da8STheodore Ts'o { 25352adf6da8STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 25362adf6da8STheodore Ts'o struct ext4_super_block *es = sbi->s_es; 253768afa7e0STyson Nottingham int def_errors, def_mount_opt = sbi->s_def_mount_opt; 25385a916be1STheodore Ts'o const struct mount_opts *m; 253966acdcf4STheodore Ts'o char sep = nodefs ? '\n' : ','; 25402adf6da8STheodore Ts'o 254166acdcf4STheodore Ts'o #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 254266acdcf4STheodore Ts'o #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 25432adf6da8STheodore Ts'o 25442adf6da8STheodore Ts'o if (sbi->s_sb_block != 1) 25455a916be1STheodore Ts'o SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 25465a916be1STheodore Ts'o 25475a916be1STheodore Ts'o for (m = ext4_mount_opts; m->token != Opt_err; m++) { 25485a916be1STheodore Ts'o int want_set = m->flags & MOPT_SET; 25495a916be1STheodore Ts'o if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 25509cb20f94SIra Weiny (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP) 25515a916be1STheodore Ts'o continue; 255268afa7e0STyson Nottingham if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 25535a916be1STheodore Ts'o continue; /* skip if same as the default */ 25545a916be1STheodore Ts'o if ((want_set && 25555a916be1STheodore Ts'o (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 25565a916be1STheodore Ts'o (!want_set && (sbi->s_mount_opt & m->mount_opt))) 25575a916be1STheodore Ts'o continue; /* select Opt_noFoo vs Opt_Foo */ 25585a916be1STheodore Ts'o SEQ_OPTS_PRINT("%s", token2str(m->token)); 25595a916be1STheodore Ts'o } 25605a916be1STheodore Ts'o 256108cefc7aSEric W. Biederman if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 25625a916be1STheodore Ts'o le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 256308cefc7aSEric W. Biederman SEQ_OPTS_PRINT("resuid=%u", 256408cefc7aSEric W. Biederman from_kuid_munged(&init_user_ns, sbi->s_resuid)); 256508cefc7aSEric W. Biederman if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 25665a916be1STheodore Ts'o le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 256708cefc7aSEric W. Biederman SEQ_OPTS_PRINT("resgid=%u", 256808cefc7aSEric W. Biederman from_kgid_munged(&init_user_ns, sbi->s_resgid)); 256966acdcf4STheodore Ts'o def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 25705a916be1STheodore Ts'o if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 25715a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=remount-ro"); 25722adf6da8STheodore Ts'o if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 25735a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=continue"); 25742adf6da8STheodore Ts'o if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 25755a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=panic"); 257666acdcf4STheodore Ts'o if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 25775a916be1STheodore Ts'o SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 257866acdcf4STheodore Ts'o if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 25795a916be1STheodore Ts'o SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 258066acdcf4STheodore Ts'o if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 25815a916be1STheodore Ts'o SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 2582357fdad0SMatthew Garrett if (sb->s_flags & SB_I_VERSION) 25835a916be1STheodore Ts'o SEQ_OPTS_PUTS("i_version"); 258466acdcf4STheodore Ts'o if (nodefs || sbi->s_stripe) 25855a916be1STheodore Ts'o SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 258668afa7e0STyson Nottingham if (nodefs || EXT4_MOUNT_DATA_FLAGS & 258768afa7e0STyson Nottingham (sbi->s_mount_opt ^ def_mount_opt)) { 25882adf6da8STheodore Ts'o if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 25895a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=journal"); 25902adf6da8STheodore Ts'o else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 25915a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=ordered"); 25922adf6da8STheodore Ts'o else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 25935a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=writeback"); 25945a916be1STheodore Ts'o } 259566acdcf4STheodore Ts'o if (nodefs || 259666acdcf4STheodore Ts'o sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 25975a916be1STheodore Ts'o SEQ_OPTS_PRINT("inode_readahead_blks=%u", 25982adf6da8STheodore Ts'o sbi->s_inode_readahead_blks); 25992adf6da8STheodore Ts'o 2600ceec0376STyson Nottingham if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 260166acdcf4STheodore Ts'o (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 26025a916be1STheodore Ts'o SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 2603df981d03STheodore Ts'o if (nodefs || sbi->s_max_dir_size_kb) 2604df981d03STheodore Ts'o SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 26057915a861SAles Novak if (test_opt(sb, DATA_ERR_ABORT)) 26067915a861SAles Novak SEQ_OPTS_PUTS("data_err=abort"); 2607ed318a6cSEric Biggers 2608ed318a6cSEric Biggers fscrypt_show_test_dummy_encryption(seq, sep, sb); 26092adf6da8STheodore Ts'o 26104f74d15fSEric Biggers if (sb->s_flags & SB_INLINECRYPT) 26114f74d15fSEric Biggers SEQ_OPTS_PUTS("inlinecrypt"); 26124f74d15fSEric Biggers 26139cb20f94SIra Weiny if (test_opt(sb, DAX_ALWAYS)) { 26149cb20f94SIra Weiny if (IS_EXT2_SB(sb)) 26159cb20f94SIra Weiny SEQ_OPTS_PUTS("dax"); 26169cb20f94SIra Weiny else 26179cb20f94SIra Weiny SEQ_OPTS_PUTS("dax=always"); 26189cb20f94SIra Weiny } else if (test_opt2(sb, DAX_NEVER)) { 26199cb20f94SIra Weiny SEQ_OPTS_PUTS("dax=never"); 26209cb20f94SIra Weiny } else if (test_opt2(sb, DAX_INODE)) { 26219cb20f94SIra Weiny SEQ_OPTS_PUTS("dax=inode"); 26229cb20f94SIra Weiny } 26232adf6da8STheodore Ts'o ext4_show_quota_options(seq, sb); 26242adf6da8STheodore Ts'o return 0; 26252adf6da8STheodore Ts'o } 26262adf6da8STheodore Ts'o 262766acdcf4STheodore Ts'o static int ext4_show_options(struct seq_file *seq, struct dentry *root) 262866acdcf4STheodore Ts'o { 262966acdcf4STheodore Ts'o return _ext4_show_options(seq, root->d_sb, 0); 263066acdcf4STheodore Ts'o } 263166acdcf4STheodore Ts'o 2632ebd173beSTheodore Ts'o int ext4_seq_options_show(struct seq_file *seq, void *offset) 263366acdcf4STheodore Ts'o { 263466acdcf4STheodore Ts'o struct super_block *sb = seq->private; 263566acdcf4STheodore Ts'o int rc; 263666acdcf4STheodore Ts'o 2637bc98a42cSDavid Howells seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 263866acdcf4STheodore Ts'o rc = _ext4_show_options(seq, sb, 1); 263966acdcf4STheodore Ts'o seq_puts(seq, "\n"); 264066acdcf4STheodore Ts'o return rc; 264166acdcf4STheodore Ts'o } 264266acdcf4STheodore Ts'o 2643617ba13bSMingming Cao static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 2644ac27a0ecSDave Kleikamp int read_only) 2645ac27a0ecSDave Kleikamp { 2646617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 2647c89128a0SJaegeuk Kim int err = 0; 2648ac27a0ecSDave Kleikamp 2649617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 2650b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "revision level too high, " 2651b31e1552SEric Sandeen "forcing read-only mode"); 2652c89128a0SJaegeuk Kim err = -EROFS; 26535adaccacSyangerkun goto done; 2654ac27a0ecSDave Kleikamp } 2655ac27a0ecSDave Kleikamp if (read_only) 2656281b5995STheodore Ts'o goto done; 2657617ba13bSMingming Cao if (!(sbi->s_mount_state & EXT4_VALID_FS)) 2658b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 2659b31e1552SEric Sandeen "running e2fsck is recommended"); 2660c8b459f4SLukas Czerner else if (sbi->s_mount_state & EXT4_ERROR_FS) 2661b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 2662b31e1552SEric Sandeen "warning: mounting fs with errors, " 2663b31e1552SEric Sandeen "running e2fsck is recommended"); 2664ed3ce80aSTao Ma else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 2665ac27a0ecSDave Kleikamp le16_to_cpu(es->s_mnt_count) >= 2666ac27a0ecSDave Kleikamp (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 2667b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 2668b31e1552SEric Sandeen "warning: maximal mount count reached, " 2669b31e1552SEric Sandeen "running e2fsck is recommended"); 2670ac27a0ecSDave Kleikamp else if (le32_to_cpu(es->s_checkinterval) && 26716a0678a7SArnd Bergmann (ext4_get_tstamp(es, s_lastcheck) + 26726a0678a7SArnd Bergmann le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 2673b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 2674b31e1552SEric Sandeen "warning: checktime reached, " 2675b31e1552SEric Sandeen "running e2fsck is recommended"); 26760390131bSFrank Mayhar if (!sbi->s_journal) 2677216c34b2SMarcin Slusarz es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 2678ac27a0ecSDave Kleikamp if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 2679617ba13bSMingming Cao es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 2680e8546d06SMarcin Slusarz le16_add_cpu(&es->s_mnt_count, 1); 26816a0678a7SArnd Bergmann ext4_update_tstamp(es, s_mtime); 26820390131bSFrank Mayhar if (sbi->s_journal) 2683e2b911c5SDarrick J. Wong ext4_set_feature_journal_needs_recovery(sb); 2684ac27a0ecSDave Kleikamp 26854392fbc4SJan Kara err = ext4_commit_super(sb); 2686281b5995STheodore Ts'o done: 2687ac27a0ecSDave Kleikamp if (test_opt(sb, DEBUG)) 2688a9df9a49STheodore Ts'o printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 2689a2595b8aSTheodore Ts'o "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 2690ac27a0ecSDave Kleikamp sb->s_blocksize, 2691ac27a0ecSDave Kleikamp sbi->s_groups_count, 2692617ba13bSMingming Cao EXT4_BLOCKS_PER_GROUP(sb), 2693617ba13bSMingming Cao EXT4_INODES_PER_GROUP(sb), 2694a2595b8aSTheodore Ts'o sbi->s_mount_opt, sbi->s_mount_opt2); 2695ac27a0ecSDave Kleikamp 26967abc52c2SDan Magenheimer cleancache_init_fs(sb); 2697c89128a0SJaegeuk Kim return err; 2698ac27a0ecSDave Kleikamp } 2699ac27a0ecSDave Kleikamp 2700117fff10STheodore Ts'o int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 2701117fff10STheodore Ts'o { 2702117fff10STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 27037c990728SSuraj Jitindar Singh struct flex_groups **old_groups, **new_groups; 270437b0b6b8SDan Carpenter int size, i, j; 2705117fff10STheodore Ts'o 2706117fff10STheodore Ts'o if (!sbi->s_log_groups_per_flex) 2707117fff10STheodore Ts'o return 0; 2708117fff10STheodore Ts'o 2709117fff10STheodore Ts'o size = ext4_flex_group(sbi, ngroup - 1) + 1; 2710117fff10STheodore Ts'o if (size <= sbi->s_flex_groups_allocated) 2711117fff10STheodore Ts'o return 0; 2712117fff10STheodore Ts'o 27137c990728SSuraj Jitindar Singh new_groups = kvzalloc(roundup_pow_of_two(size * 27147c990728SSuraj Jitindar Singh sizeof(*sbi->s_flex_groups)), GFP_KERNEL); 2715117fff10STheodore Ts'o if (!new_groups) { 27167c990728SSuraj Jitindar Singh ext4_msg(sb, KERN_ERR, 27177c990728SSuraj Jitindar Singh "not enough memory for %d flex group pointers", size); 2718117fff10STheodore Ts'o return -ENOMEM; 2719117fff10STheodore Ts'o } 27207c990728SSuraj Jitindar Singh for (i = sbi->s_flex_groups_allocated; i < size; i++) { 27217c990728SSuraj Jitindar Singh new_groups[i] = kvzalloc(roundup_pow_of_two( 27227c990728SSuraj Jitindar Singh sizeof(struct flex_groups)), 27237c990728SSuraj Jitindar Singh GFP_KERNEL); 27247c990728SSuraj Jitindar Singh if (!new_groups[i]) { 272537b0b6b8SDan Carpenter for (j = sbi->s_flex_groups_allocated; j < i; j++) 272637b0b6b8SDan Carpenter kvfree(new_groups[j]); 27277c990728SSuraj Jitindar Singh kvfree(new_groups); 27287c990728SSuraj Jitindar Singh ext4_msg(sb, KERN_ERR, 27297c990728SSuraj Jitindar Singh "not enough memory for %d flex groups", size); 27307c990728SSuraj Jitindar Singh return -ENOMEM; 2731117fff10STheodore Ts'o } 27327c990728SSuraj Jitindar Singh } 27337c990728SSuraj Jitindar Singh rcu_read_lock(); 27347c990728SSuraj Jitindar Singh old_groups = rcu_dereference(sbi->s_flex_groups); 27357c990728SSuraj Jitindar Singh if (old_groups) 27367c990728SSuraj Jitindar Singh memcpy(new_groups, old_groups, 27377c990728SSuraj Jitindar Singh (sbi->s_flex_groups_allocated * 27387c990728SSuraj Jitindar Singh sizeof(struct flex_groups *))); 27397c990728SSuraj Jitindar Singh rcu_read_unlock(); 27407c990728SSuraj Jitindar Singh rcu_assign_pointer(sbi->s_flex_groups, new_groups); 27417c990728SSuraj Jitindar Singh sbi->s_flex_groups_allocated = size; 27427c990728SSuraj Jitindar Singh if (old_groups) 27437c990728SSuraj Jitindar Singh ext4_kvfree_array_rcu(old_groups); 2744117fff10STheodore Ts'o return 0; 2745117fff10STheodore Ts'o } 2746117fff10STheodore Ts'o 2747772cb7c8SJose R. Santos static int ext4_fill_flex_info(struct super_block *sb) 2748772cb7c8SJose R. Santos { 2749772cb7c8SJose R. Santos struct ext4_sb_info *sbi = EXT4_SB(sb); 2750772cb7c8SJose R. Santos struct ext4_group_desc *gdp = NULL; 27517c990728SSuraj Jitindar Singh struct flex_groups *fg; 2752772cb7c8SJose R. Santos ext4_group_t flex_group; 2753117fff10STheodore Ts'o int i, err; 2754772cb7c8SJose R. Santos 2755503358aeSTheodore Ts'o sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 2756d50f2ab6SXi Wang if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 2757772cb7c8SJose R. Santos sbi->s_log_groups_per_flex = 0; 2758772cb7c8SJose R. Santos return 1; 2759772cb7c8SJose R. Santos } 2760772cb7c8SJose R. Santos 2761117fff10STheodore Ts'o err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 2762117fff10STheodore Ts'o if (err) 2763772cb7c8SJose R. Santos goto failed; 2764772cb7c8SJose R. Santos 2765772cb7c8SJose R. Santos for (i = 0; i < sbi->s_groups_count; i++) { 276688b6edd1STheodore Ts'o gdp = ext4_get_group_desc(sb, i, NULL); 2767772cb7c8SJose R. Santos 2768772cb7c8SJose R. Santos flex_group = ext4_flex_group(sbi, i); 27697c990728SSuraj Jitindar Singh fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group); 27707c990728SSuraj Jitindar Singh atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes); 277190ba983fSTheodore Ts'o atomic64_add(ext4_free_group_clusters(sb, gdp), 27727c990728SSuraj Jitindar Singh &fg->free_clusters); 27737c990728SSuraj Jitindar Singh atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs); 2774772cb7c8SJose R. Santos } 2775772cb7c8SJose R. Santos 2776772cb7c8SJose R. Santos return 1; 2777772cb7c8SJose R. Santos failed: 2778772cb7c8SJose R. Santos return 0; 2779772cb7c8SJose R. Santos } 2780772cb7c8SJose R. Santos 2781e2b911c5SDarrick J. Wong static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 2782717d50e4SAndreas Dilger struct ext4_group_desc *gdp) 2783717d50e4SAndreas Dilger { 2784b47820edSDaeho Jeong int offset = offsetof(struct ext4_group_desc, bg_checksum); 2785717d50e4SAndreas Dilger __u16 crc = 0; 2786717d50e4SAndreas Dilger __le32 le_group = cpu_to_le32(block_group); 2787e2b911c5SDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 2788717d50e4SAndreas Dilger 27899aa5d32bSDmitry Monakhov if (ext4_has_metadata_csum(sbi->s_sb)) { 2790feb0ab32SDarrick J. Wong /* Use new metadata_csum algorithm */ 2791feb0ab32SDarrick J. Wong __u32 csum32; 2792b47820edSDaeho Jeong __u16 dummy_csum = 0; 2793feb0ab32SDarrick J. Wong 2794feb0ab32SDarrick J. Wong csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 2795feb0ab32SDarrick J. Wong sizeof(le_group)); 2796b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 2797b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 2798b47820edSDaeho Jeong sizeof(dummy_csum)); 2799b47820edSDaeho Jeong offset += sizeof(dummy_csum); 2800b47820edSDaeho Jeong if (offset < sbi->s_desc_size) 2801b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 2802b47820edSDaeho Jeong sbi->s_desc_size - offset); 2803feb0ab32SDarrick J. Wong 2804feb0ab32SDarrick J. Wong crc = csum32 & 0xFFFF; 2805feb0ab32SDarrick J. Wong goto out; 2806feb0ab32SDarrick J. Wong } 2807feb0ab32SDarrick J. Wong 2808feb0ab32SDarrick J. Wong /* old crc16 code */ 2809e2b911c5SDarrick J. Wong if (!ext4_has_feature_gdt_csum(sb)) 2810813d32f9SDarrick J. Wong return 0; 2811813d32f9SDarrick J. Wong 2812717d50e4SAndreas Dilger crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2813717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2814717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)gdp, offset); 2815717d50e4SAndreas Dilger offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2816717d50e4SAndreas Dilger /* for checksum of struct ext4_group_desc do the rest...*/ 2817e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb) && 2818717d50e4SAndreas Dilger offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2819717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)gdp + offset, 2820717d50e4SAndreas Dilger le16_to_cpu(sbi->s_es->s_desc_size) - 2821717d50e4SAndreas Dilger offset); 2822717d50e4SAndreas Dilger 2823feb0ab32SDarrick J. Wong out: 2824717d50e4SAndreas Dilger return cpu_to_le16(crc); 2825717d50e4SAndreas Dilger } 2826717d50e4SAndreas Dilger 2827feb0ab32SDarrick J. Wong int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2828717d50e4SAndreas Dilger struct ext4_group_desc *gdp) 2829717d50e4SAndreas Dilger { 2830feb0ab32SDarrick J. Wong if (ext4_has_group_desc_csum(sb) && 2831e2b911c5SDarrick J. Wong (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 2832717d50e4SAndreas Dilger return 0; 2833717d50e4SAndreas Dilger 2834717d50e4SAndreas Dilger return 1; 2835717d50e4SAndreas Dilger } 2836717d50e4SAndreas Dilger 2837feb0ab32SDarrick J. Wong void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2838feb0ab32SDarrick J. Wong struct ext4_group_desc *gdp) 2839feb0ab32SDarrick J. Wong { 2840feb0ab32SDarrick J. Wong if (!ext4_has_group_desc_csum(sb)) 2841feb0ab32SDarrick J. Wong return; 2842e2b911c5SDarrick J. Wong gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 2843feb0ab32SDarrick J. Wong } 2844feb0ab32SDarrick J. Wong 2845ac27a0ecSDave Kleikamp /* Called at mount-time, super-block is locked */ 2846bfff6873SLukas Czerner static int ext4_check_descriptors(struct super_block *sb, 2847829fa70dSTheodore Ts'o ext4_fsblk_t sb_block, 2848bfff6873SLukas Czerner ext4_group_t *first_not_zeroed) 2849ac27a0ecSDave Kleikamp { 2850617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 2851617ba13bSMingming Cao ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2852617ba13bSMingming Cao ext4_fsblk_t last_block; 285344de022cSTheodore Ts'o ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 2854bd81d8eeSLaurent Vivier ext4_fsblk_t block_bitmap; 2855bd81d8eeSLaurent Vivier ext4_fsblk_t inode_bitmap; 2856bd81d8eeSLaurent Vivier ext4_fsblk_t inode_table; 2857ce421581SJose R. Santos int flexbg_flag = 0; 2858bfff6873SLukas Czerner ext4_group_t i, grp = sbi->s_groups_count; 2859ac27a0ecSDave Kleikamp 2860e2b911c5SDarrick J. Wong if (ext4_has_feature_flex_bg(sb)) 2861ce421581SJose R. Santos flexbg_flag = 1; 2862ce421581SJose R. Santos 2863617ba13bSMingming Cao ext4_debug("Checking group descriptors"); 2864ac27a0ecSDave Kleikamp 2865197cd65aSAkinobu Mita for (i = 0; i < sbi->s_groups_count; i++) { 2866197cd65aSAkinobu Mita struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2867197cd65aSAkinobu Mita 2868ce421581SJose R. Santos if (i == sbi->s_groups_count - 1 || flexbg_flag) 2869bd81d8eeSLaurent Vivier last_block = ext4_blocks_count(sbi->s_es) - 1; 2870ac27a0ecSDave Kleikamp else 2871ac27a0ecSDave Kleikamp last_block = first_block + 2872617ba13bSMingming Cao (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2873ac27a0ecSDave Kleikamp 2874bfff6873SLukas Czerner if ((grp == sbi->s_groups_count) && 2875bfff6873SLukas Czerner !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2876bfff6873SLukas Czerner grp = i; 2877bfff6873SLukas Czerner 28788fadc143SAlexandre Ratchov block_bitmap = ext4_block_bitmap(sb, gdp); 2879829fa70dSTheodore Ts'o if (block_bitmap == sb_block) { 2880829fa70dSTheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2881829fa70dSTheodore Ts'o "Block bitmap for group %u overlaps " 2882829fa70dSTheodore Ts'o "superblock", i); 288318db4b4eSTheodore Ts'o if (!sb_rdonly(sb)) 288418db4b4eSTheodore Ts'o return 0; 2885829fa70dSTheodore Ts'o } 288677260807STheodore Ts'o if (block_bitmap >= sb_block + 1 && 288777260807STheodore Ts'o block_bitmap <= last_bg_block) { 288877260807STheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 288977260807STheodore Ts'o "Block bitmap for group %u overlaps " 289077260807STheodore Ts'o "block group descriptors", i); 289177260807STheodore Ts'o if (!sb_rdonly(sb)) 289277260807STheodore Ts'o return 0; 289377260807STheodore Ts'o } 28942b2d6d01STheodore Ts'o if (block_bitmap < first_block || block_bitmap > last_block) { 2895b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2896a9df9a49STheodore Ts'o "Block bitmap for group %u not in group " 2897b31e1552SEric Sandeen "(block %llu)!", i, block_bitmap); 2898ac27a0ecSDave Kleikamp return 0; 2899ac27a0ecSDave Kleikamp } 29008fadc143SAlexandre Ratchov inode_bitmap = ext4_inode_bitmap(sb, gdp); 2901829fa70dSTheodore Ts'o if (inode_bitmap == sb_block) { 2902829fa70dSTheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2903829fa70dSTheodore Ts'o "Inode bitmap for group %u overlaps " 2904829fa70dSTheodore Ts'o "superblock", i); 290518db4b4eSTheodore Ts'o if (!sb_rdonly(sb)) 290618db4b4eSTheodore Ts'o return 0; 2907829fa70dSTheodore Ts'o } 290877260807STheodore Ts'o if (inode_bitmap >= sb_block + 1 && 290977260807STheodore Ts'o inode_bitmap <= last_bg_block) { 291077260807STheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 291177260807STheodore Ts'o "Inode bitmap for group %u overlaps " 291277260807STheodore Ts'o "block group descriptors", i); 291377260807STheodore Ts'o if (!sb_rdonly(sb)) 291477260807STheodore Ts'o return 0; 291577260807STheodore Ts'o } 29162b2d6d01STheodore Ts'o if (inode_bitmap < first_block || inode_bitmap > last_block) { 2917b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2918a9df9a49STheodore Ts'o "Inode bitmap for group %u not in group " 2919b31e1552SEric Sandeen "(block %llu)!", i, inode_bitmap); 2920ac27a0ecSDave Kleikamp return 0; 2921ac27a0ecSDave Kleikamp } 29228fadc143SAlexandre Ratchov inode_table = ext4_inode_table(sb, gdp); 2923829fa70dSTheodore Ts'o if (inode_table == sb_block) { 2924829fa70dSTheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2925829fa70dSTheodore Ts'o "Inode table for group %u overlaps " 2926829fa70dSTheodore Ts'o "superblock", i); 292718db4b4eSTheodore Ts'o if (!sb_rdonly(sb)) 292818db4b4eSTheodore Ts'o return 0; 2929829fa70dSTheodore Ts'o } 293077260807STheodore Ts'o if (inode_table >= sb_block + 1 && 293177260807STheodore Ts'o inode_table <= last_bg_block) { 293277260807STheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 293377260807STheodore Ts'o "Inode table for group %u overlaps " 293477260807STheodore Ts'o "block group descriptors", i); 293577260807STheodore Ts'o if (!sb_rdonly(sb)) 293677260807STheodore Ts'o return 0; 293777260807STheodore Ts'o } 2938bd81d8eeSLaurent Vivier if (inode_table < first_block || 29392b2d6d01STheodore Ts'o inode_table + sbi->s_itb_per_group - 1 > last_block) { 2940b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2941a9df9a49STheodore Ts'o "Inode table for group %u not in group " 2942b31e1552SEric Sandeen "(block %llu)!", i, inode_table); 2943ac27a0ecSDave Kleikamp return 0; 2944ac27a0ecSDave Kleikamp } 2945955ce5f5SAneesh Kumar K.V ext4_lock_group(sb, i); 2946feb0ab32SDarrick J. Wong if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2947b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2948b31e1552SEric Sandeen "Checksum for group %u failed (%u!=%u)", 2949e2b911c5SDarrick J. Wong i, le16_to_cpu(ext4_group_desc_csum(sb, i, 2950fd2d4291SAvantika Mathur gdp)), le16_to_cpu(gdp->bg_checksum)); 2951bc98a42cSDavid Howells if (!sb_rdonly(sb)) { 2952955ce5f5SAneesh Kumar K.V ext4_unlock_group(sb, i); 2953717d50e4SAndreas Dilger return 0; 2954717d50e4SAndreas Dilger } 29557ee1ec4cSLi Zefan } 2956955ce5f5SAneesh Kumar K.V ext4_unlock_group(sb, i); 2957ce421581SJose R. Santos if (!flexbg_flag) 2958617ba13bSMingming Cao first_block += EXT4_BLOCKS_PER_GROUP(sb); 2959ac27a0ecSDave Kleikamp } 2960bfff6873SLukas Czerner if (NULL != first_not_zeroed) 2961bfff6873SLukas Czerner *first_not_zeroed = grp; 2962ac27a0ecSDave Kleikamp return 1; 2963ac27a0ecSDave Kleikamp } 2964ac27a0ecSDave Kleikamp 2965617ba13bSMingming Cao /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2966ac27a0ecSDave Kleikamp * the superblock) which were deleted from all directories, but held open by 2967ac27a0ecSDave Kleikamp * a process at the time of a crash. We walk the list and try to delete these 2968ac27a0ecSDave Kleikamp * inodes at recovery time (only with a read-write filesystem). 2969ac27a0ecSDave Kleikamp * 2970ac27a0ecSDave Kleikamp * In order to keep the orphan inode chain consistent during traversal (in 2971ac27a0ecSDave Kleikamp * case of crash during recovery), we link each inode into the superblock 2972ac27a0ecSDave Kleikamp * orphan list_head and handle it the same way as an inode deletion during 2973ac27a0ecSDave Kleikamp * normal operation (which journals the operations for us). 2974ac27a0ecSDave Kleikamp * 2975ac27a0ecSDave Kleikamp * We only do an iget() and an iput() on each inode, which is very safe if we 2976ac27a0ecSDave Kleikamp * accidentally point at an in-use or already deleted inode. The worst that 2977ac27a0ecSDave Kleikamp * can happen in this case is that we get a "bit already cleared" message from 2978617ba13bSMingming Cao * ext4_free_inode(). The only reason we would point at a wrong inode is if 2979ac27a0ecSDave Kleikamp * e2fsck was run on this filesystem, and it must have already done the orphan 2980ac27a0ecSDave Kleikamp * inode cleanup for us, so we can safely abort without any further action. 2981ac27a0ecSDave Kleikamp */ 2982617ba13bSMingming Cao static void ext4_orphan_cleanup(struct super_block *sb, 2983617ba13bSMingming Cao struct ext4_super_block *es) 2984ac27a0ecSDave Kleikamp { 2985ac27a0ecSDave Kleikamp unsigned int s_flags = sb->s_flags; 29862c98eb5eSTheodore Ts'o int ret, nr_orphans = 0, nr_truncates = 0; 2987ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 298895f1fda4Szhangyi (F) int quota_update = 0; 2989ac27a0ecSDave Kleikamp int i; 2990ac27a0ecSDave Kleikamp #endif 2991ac27a0ecSDave Kleikamp if (!es->s_last_orphan) { 2992ac27a0ecSDave Kleikamp jbd_debug(4, "no orphan inodes to clean up\n"); 2993ac27a0ecSDave Kleikamp return; 2994ac27a0ecSDave Kleikamp } 2995ac27a0ecSDave Kleikamp 2996a8f48a95SEric Sandeen if (bdev_read_only(sb->s_bdev)) { 2997b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "write access " 2998b31e1552SEric Sandeen "unavailable, skipping orphan cleanup"); 2999a8f48a95SEric Sandeen return; 3000a8f48a95SEric Sandeen } 3001a8f48a95SEric Sandeen 3002d39195c3SAmir Goldstein /* Check if feature set would not allow a r/w mount */ 3003d39195c3SAmir Goldstein if (!ext4_feature_set_ok(sb, 0)) { 3004d39195c3SAmir Goldstein ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 3005d39195c3SAmir Goldstein "unknown ROCOMPAT features"); 3006d39195c3SAmir Goldstein return; 3007d39195c3SAmir Goldstein } 3008d39195c3SAmir Goldstein 3009617ba13bSMingming Cao if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 3010c25f9bc6SEric Sandeen /* don't clear list on RO mount w/ errors */ 30111751e8a6SLinus Torvalds if (es->s_last_orphan && !(s_flags & SB_RDONLY)) { 301284474976SDmitry Monakhov ext4_msg(sb, KERN_INFO, "Errors on filesystem, " 3013ac27a0ecSDave Kleikamp "clearing orphan list.\n"); 3014ac27a0ecSDave Kleikamp es->s_last_orphan = 0; 3015c25f9bc6SEric Sandeen } 3016ac27a0ecSDave Kleikamp jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 3017ac27a0ecSDave Kleikamp return; 3018ac27a0ecSDave Kleikamp } 3019ac27a0ecSDave Kleikamp 30201751e8a6SLinus Torvalds if (s_flags & SB_RDONLY) { 3021b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 30221751e8a6SLinus Torvalds sb->s_flags &= ~SB_RDONLY; 3023ac27a0ecSDave Kleikamp } 3024ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 3025ac27a0ecSDave Kleikamp /* Needed for iput() to work correctly and not trash data */ 30261751e8a6SLinus Torvalds sb->s_flags |= SB_ACTIVE; 302795f1fda4Szhangyi (F) 302895f1fda4Szhangyi (F) /* 302995f1fda4Szhangyi (F) * Turn on quotas which were not enabled for read-only mounts if 303095f1fda4Szhangyi (F) * filesystem has quota feature, so that they are updated correctly. 303195f1fda4Szhangyi (F) */ 30321751e8a6SLinus Torvalds if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) { 303395f1fda4Szhangyi (F) int ret = ext4_enable_quotas(sb); 303495f1fda4Szhangyi (F) 303595f1fda4Szhangyi (F) if (!ret) 303695f1fda4Szhangyi (F) quota_update = 1; 303795f1fda4Szhangyi (F) else 303895f1fda4Szhangyi (F) ext4_msg(sb, KERN_ERR, 303995f1fda4Szhangyi (F) "Cannot turn on quotas: error %d", ret); 304095f1fda4Szhangyi (F) } 304195f1fda4Szhangyi (F) 304295f1fda4Szhangyi (F) /* Turn on journaled quotas used for old sytle */ 3043a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 3044617ba13bSMingming Cao if (EXT4_SB(sb)->s_qf_names[i]) { 3045617ba13bSMingming Cao int ret = ext4_quota_on_mount(sb, i); 304695f1fda4Szhangyi (F) 304795f1fda4Szhangyi (F) if (!ret) 304895f1fda4Szhangyi (F) quota_update = 1; 304995f1fda4Szhangyi (F) else 3050b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 3051b31e1552SEric Sandeen "Cannot turn on journaled " 305295f1fda4Szhangyi (F) "quota: type %d: error %d", i, ret); 3053ac27a0ecSDave Kleikamp } 3054ac27a0ecSDave Kleikamp } 3055ac27a0ecSDave Kleikamp #endif 3056ac27a0ecSDave Kleikamp 3057ac27a0ecSDave Kleikamp while (es->s_last_orphan) { 3058ac27a0ecSDave Kleikamp struct inode *inode; 3059ac27a0ecSDave Kleikamp 3060c65d5c6cSVegard Nossum /* 3061c65d5c6cSVegard Nossum * We may have encountered an error during cleanup; if 3062c65d5c6cSVegard Nossum * so, skip the rest. 3063c65d5c6cSVegard Nossum */ 3064c65d5c6cSVegard Nossum if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 3065c65d5c6cSVegard Nossum jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 3066c65d5c6cSVegard Nossum es->s_last_orphan = 0; 3067c65d5c6cSVegard Nossum break; 3068c65d5c6cSVegard Nossum } 3069c65d5c6cSVegard Nossum 307097bd42b9SJosef Bacik inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 307197bd42b9SJosef Bacik if (IS_ERR(inode)) { 3072ac27a0ecSDave Kleikamp es->s_last_orphan = 0; 3073ac27a0ecSDave Kleikamp break; 3074ac27a0ecSDave Kleikamp } 3075ac27a0ecSDave Kleikamp 3076617ba13bSMingming Cao list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 3077871a2931SChristoph Hellwig dquot_initialize(inode); 3078ac27a0ecSDave Kleikamp if (inode->i_nlink) { 3079566370a2SPaul Taysom if (test_opt(sb, DEBUG)) 3080b31e1552SEric Sandeen ext4_msg(sb, KERN_DEBUG, 3081b31e1552SEric Sandeen "%s: truncating inode %lu to %lld bytes", 308246e665e9SHarvey Harrison __func__, inode->i_ino, inode->i_size); 3083e5f8eab8STheodore Ts'o jbd_debug(2, "truncating inode %lu to %lld bytes\n", 3084ac27a0ecSDave Kleikamp inode->i_ino, inode->i_size); 30855955102cSAl Viro inode_lock(inode); 308655f252c9SLukas Czerner truncate_inode_pages(inode->i_mapping, inode->i_size); 30872c98eb5eSTheodore Ts'o ret = ext4_truncate(inode); 30882c98eb5eSTheodore Ts'o if (ret) 30892c98eb5eSTheodore Ts'o ext4_std_error(inode->i_sb, ret); 30905955102cSAl Viro inode_unlock(inode); 3091ac27a0ecSDave Kleikamp nr_truncates++; 3092ac27a0ecSDave Kleikamp } else { 3093566370a2SPaul Taysom if (test_opt(sb, DEBUG)) 3094b31e1552SEric Sandeen ext4_msg(sb, KERN_DEBUG, 3095b31e1552SEric Sandeen "%s: deleting unreferenced inode %lu", 309646e665e9SHarvey Harrison __func__, inode->i_ino); 3097ac27a0ecSDave Kleikamp jbd_debug(2, "deleting unreferenced inode %lu\n", 3098ac27a0ecSDave Kleikamp inode->i_ino); 3099ac27a0ecSDave Kleikamp nr_orphans++; 3100ac27a0ecSDave Kleikamp } 3101ac27a0ecSDave Kleikamp iput(inode); /* The delete magic happens here! */ 3102ac27a0ecSDave Kleikamp } 3103ac27a0ecSDave Kleikamp 3104ac27a0ecSDave Kleikamp #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 3105ac27a0ecSDave Kleikamp 3106ac27a0ecSDave Kleikamp if (nr_orphans) 3107b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 3108b31e1552SEric Sandeen PLURAL(nr_orphans)); 3109ac27a0ecSDave Kleikamp if (nr_truncates) 3110b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 3111b31e1552SEric Sandeen PLURAL(nr_truncates)); 3112ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 311395f1fda4Szhangyi (F) /* Turn off quotas if they were enabled for orphan cleanup */ 311495f1fda4Szhangyi (F) if (quota_update) { 3115a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 311695f1fda4Szhangyi (F) if (sb_dqopt(sb)->files[i]) 3117287a8095SChristoph Hellwig dquot_quota_off(sb, i); 3118ac27a0ecSDave Kleikamp } 311995f1fda4Szhangyi (F) } 3120ac27a0ecSDave Kleikamp #endif 31211751e8a6SLinus Torvalds sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 3122ac27a0ecSDave Kleikamp } 31230b8e58a1SAndreas Dilger 3124cd2291a4SEric Sandeen /* 3125cd2291a4SEric Sandeen * Maximal extent format file size. 3126cd2291a4SEric Sandeen * Resulting logical blkno at s_maxbytes must fit in our on-disk 3127cd2291a4SEric Sandeen * extent format containers, within a sector_t, and within i_blocks 3128cd2291a4SEric Sandeen * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 3129cd2291a4SEric Sandeen * so that won't be a limiting factor. 3130cd2291a4SEric Sandeen * 3131f17722f9SLukas Czerner * However there is other limiting factor. We do store extents in the form 3132f17722f9SLukas Czerner * of starting block and length, hence the resulting length of the extent 3133f17722f9SLukas Czerner * covering maximum file size must fit into on-disk format containers as 3134f17722f9SLukas Czerner * well. Given that length is always by 1 unit bigger than max unit (because 3135f17722f9SLukas Czerner * we count 0 as well) we have to lower the s_maxbytes by one fs block. 3136f17722f9SLukas Czerner * 3137cd2291a4SEric Sandeen * Note, this does *not* consider any metadata overhead for vfs i_blocks. 3138cd2291a4SEric Sandeen */ 3139f287a1a5STheodore Ts'o static loff_t ext4_max_size(int blkbits, int has_huge_files) 3140cd2291a4SEric Sandeen { 3141cd2291a4SEric Sandeen loff_t res; 3142cd2291a4SEric Sandeen loff_t upper_limit = MAX_LFS_FILESIZE; 3143cd2291a4SEric Sandeen 314472deb455SChristoph Hellwig BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); 314572deb455SChristoph Hellwig 314672deb455SChristoph Hellwig if (!has_huge_files) { 3147cd2291a4SEric Sandeen upper_limit = (1LL << 32) - 1; 3148cd2291a4SEric Sandeen 3149cd2291a4SEric Sandeen /* total blocks in file system block size */ 3150cd2291a4SEric Sandeen upper_limit >>= (blkbits - 9); 3151cd2291a4SEric Sandeen upper_limit <<= blkbits; 3152cd2291a4SEric Sandeen } 3153cd2291a4SEric Sandeen 3154f17722f9SLukas Czerner /* 3155f17722f9SLukas Czerner * 32-bit extent-start container, ee_block. We lower the maxbytes 3156f17722f9SLukas Czerner * by one fs block, so ee_len can cover the extent of maximum file 3157f17722f9SLukas Czerner * size 3158f17722f9SLukas Czerner */ 3159f17722f9SLukas Czerner res = (1LL << 32) - 1; 3160cd2291a4SEric Sandeen res <<= blkbits; 3161cd2291a4SEric Sandeen 3162cd2291a4SEric Sandeen /* Sanity check against vm- & vfs- imposed limits */ 3163cd2291a4SEric Sandeen if (res > upper_limit) 3164cd2291a4SEric Sandeen res = upper_limit; 3165cd2291a4SEric Sandeen 3166cd2291a4SEric Sandeen return res; 3167cd2291a4SEric Sandeen } 3168ac27a0ecSDave Kleikamp 3169ac27a0ecSDave Kleikamp /* 3170cd2291a4SEric Sandeen * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 31710fc1b451SAneesh Kumar K.V * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 31720fc1b451SAneesh Kumar K.V * We need to be 1 filesystem block less than the 2^48 sector limit. 3173ac27a0ecSDave Kleikamp */ 3174f287a1a5STheodore Ts'o static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 3175ac27a0ecSDave Kleikamp { 3176617ba13bSMingming Cao loff_t res = EXT4_NDIR_BLOCKS; 31770fc1b451SAneesh Kumar K.V int meta_blocks; 31780fc1b451SAneesh Kumar K.V loff_t upper_limit; 31790b8e58a1SAndreas Dilger /* This is calculated to be the largest file size for a dense, block 31800b8e58a1SAndreas Dilger * mapped file such that the file's total number of 512-byte sectors, 31810b8e58a1SAndreas Dilger * including data and all indirect blocks, does not exceed (2^48 - 1). 31820b8e58a1SAndreas Dilger * 31830b8e58a1SAndreas Dilger * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 31840b8e58a1SAndreas Dilger * number of 512-byte sectors of the file. 31850fc1b451SAneesh Kumar K.V */ 31860fc1b451SAneesh Kumar K.V 318772deb455SChristoph Hellwig if (!has_huge_files) { 31880fc1b451SAneesh Kumar K.V /* 318972deb455SChristoph Hellwig * !has_huge_files or implies that the inode i_block field 319072deb455SChristoph Hellwig * represents total file blocks in 2^32 512-byte sectors == 319172deb455SChristoph Hellwig * size of vfs inode i_blocks * 8 31920fc1b451SAneesh Kumar K.V */ 31930fc1b451SAneesh Kumar K.V upper_limit = (1LL << 32) - 1; 31940fc1b451SAneesh Kumar K.V 31950fc1b451SAneesh Kumar K.V /* total blocks in file system block size */ 31960fc1b451SAneesh Kumar K.V upper_limit >>= (bits - 9); 31970fc1b451SAneesh Kumar K.V 31980fc1b451SAneesh Kumar K.V } else { 31998180a562SAneesh Kumar K.V /* 32008180a562SAneesh Kumar K.V * We use 48 bit ext4_inode i_blocks 32018180a562SAneesh Kumar K.V * With EXT4_HUGE_FILE_FL set the i_blocks 32028180a562SAneesh Kumar K.V * represent total number of blocks in 32038180a562SAneesh Kumar K.V * file system block size 32048180a562SAneesh Kumar K.V */ 32050fc1b451SAneesh Kumar K.V upper_limit = (1LL << 48) - 1; 32060fc1b451SAneesh Kumar K.V 32070fc1b451SAneesh Kumar K.V } 32080fc1b451SAneesh Kumar K.V 32090fc1b451SAneesh Kumar K.V /* indirect blocks */ 32100fc1b451SAneesh Kumar K.V meta_blocks = 1; 32110fc1b451SAneesh Kumar K.V /* double indirect blocks */ 32120fc1b451SAneesh Kumar K.V meta_blocks += 1 + (1LL << (bits-2)); 32130fc1b451SAneesh Kumar K.V /* tripple indirect blocks */ 32140fc1b451SAneesh Kumar K.V meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 32150fc1b451SAneesh Kumar K.V 32160fc1b451SAneesh Kumar K.V upper_limit -= meta_blocks; 32170fc1b451SAneesh Kumar K.V upper_limit <<= bits; 3218ac27a0ecSDave Kleikamp 3219ac27a0ecSDave Kleikamp res += 1LL << (bits-2); 3220ac27a0ecSDave Kleikamp res += 1LL << (2*(bits-2)); 3221ac27a0ecSDave Kleikamp res += 1LL << (3*(bits-2)); 3222ac27a0ecSDave Kleikamp res <<= bits; 3223ac27a0ecSDave Kleikamp if (res > upper_limit) 3224ac27a0ecSDave Kleikamp res = upper_limit; 32250fc1b451SAneesh Kumar K.V 32260fc1b451SAneesh Kumar K.V if (res > MAX_LFS_FILESIZE) 32270fc1b451SAneesh Kumar K.V res = MAX_LFS_FILESIZE; 32280fc1b451SAneesh Kumar K.V 3229ac27a0ecSDave Kleikamp return res; 3230ac27a0ecSDave Kleikamp } 3231ac27a0ecSDave Kleikamp 3232617ba13bSMingming Cao static ext4_fsblk_t descriptor_loc(struct super_block *sb, 323370bbb3e0SAndrew Morton ext4_fsblk_t logical_sb_block, int nr) 3234ac27a0ecSDave Kleikamp { 3235617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 3236fd2d4291SAvantika Mathur ext4_group_t bg, first_meta_bg; 3237ac27a0ecSDave Kleikamp int has_super = 0; 3238ac27a0ecSDave Kleikamp 3239ac27a0ecSDave Kleikamp first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 3240ac27a0ecSDave Kleikamp 3241e2b911c5SDarrick J. Wong if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 324270bbb3e0SAndrew Morton return logical_sb_block + nr + 1; 3243ac27a0ecSDave Kleikamp bg = sbi->s_desc_per_block * nr; 3244617ba13bSMingming Cao if (ext4_bg_has_super(sb, bg)) 3245ac27a0ecSDave Kleikamp has_super = 1; 32460b8e58a1SAndreas Dilger 3247bd63f6b0SDarrick J. Wong /* 3248bd63f6b0SDarrick J. Wong * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 3249bd63f6b0SDarrick J. Wong * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 3250bd63f6b0SDarrick J. Wong * on modern mke2fs or blksize > 1k on older mke2fs) then we must 3251bd63f6b0SDarrick J. Wong * compensate. 3252bd63f6b0SDarrick J. Wong */ 3253bd63f6b0SDarrick J. Wong if (sb->s_blocksize == 1024 && nr == 0 && 325449598e04SJun Piao le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 3255bd63f6b0SDarrick J. Wong has_super++; 3256bd63f6b0SDarrick J. Wong 3257617ba13bSMingming Cao return (has_super + ext4_group_first_block_no(sb, bg)); 3258ac27a0ecSDave Kleikamp } 3259ac27a0ecSDave Kleikamp 3260c9de560dSAlex Tomas /** 3261c9de560dSAlex Tomas * ext4_get_stripe_size: Get the stripe size. 3262c9de560dSAlex Tomas * @sbi: In memory super block info 3263c9de560dSAlex Tomas * 3264c9de560dSAlex Tomas * If we have specified it via mount option, then 3265c9de560dSAlex Tomas * use the mount option value. If the value specified at mount time is 3266c9de560dSAlex Tomas * greater than the blocks per group use the super block value. 3267c9de560dSAlex Tomas * If the super block value is greater than blocks per group return 0. 3268c9de560dSAlex Tomas * Allocator needs it be less than blocks per group. 3269c9de560dSAlex Tomas * 3270c9de560dSAlex Tomas */ 3271c9de560dSAlex Tomas static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 3272c9de560dSAlex Tomas { 3273c9de560dSAlex Tomas unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 3274c9de560dSAlex Tomas unsigned long stripe_width = 3275c9de560dSAlex Tomas le32_to_cpu(sbi->s_es->s_raid_stripe_width); 32763eb08658SDan Ehrenberg int ret; 3277c9de560dSAlex Tomas 3278c9de560dSAlex Tomas if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 32793eb08658SDan Ehrenberg ret = sbi->s_stripe; 32805469d7c3SJan Kara else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 32813eb08658SDan Ehrenberg ret = stripe_width; 32825469d7c3SJan Kara else if (stride && stride <= sbi->s_blocks_per_group) 32833eb08658SDan Ehrenberg ret = stride; 32843eb08658SDan Ehrenberg else 32853eb08658SDan Ehrenberg ret = 0; 3286c9de560dSAlex Tomas 32873eb08658SDan Ehrenberg /* 32883eb08658SDan Ehrenberg * If the stripe width is 1, this makes no sense and 32893eb08658SDan Ehrenberg * we set it to 0 to turn off stripe handling code. 32903eb08658SDan Ehrenberg */ 32913eb08658SDan Ehrenberg if (ret <= 1) 32923eb08658SDan Ehrenberg ret = 0; 3293c9de560dSAlex Tomas 32943eb08658SDan Ehrenberg return ret; 3295c9de560dSAlex Tomas } 3296ac27a0ecSDave Kleikamp 3297a13fb1a4SEric Sandeen /* 3298a13fb1a4SEric Sandeen * Check whether this filesystem can be mounted based on 3299a13fb1a4SEric Sandeen * the features present and the RDONLY/RDWR mount requested. 3300a13fb1a4SEric Sandeen * Returns 1 if this filesystem can be mounted as requested, 3301a13fb1a4SEric Sandeen * 0 if it cannot be. 3302a13fb1a4SEric Sandeen */ 3303a13fb1a4SEric Sandeen static int ext4_feature_set_ok(struct super_block *sb, int readonly) 3304a13fb1a4SEric Sandeen { 3305e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext4_incompat_features(sb)) { 3306a13fb1a4SEric Sandeen ext4_msg(sb, KERN_ERR, 3307a13fb1a4SEric Sandeen "Couldn't mount because of " 3308a13fb1a4SEric Sandeen "unsupported optional features (%x)", 3309a13fb1a4SEric Sandeen (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 3310a13fb1a4SEric Sandeen ~EXT4_FEATURE_INCOMPAT_SUPP)); 3311a13fb1a4SEric Sandeen return 0; 3312a13fb1a4SEric Sandeen } 3313a13fb1a4SEric Sandeen 3314c83ad55eSGabriel Krisman Bertazi #ifndef CONFIG_UNICODE 3315c83ad55eSGabriel Krisman Bertazi if (ext4_has_feature_casefold(sb)) { 3316c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 3317c83ad55eSGabriel Krisman Bertazi "Filesystem with casefold feature cannot be " 3318c83ad55eSGabriel Krisman Bertazi "mounted without CONFIG_UNICODE"); 3319c83ad55eSGabriel Krisman Bertazi return 0; 3320c83ad55eSGabriel Krisman Bertazi } 3321c83ad55eSGabriel Krisman Bertazi #endif 3322c83ad55eSGabriel Krisman Bertazi 3323a13fb1a4SEric Sandeen if (readonly) 3324a13fb1a4SEric Sandeen return 1; 3325a13fb1a4SEric Sandeen 3326e2b911c5SDarrick J. Wong if (ext4_has_feature_readonly(sb)) { 33272cb5cc8bSDarrick J. Wong ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 33281751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 33292cb5cc8bSDarrick J. Wong return 1; 33302cb5cc8bSDarrick J. Wong } 33312cb5cc8bSDarrick J. Wong 3332a13fb1a4SEric Sandeen /* Check that feature set is OK for a read-write mount */ 3333e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 3334a13fb1a4SEric Sandeen ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 3335a13fb1a4SEric Sandeen "unsupported optional features (%x)", 3336a13fb1a4SEric Sandeen (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 3337a13fb1a4SEric Sandeen ~EXT4_FEATURE_RO_COMPAT_SUPP)); 3338a13fb1a4SEric Sandeen return 0; 3339a13fb1a4SEric Sandeen } 3340e2b911c5SDarrick J. Wong if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 3341bab08ab9STheodore Ts'o ext4_msg(sb, KERN_ERR, 3342bab08ab9STheodore Ts'o "Can't support bigalloc feature without " 3343bab08ab9STheodore Ts'o "extents feature\n"); 3344bab08ab9STheodore Ts'o return 0; 3345bab08ab9STheodore Ts'o } 33467c319d32SAditya Kali 33479db176bcSJan Kara #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2) 3348d65d87a0STheodore Ts'o if (!readonly && (ext4_has_feature_quota(sb) || 3349d65d87a0STheodore Ts'o ext4_has_feature_project(sb))) { 33507c319d32SAditya Kali ext4_msg(sb, KERN_ERR, 3351d65d87a0STheodore Ts'o "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2"); 3352689c958cSLi Xi return 0; 3353689c958cSLi Xi } 33547c319d32SAditya Kali #endif /* CONFIG_QUOTA */ 3355a13fb1a4SEric Sandeen return 1; 3356a13fb1a4SEric Sandeen } 3357a13fb1a4SEric Sandeen 335866e61a9eSTheodore Ts'o /* 335966e61a9eSTheodore Ts'o * This function is called once a day if we have errors logged 336066e61a9eSTheodore Ts'o * on the file system 336166e61a9eSTheodore Ts'o */ 3362235699a8SKees Cook static void print_daily_error_info(struct timer_list *t) 336366e61a9eSTheodore Ts'o { 3364235699a8SKees Cook struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report); 3365235699a8SKees Cook struct super_block *sb = sbi->s_sb; 3366235699a8SKees Cook struct ext4_super_block *es = sbi->s_es; 336766e61a9eSTheodore Ts'o 336866e61a9eSTheodore Ts'o if (es->s_error_count) 3369ae0f78deSTheodore Ts'o /* fsck newer than v1.41.13 is needed to clean this condition. */ 3370ae0f78deSTheodore Ts'o ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 337166e61a9eSTheodore Ts'o le32_to_cpu(es->s_error_count)); 337266e61a9eSTheodore Ts'o if (es->s_first_error_time) { 33736a0678a7SArnd Bergmann printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 33746a0678a7SArnd Bergmann sb->s_id, 33756a0678a7SArnd Bergmann ext4_get_tstamp(es, s_first_error_time), 337666e61a9eSTheodore Ts'o (int) sizeof(es->s_first_error_func), 337766e61a9eSTheodore Ts'o es->s_first_error_func, 337866e61a9eSTheodore Ts'o le32_to_cpu(es->s_first_error_line)); 337966e61a9eSTheodore Ts'o if (es->s_first_error_ino) 3380651e1c3bSJoe Perches printk(KERN_CONT ": inode %u", 338166e61a9eSTheodore Ts'o le32_to_cpu(es->s_first_error_ino)); 338266e61a9eSTheodore Ts'o if (es->s_first_error_block) 3383651e1c3bSJoe Perches printk(KERN_CONT ": block %llu", (unsigned long long) 338466e61a9eSTheodore Ts'o le64_to_cpu(es->s_first_error_block)); 3385651e1c3bSJoe Perches printk(KERN_CONT "\n"); 338666e61a9eSTheodore Ts'o } 338766e61a9eSTheodore Ts'o if (es->s_last_error_time) { 33886a0678a7SArnd Bergmann printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 33896a0678a7SArnd Bergmann sb->s_id, 33906a0678a7SArnd Bergmann ext4_get_tstamp(es, s_last_error_time), 339166e61a9eSTheodore Ts'o (int) sizeof(es->s_last_error_func), 339266e61a9eSTheodore Ts'o es->s_last_error_func, 339366e61a9eSTheodore Ts'o le32_to_cpu(es->s_last_error_line)); 339466e61a9eSTheodore Ts'o if (es->s_last_error_ino) 3395651e1c3bSJoe Perches printk(KERN_CONT ": inode %u", 339666e61a9eSTheodore Ts'o le32_to_cpu(es->s_last_error_ino)); 339766e61a9eSTheodore Ts'o if (es->s_last_error_block) 3398651e1c3bSJoe Perches printk(KERN_CONT ": block %llu", (unsigned long long) 339966e61a9eSTheodore Ts'o le64_to_cpu(es->s_last_error_block)); 3400651e1c3bSJoe Perches printk(KERN_CONT "\n"); 340166e61a9eSTheodore Ts'o } 340266e61a9eSTheodore Ts'o mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 340366e61a9eSTheodore Ts'o } 340466e61a9eSTheodore Ts'o 3405bfff6873SLukas Czerner /* Find next suitable group and run ext4_init_inode_table */ 3406bfff6873SLukas Czerner static int ext4_run_li_request(struct ext4_li_request *elr) 3407bfff6873SLukas Czerner { 3408bfff6873SLukas Czerner struct ext4_group_desc *gdp = NULL; 34093d392b26STheodore Ts'o struct super_block *sb = elr->lr_super; 34103d392b26STheodore Ts'o ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 34113d392b26STheodore Ts'o ext4_group_t group = elr->lr_next_group; 3412bfff6873SLukas Czerner unsigned long timeout = 0; 34133d392b26STheodore Ts'o unsigned int prefetch_ios = 0; 3414bfff6873SLukas Czerner int ret = 0; 3415bfff6873SLukas Czerner 34163d392b26STheodore Ts'o if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) { 34173d392b26STheodore Ts'o elr->lr_next_group = ext4_mb_prefetch(sb, group, 34183d392b26STheodore Ts'o EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios); 34193d392b26STheodore Ts'o if (prefetch_ios) 34203d392b26STheodore Ts'o ext4_mb_prefetch_fini(sb, elr->lr_next_group, 34213d392b26STheodore Ts'o prefetch_ios); 34223d392b26STheodore Ts'o trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, 34233d392b26STheodore Ts'o prefetch_ios); 34243d392b26STheodore Ts'o if (group >= elr->lr_next_group) { 34253d392b26STheodore Ts'o ret = 1; 34263d392b26STheodore Ts'o if (elr->lr_first_not_zeroed != ngroups && 34273d392b26STheodore Ts'o !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) { 34283d392b26STheodore Ts'o elr->lr_next_group = elr->lr_first_not_zeroed; 34293d392b26STheodore Ts'o elr->lr_mode = EXT4_LI_MODE_ITABLE; 34303d392b26STheodore Ts'o ret = 0; 34313d392b26STheodore Ts'o } 34323d392b26STheodore Ts'o } 34333d392b26STheodore Ts'o return ret; 34343d392b26STheodore Ts'o } 3435bfff6873SLukas Czerner 34363d392b26STheodore Ts'o for (; group < ngroups; group++) { 3437bfff6873SLukas Czerner gdp = ext4_get_group_desc(sb, group, NULL); 3438bfff6873SLukas Czerner if (!gdp) { 3439bfff6873SLukas Czerner ret = 1; 3440bfff6873SLukas Czerner break; 3441bfff6873SLukas Czerner } 3442bfff6873SLukas Czerner 3443bfff6873SLukas Czerner if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3444bfff6873SLukas Czerner break; 3445bfff6873SLukas Czerner } 3446bfff6873SLukas Czerner 34477f511862STheodore Ts'o if (group >= ngroups) 3448bfff6873SLukas Czerner ret = 1; 3449bfff6873SLukas Czerner 3450bfff6873SLukas Czerner if (!ret) { 3451bfff6873SLukas Czerner timeout = jiffies; 3452bfff6873SLukas Czerner ret = ext4_init_inode_table(sb, group, 3453bfff6873SLukas Czerner elr->lr_timeout ? 0 : 1); 34543d392b26STheodore Ts'o trace_ext4_lazy_itable_init(sb, group); 3455bfff6873SLukas Czerner if (elr->lr_timeout == 0) { 345651ce6511SLukas Czerner timeout = (jiffies - timeout) * 34573d392b26STheodore Ts'o EXT4_SB(elr->lr_super)->s_li_wait_mult; 3458bfff6873SLukas Czerner elr->lr_timeout = timeout; 3459bfff6873SLukas Czerner } 3460bfff6873SLukas Czerner elr->lr_next_sched = jiffies + elr->lr_timeout; 3461bfff6873SLukas Czerner elr->lr_next_group = group + 1; 3462bfff6873SLukas Czerner } 3463bfff6873SLukas Czerner return ret; 3464bfff6873SLukas Czerner } 3465bfff6873SLukas Czerner 3466bfff6873SLukas Czerner /* 3467bfff6873SLukas Czerner * Remove lr_request from the list_request and free the 34684ed5c033SLukas Czerner * request structure. Should be called with li_list_mtx held 3469bfff6873SLukas Czerner */ 3470bfff6873SLukas Czerner static void ext4_remove_li_request(struct ext4_li_request *elr) 3471bfff6873SLukas Czerner { 3472bfff6873SLukas Czerner if (!elr) 3473bfff6873SLukas Czerner return; 3474bfff6873SLukas Czerner 3475bfff6873SLukas Czerner list_del(&elr->lr_request); 34763d392b26STheodore Ts'o EXT4_SB(elr->lr_super)->s_li_request = NULL; 3477bfff6873SLukas Czerner kfree(elr); 3478bfff6873SLukas Czerner } 3479bfff6873SLukas Czerner 3480bfff6873SLukas Czerner static void ext4_unregister_li_request(struct super_block *sb) 3481bfff6873SLukas Czerner { 34821bb933fbSLukas Czerner mutex_lock(&ext4_li_mtx); 34831bb933fbSLukas Czerner if (!ext4_li_info) { 34841bb933fbSLukas Czerner mutex_unlock(&ext4_li_mtx); 3485bfff6873SLukas Czerner return; 34861bb933fbSLukas Czerner } 3487bfff6873SLukas Czerner 3488bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 34891bb933fbSLukas Czerner ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3490bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 34911bb933fbSLukas Czerner mutex_unlock(&ext4_li_mtx); 3492bfff6873SLukas Czerner } 3493bfff6873SLukas Czerner 34948f1f7453SEric Sandeen static struct task_struct *ext4_lazyinit_task; 34958f1f7453SEric Sandeen 3496bfff6873SLukas Czerner /* 3497bfff6873SLukas Czerner * This is the function where ext4lazyinit thread lives. It walks 3498bfff6873SLukas Czerner * through the request list searching for next scheduled filesystem. 3499bfff6873SLukas Czerner * When such a fs is found, run the lazy initialization request 3500bfff6873SLukas Czerner * (ext4_rn_li_request) and keep track of the time spend in this 3501bfff6873SLukas Czerner * function. Based on that time we compute next schedule time of 3502bfff6873SLukas Czerner * the request. When walking through the list is complete, compute 3503bfff6873SLukas Czerner * next waking time and put itself into sleep. 3504bfff6873SLukas Czerner */ 3505bfff6873SLukas Czerner static int ext4_lazyinit_thread(void *arg) 3506bfff6873SLukas Czerner { 3507bfff6873SLukas Czerner struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 3508bfff6873SLukas Czerner struct list_head *pos, *n; 3509bfff6873SLukas Czerner struct ext4_li_request *elr; 35104ed5c033SLukas Czerner unsigned long next_wakeup, cur; 3511bfff6873SLukas Czerner 3512bfff6873SLukas Czerner BUG_ON(NULL == eli); 3513bfff6873SLukas Czerner 3514bfff6873SLukas Czerner cont_thread: 3515bfff6873SLukas Czerner while (true) { 3516bfff6873SLukas Czerner next_wakeup = MAX_JIFFY_OFFSET; 3517bfff6873SLukas Czerner 3518bfff6873SLukas Czerner mutex_lock(&eli->li_list_mtx); 3519bfff6873SLukas Czerner if (list_empty(&eli->li_request_list)) { 3520bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3521bfff6873SLukas Czerner goto exit_thread; 3522bfff6873SLukas Czerner } 3523bfff6873SLukas Czerner list_for_each_safe(pos, n, &eli->li_request_list) { 3524e22834f0SDmitry Monakhov int err = 0; 3525e22834f0SDmitry Monakhov int progress = 0; 3526bfff6873SLukas Czerner elr = list_entry(pos, struct ext4_li_request, 3527bfff6873SLukas Czerner lr_request); 3528bfff6873SLukas Czerner 3529e22834f0SDmitry Monakhov if (time_before(jiffies, elr->lr_next_sched)) { 3530e22834f0SDmitry Monakhov if (time_before(elr->lr_next_sched, next_wakeup)) 3531e22834f0SDmitry Monakhov next_wakeup = elr->lr_next_sched; 3532e22834f0SDmitry Monakhov continue; 3533e22834f0SDmitry Monakhov } 3534e22834f0SDmitry Monakhov if (down_read_trylock(&elr->lr_super->s_umount)) { 3535e22834f0SDmitry Monakhov if (sb_start_write_trylock(elr->lr_super)) { 3536e22834f0SDmitry Monakhov progress = 1; 3537e22834f0SDmitry Monakhov /* 3538e22834f0SDmitry Monakhov * We hold sb->s_umount, sb can not 3539e22834f0SDmitry Monakhov * be removed from the list, it is 3540e22834f0SDmitry Monakhov * now safe to drop li_list_mtx 3541e22834f0SDmitry Monakhov */ 3542e22834f0SDmitry Monakhov mutex_unlock(&eli->li_list_mtx); 3543e22834f0SDmitry Monakhov err = ext4_run_li_request(elr); 3544e22834f0SDmitry Monakhov sb_end_write(elr->lr_super); 3545e22834f0SDmitry Monakhov mutex_lock(&eli->li_list_mtx); 3546e22834f0SDmitry Monakhov n = pos->next; 3547e22834f0SDmitry Monakhov } 3548e22834f0SDmitry Monakhov up_read((&elr->lr_super->s_umount)); 3549e22834f0SDmitry Monakhov } 3550b2c78cd0STheodore Ts'o /* error, remove the lazy_init job */ 3551e22834f0SDmitry Monakhov if (err) { 3552bfff6873SLukas Czerner ext4_remove_li_request(elr); 3553bfff6873SLukas Czerner continue; 3554bfff6873SLukas Czerner } 3555e22834f0SDmitry Monakhov if (!progress) { 3556e22834f0SDmitry Monakhov elr->lr_next_sched = jiffies + 3557e22834f0SDmitry Monakhov (prandom_u32() 3558e22834f0SDmitry Monakhov % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3559b2c78cd0STheodore Ts'o } 3560bfff6873SLukas Czerner if (time_before(elr->lr_next_sched, next_wakeup)) 3561bfff6873SLukas Czerner next_wakeup = elr->lr_next_sched; 3562bfff6873SLukas Czerner } 3563bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3564bfff6873SLukas Czerner 3565a0acae0eSTejun Heo try_to_freeze(); 3566bfff6873SLukas Czerner 35674ed5c033SLukas Czerner cur = jiffies; 35684ed5c033SLukas Czerner if ((time_after_eq(cur, next_wakeup)) || 3569f4245bd4SLukas Czerner (MAX_JIFFY_OFFSET == next_wakeup)) { 3570bfff6873SLukas Czerner cond_resched(); 3571bfff6873SLukas Czerner continue; 3572bfff6873SLukas Czerner } 3573bfff6873SLukas Czerner 35744ed5c033SLukas Czerner schedule_timeout_interruptible(next_wakeup - cur); 35754ed5c033SLukas Czerner 35768f1f7453SEric Sandeen if (kthread_should_stop()) { 35778f1f7453SEric Sandeen ext4_clear_request_list(); 35788f1f7453SEric Sandeen goto exit_thread; 35798f1f7453SEric Sandeen } 3580bfff6873SLukas Czerner } 3581bfff6873SLukas Czerner 3582bfff6873SLukas Czerner exit_thread: 3583bfff6873SLukas Czerner /* 3584bfff6873SLukas Czerner * It looks like the request list is empty, but we need 3585bfff6873SLukas Czerner * to check it under the li_list_mtx lock, to prevent any 3586bfff6873SLukas Czerner * additions into it, and of course we should lock ext4_li_mtx 3587bfff6873SLukas Czerner * to atomically free the list and ext4_li_info, because at 3588bfff6873SLukas Czerner * this point another ext4 filesystem could be registering 3589bfff6873SLukas Czerner * new one. 3590bfff6873SLukas Czerner */ 3591bfff6873SLukas Czerner mutex_lock(&ext4_li_mtx); 3592bfff6873SLukas Czerner mutex_lock(&eli->li_list_mtx); 3593bfff6873SLukas Czerner if (!list_empty(&eli->li_request_list)) { 3594bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3595bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 3596bfff6873SLukas Czerner goto cont_thread; 3597bfff6873SLukas Czerner } 3598bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3599bfff6873SLukas Czerner kfree(ext4_li_info); 3600bfff6873SLukas Czerner ext4_li_info = NULL; 3601bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 3602bfff6873SLukas Czerner 3603bfff6873SLukas Czerner return 0; 3604bfff6873SLukas Czerner } 3605bfff6873SLukas Czerner 3606bfff6873SLukas Czerner static void ext4_clear_request_list(void) 3607bfff6873SLukas Czerner { 3608bfff6873SLukas Czerner struct list_head *pos, *n; 3609bfff6873SLukas Czerner struct ext4_li_request *elr; 3610bfff6873SLukas Czerner 3611bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 3612bfff6873SLukas Czerner list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3613bfff6873SLukas Czerner elr = list_entry(pos, struct ext4_li_request, 3614bfff6873SLukas Czerner lr_request); 3615bfff6873SLukas Czerner ext4_remove_li_request(elr); 3616bfff6873SLukas Czerner } 3617bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 3618bfff6873SLukas Czerner } 3619bfff6873SLukas Czerner 3620bfff6873SLukas Czerner static int ext4_run_lazyinit_thread(void) 3621bfff6873SLukas Czerner { 36228f1f7453SEric Sandeen ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 36238f1f7453SEric Sandeen ext4_li_info, "ext4lazyinit"); 36248f1f7453SEric Sandeen if (IS_ERR(ext4_lazyinit_task)) { 36258f1f7453SEric Sandeen int err = PTR_ERR(ext4_lazyinit_task); 3626bfff6873SLukas Czerner ext4_clear_request_list(); 3627bfff6873SLukas Czerner kfree(ext4_li_info); 3628bfff6873SLukas Czerner ext4_li_info = NULL; 362992b97816STheodore Ts'o printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3630bfff6873SLukas Czerner "initialization thread\n", 3631bfff6873SLukas Czerner err); 3632bfff6873SLukas Czerner return err; 3633bfff6873SLukas Czerner } 3634bfff6873SLukas Czerner ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3635bfff6873SLukas Czerner return 0; 3636bfff6873SLukas Czerner } 3637bfff6873SLukas Czerner 3638bfff6873SLukas Czerner /* 3639bfff6873SLukas Czerner * Check whether it make sense to run itable init. thread or not. 3640bfff6873SLukas Czerner * If there is at least one uninitialized inode table, return 3641bfff6873SLukas Czerner * corresponding group number, else the loop goes through all 3642bfff6873SLukas Czerner * groups and return total number of groups. 3643bfff6873SLukas Czerner */ 3644bfff6873SLukas Czerner static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3645bfff6873SLukas Czerner { 3646bfff6873SLukas Czerner ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3647bfff6873SLukas Czerner struct ext4_group_desc *gdp = NULL; 3648bfff6873SLukas Czerner 36498844618dSTheodore Ts'o if (!ext4_has_group_desc_csum(sb)) 36508844618dSTheodore Ts'o return ngroups; 36518844618dSTheodore Ts'o 3652bfff6873SLukas Czerner for (group = 0; group < ngroups; group++) { 3653bfff6873SLukas Czerner gdp = ext4_get_group_desc(sb, group, NULL); 3654bfff6873SLukas Czerner if (!gdp) 3655bfff6873SLukas Czerner continue; 3656bfff6873SLukas Czerner 365750122847STheodore Ts'o if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3658bfff6873SLukas Czerner break; 3659bfff6873SLukas Czerner } 3660bfff6873SLukas Czerner 3661bfff6873SLukas Czerner return group; 3662bfff6873SLukas Czerner } 3663bfff6873SLukas Czerner 3664bfff6873SLukas Czerner static int ext4_li_info_new(void) 3665bfff6873SLukas Czerner { 3666bfff6873SLukas Czerner struct ext4_lazy_init *eli = NULL; 3667bfff6873SLukas Czerner 3668bfff6873SLukas Czerner eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3669bfff6873SLukas Czerner if (!eli) 3670bfff6873SLukas Czerner return -ENOMEM; 3671bfff6873SLukas Czerner 3672bfff6873SLukas Czerner INIT_LIST_HEAD(&eli->li_request_list); 3673bfff6873SLukas Czerner mutex_init(&eli->li_list_mtx); 3674bfff6873SLukas Czerner 3675bfff6873SLukas Czerner eli->li_state |= EXT4_LAZYINIT_QUIT; 3676bfff6873SLukas Czerner 3677bfff6873SLukas Czerner ext4_li_info = eli; 3678bfff6873SLukas Czerner 3679bfff6873SLukas Czerner return 0; 3680bfff6873SLukas Czerner } 3681bfff6873SLukas Czerner 3682bfff6873SLukas Czerner static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3683bfff6873SLukas Czerner ext4_group_t start) 3684bfff6873SLukas Czerner { 3685bfff6873SLukas Czerner struct ext4_li_request *elr; 3686bfff6873SLukas Czerner 3687bfff6873SLukas Czerner elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3688bfff6873SLukas Czerner if (!elr) 3689bfff6873SLukas Czerner return NULL; 3690bfff6873SLukas Czerner 3691bfff6873SLukas Czerner elr->lr_super = sb; 36923d392b26STheodore Ts'o elr->lr_first_not_zeroed = start; 36933d392b26STheodore Ts'o if (test_opt(sb, PREFETCH_BLOCK_BITMAPS)) 36943d392b26STheodore Ts'o elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP; 36953d392b26STheodore Ts'o else { 36963d392b26STheodore Ts'o elr->lr_mode = EXT4_LI_MODE_ITABLE; 3697bfff6873SLukas Czerner elr->lr_next_group = start; 36983d392b26STheodore Ts'o } 3699bfff6873SLukas Czerner 3700bfff6873SLukas Czerner /* 3701bfff6873SLukas Czerner * Randomize first schedule time of the request to 3702bfff6873SLukas Czerner * spread the inode table initialization requests 3703bfff6873SLukas Czerner * better. 3704bfff6873SLukas Czerner */ 3705dd1f723bSTheodore Ts'o elr->lr_next_sched = jiffies + (prandom_u32() % 3706dd1f723bSTheodore Ts'o (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3707bfff6873SLukas Czerner return elr; 3708bfff6873SLukas Czerner } 3709bfff6873SLukas Czerner 37107f511862STheodore Ts'o int ext4_register_li_request(struct super_block *sb, 3711bfff6873SLukas Czerner ext4_group_t first_not_zeroed) 3712bfff6873SLukas Czerner { 3713bfff6873SLukas Czerner struct ext4_sb_info *sbi = EXT4_SB(sb); 37147f511862STheodore Ts'o struct ext4_li_request *elr = NULL; 371549598e04SJun Piao ext4_group_t ngroups = sbi->s_groups_count; 37166c5a6cb9SAndrew Morton int ret = 0; 3717bfff6873SLukas Czerner 37187f511862STheodore Ts'o mutex_lock(&ext4_li_mtx); 371951ce6511SLukas Czerner if (sbi->s_li_request != NULL) { 372051ce6511SLukas Czerner /* 372151ce6511SLukas Czerner * Reset timeout so it can be computed again, because 372251ce6511SLukas Czerner * s_li_wait_mult might have changed. 372351ce6511SLukas Czerner */ 372451ce6511SLukas Czerner sbi->s_li_request->lr_timeout = 0; 37257f511862STheodore Ts'o goto out; 372651ce6511SLukas Czerner } 3727bfff6873SLukas Czerner 37283d392b26STheodore Ts'o if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) && 37293d392b26STheodore Ts'o (first_not_zeroed == ngroups || sb_rdonly(sb) || 37303d392b26STheodore Ts'o !test_opt(sb, INIT_INODE_TABLE))) 37317f511862STheodore Ts'o goto out; 3732bfff6873SLukas Czerner 3733bfff6873SLukas Czerner elr = ext4_li_request_new(sb, first_not_zeroed); 37347f511862STheodore Ts'o if (!elr) { 37357f511862STheodore Ts'o ret = -ENOMEM; 37367f511862STheodore Ts'o goto out; 37377f511862STheodore Ts'o } 3738bfff6873SLukas Czerner 3739bfff6873SLukas Czerner if (NULL == ext4_li_info) { 3740bfff6873SLukas Czerner ret = ext4_li_info_new(); 3741bfff6873SLukas Czerner if (ret) 3742bfff6873SLukas Czerner goto out; 3743bfff6873SLukas Czerner } 3744bfff6873SLukas Czerner 3745bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 3746bfff6873SLukas Czerner list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3747bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 3748bfff6873SLukas Czerner 3749bfff6873SLukas Czerner sbi->s_li_request = elr; 375046e4690bSTao Ma /* 375146e4690bSTao Ma * set elr to NULL here since it has been inserted to 375246e4690bSTao Ma * the request_list and the removal and free of it is 375346e4690bSTao Ma * handled by ext4_clear_request_list from now on. 375446e4690bSTao Ma */ 375546e4690bSTao Ma elr = NULL; 3756bfff6873SLukas Czerner 3757bfff6873SLukas Czerner if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3758bfff6873SLukas Czerner ret = ext4_run_lazyinit_thread(); 3759bfff6873SLukas Czerner if (ret) 3760bfff6873SLukas Czerner goto out; 3761bfff6873SLukas Czerner } 3762bfff6873SLukas Czerner out: 3763bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 3764beed5ecbSNicolas Kaiser if (ret) 3765bfff6873SLukas Czerner kfree(elr); 3766bfff6873SLukas Czerner return ret; 3767bfff6873SLukas Czerner } 3768bfff6873SLukas Czerner 3769bfff6873SLukas Czerner /* 3770bfff6873SLukas Czerner * We do not need to lock anything since this is called on 3771bfff6873SLukas Czerner * module unload. 3772bfff6873SLukas Czerner */ 3773bfff6873SLukas Czerner static void ext4_destroy_lazyinit_thread(void) 3774bfff6873SLukas Czerner { 3775bfff6873SLukas Czerner /* 3776bfff6873SLukas Czerner * If thread exited earlier 3777bfff6873SLukas Czerner * there's nothing to be done. 3778bfff6873SLukas Czerner */ 37798f1f7453SEric Sandeen if (!ext4_li_info || !ext4_lazyinit_task) 3780bfff6873SLukas Czerner return; 3781bfff6873SLukas Czerner 37828f1f7453SEric Sandeen kthread_stop(ext4_lazyinit_task); 3783bfff6873SLukas Czerner } 3784bfff6873SLukas Czerner 378525ed6e8aSDarrick J. Wong static int set_journal_csum_feature_set(struct super_block *sb) 378625ed6e8aSDarrick J. Wong { 378725ed6e8aSDarrick J. Wong int ret = 1; 378825ed6e8aSDarrick J. Wong int compat, incompat; 378925ed6e8aSDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 379025ed6e8aSDarrick J. Wong 37919aa5d32bSDmitry Monakhov if (ext4_has_metadata_csum(sb)) { 3792db9ee220SDarrick J. Wong /* journal checksum v3 */ 379325ed6e8aSDarrick J. Wong compat = 0; 3794db9ee220SDarrick J. Wong incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 379525ed6e8aSDarrick J. Wong } else { 379625ed6e8aSDarrick J. Wong /* journal checksum v1 */ 379725ed6e8aSDarrick J. Wong compat = JBD2_FEATURE_COMPAT_CHECKSUM; 379825ed6e8aSDarrick J. Wong incompat = 0; 379925ed6e8aSDarrick J. Wong } 380025ed6e8aSDarrick J. Wong 3801feb8c6d3SDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 3802feb8c6d3SDarrick J. Wong JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3803feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_CSUM_V3 | 3804feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_CSUM_V2); 380525ed6e8aSDarrick J. Wong if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 380625ed6e8aSDarrick J. Wong ret = jbd2_journal_set_features(sbi->s_journal, 380725ed6e8aSDarrick J. Wong compat, 0, 380825ed6e8aSDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 380925ed6e8aSDarrick J. Wong incompat); 381025ed6e8aSDarrick J. Wong } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 381125ed6e8aSDarrick J. Wong ret = jbd2_journal_set_features(sbi->s_journal, 381225ed6e8aSDarrick J. Wong compat, 0, 381325ed6e8aSDarrick J. Wong incompat); 381425ed6e8aSDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 0, 0, 381525ed6e8aSDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 381625ed6e8aSDarrick J. Wong } else { 3817feb8c6d3SDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3818feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 381925ed6e8aSDarrick J. Wong } 382025ed6e8aSDarrick J. Wong 382125ed6e8aSDarrick J. Wong return ret; 382225ed6e8aSDarrick J. Wong } 382325ed6e8aSDarrick J. Wong 3824952fc18eSTheodore Ts'o /* 3825952fc18eSTheodore Ts'o * Note: calculating the overhead so we can be compatible with 3826952fc18eSTheodore Ts'o * historical BSD practice is quite difficult in the face of 3827952fc18eSTheodore Ts'o * clusters/bigalloc. This is because multiple metadata blocks from 3828952fc18eSTheodore Ts'o * different block group can end up in the same allocation cluster. 3829952fc18eSTheodore Ts'o * Calculating the exact overhead in the face of clustered allocation 3830952fc18eSTheodore Ts'o * requires either O(all block bitmaps) in memory or O(number of block 3831952fc18eSTheodore Ts'o * groups**2) in time. We will still calculate the superblock for 3832952fc18eSTheodore Ts'o * older file systems --- and if we come across with a bigalloc file 3833952fc18eSTheodore Ts'o * system with zero in s_overhead_clusters the estimate will be close to 3834952fc18eSTheodore Ts'o * correct especially for very large cluster sizes --- but for newer 3835952fc18eSTheodore Ts'o * file systems, it's better to calculate this figure once at mkfs 3836952fc18eSTheodore Ts'o * time, and store it in the superblock. If the superblock value is 3837952fc18eSTheodore Ts'o * present (even for non-bigalloc file systems), we will use it. 3838952fc18eSTheodore Ts'o */ 3839952fc18eSTheodore Ts'o static int count_overhead(struct super_block *sb, ext4_group_t grp, 3840952fc18eSTheodore Ts'o char *buf) 3841952fc18eSTheodore Ts'o { 3842952fc18eSTheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 3843952fc18eSTheodore Ts'o struct ext4_group_desc *gdp; 3844952fc18eSTheodore Ts'o ext4_fsblk_t first_block, last_block, b; 3845952fc18eSTheodore Ts'o ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3846952fc18eSTheodore Ts'o int s, j, count = 0; 3847952fc18eSTheodore Ts'o 3848e2b911c5SDarrick J. Wong if (!ext4_has_feature_bigalloc(sb)) 38490548bbb8STheodore Ts'o return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 38500548bbb8STheodore Ts'o sbi->s_itb_per_group + 2); 38510548bbb8STheodore Ts'o 3852952fc18eSTheodore Ts'o first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3853952fc18eSTheodore Ts'o (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3854952fc18eSTheodore Ts'o last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3855952fc18eSTheodore Ts'o for (i = 0; i < ngroups; i++) { 3856952fc18eSTheodore Ts'o gdp = ext4_get_group_desc(sb, i, NULL); 3857952fc18eSTheodore Ts'o b = ext4_block_bitmap(sb, gdp); 3858952fc18eSTheodore Ts'o if (b >= first_block && b <= last_block) { 3859952fc18eSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3860952fc18eSTheodore Ts'o count++; 3861952fc18eSTheodore Ts'o } 3862952fc18eSTheodore Ts'o b = ext4_inode_bitmap(sb, gdp); 3863952fc18eSTheodore Ts'o if (b >= first_block && b <= last_block) { 3864952fc18eSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3865952fc18eSTheodore Ts'o count++; 3866952fc18eSTheodore Ts'o } 3867952fc18eSTheodore Ts'o b = ext4_inode_table(sb, gdp); 3868952fc18eSTheodore Ts'o if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3869952fc18eSTheodore Ts'o for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3870952fc18eSTheodore Ts'o int c = EXT4_B2C(sbi, b - first_block); 3871952fc18eSTheodore Ts'o ext4_set_bit(c, buf); 3872952fc18eSTheodore Ts'o count++; 3873952fc18eSTheodore Ts'o } 3874952fc18eSTheodore Ts'o if (i != grp) 3875952fc18eSTheodore Ts'o continue; 3876952fc18eSTheodore Ts'o s = 0; 3877952fc18eSTheodore Ts'o if (ext4_bg_has_super(sb, grp)) { 3878952fc18eSTheodore Ts'o ext4_set_bit(s++, buf); 3879952fc18eSTheodore Ts'o count++; 3880952fc18eSTheodore Ts'o } 3881c48ae41bSTheodore Ts'o j = ext4_bg_num_gdb(sb, grp); 3882c48ae41bSTheodore Ts'o if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 3883c48ae41bSTheodore Ts'o ext4_error(sb, "Invalid number of block group " 3884c48ae41bSTheodore Ts'o "descriptor blocks: %d", j); 3885c48ae41bSTheodore Ts'o j = EXT4_BLOCKS_PER_GROUP(sb) - s; 3886952fc18eSTheodore Ts'o } 3887c48ae41bSTheodore Ts'o count += j; 3888c48ae41bSTheodore Ts'o for (; j > 0; j--) 3889c48ae41bSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3890952fc18eSTheodore Ts'o } 3891952fc18eSTheodore Ts'o if (!count) 3892952fc18eSTheodore Ts'o return 0; 3893952fc18eSTheodore Ts'o return EXT4_CLUSTERS_PER_GROUP(sb) - 3894952fc18eSTheodore Ts'o ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3895952fc18eSTheodore Ts'o } 3896952fc18eSTheodore Ts'o 3897952fc18eSTheodore Ts'o /* 3898952fc18eSTheodore Ts'o * Compute the overhead and stash it in sbi->s_overhead 3899952fc18eSTheodore Ts'o */ 3900952fc18eSTheodore Ts'o int ext4_calculate_overhead(struct super_block *sb) 3901952fc18eSTheodore Ts'o { 3902952fc18eSTheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 3903952fc18eSTheodore Ts'o struct ext4_super_block *es = sbi->s_es; 39043c816dedSEric Whitney struct inode *j_inode; 39053c816dedSEric Whitney unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 3906952fc18eSTheodore Ts'o ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3907952fc18eSTheodore Ts'o ext4_fsblk_t overhead = 0; 39084fdb5543SDmitry Monakhov char *buf = (char *) get_zeroed_page(GFP_NOFS); 3909952fc18eSTheodore Ts'o 3910952fc18eSTheodore Ts'o if (!buf) 3911952fc18eSTheodore Ts'o return -ENOMEM; 3912952fc18eSTheodore Ts'o 3913952fc18eSTheodore Ts'o /* 3914952fc18eSTheodore Ts'o * Compute the overhead (FS structures). This is constant 3915952fc18eSTheodore Ts'o * for a given filesystem unless the number of block groups 3916952fc18eSTheodore Ts'o * changes so we cache the previous value until it does. 3917952fc18eSTheodore Ts'o */ 3918952fc18eSTheodore Ts'o 3919952fc18eSTheodore Ts'o /* 3920952fc18eSTheodore Ts'o * All of the blocks before first_data_block are overhead 3921952fc18eSTheodore Ts'o */ 3922952fc18eSTheodore Ts'o overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3923952fc18eSTheodore Ts'o 3924952fc18eSTheodore Ts'o /* 3925952fc18eSTheodore Ts'o * Add the overhead found in each block group 3926952fc18eSTheodore Ts'o */ 3927952fc18eSTheodore Ts'o for (i = 0; i < ngroups; i++) { 3928952fc18eSTheodore Ts'o int blks; 3929952fc18eSTheodore Ts'o 3930952fc18eSTheodore Ts'o blks = count_overhead(sb, i, buf); 3931952fc18eSTheodore Ts'o overhead += blks; 3932952fc18eSTheodore Ts'o if (blks) 3933952fc18eSTheodore Ts'o memset(buf, 0, PAGE_SIZE); 3934952fc18eSTheodore Ts'o cond_resched(); 3935952fc18eSTheodore Ts'o } 39363c816dedSEric Whitney 39373c816dedSEric Whitney /* 39383c816dedSEric Whitney * Add the internal journal blocks whether the journal has been 39393c816dedSEric Whitney * loaded or not 39403c816dedSEric Whitney */ 3941ee7ed3aaSChunguang Xu if (sbi->s_journal && !sbi->s_journal_bdev) 3942ede7dc7fSHarshad Shirwadkar overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len); 3943f1eec3b0SRitesh Harjani else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) { 3944f1eec3b0SRitesh Harjani /* j_inum for internal journal is non-zero */ 39453c816dedSEric Whitney j_inode = ext4_get_journal_inode(sb, j_inum); 39463c816dedSEric Whitney if (j_inode) { 39473c816dedSEric Whitney j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 39483c816dedSEric Whitney overhead += EXT4_NUM_B2C(sbi, j_blocks); 39493c816dedSEric Whitney iput(j_inode); 39503c816dedSEric Whitney } else { 39513c816dedSEric Whitney ext4_msg(sb, KERN_ERR, "can't get journal size"); 39523c816dedSEric Whitney } 39533c816dedSEric Whitney } 3954952fc18eSTheodore Ts'o sbi->s_overhead = overhead; 3955952fc18eSTheodore Ts'o smp_wmb(); 3956952fc18eSTheodore Ts'o free_page((unsigned long) buf); 3957952fc18eSTheodore Ts'o return 0; 3958952fc18eSTheodore Ts'o } 3959952fc18eSTheodore Ts'o 3960b5799018STheodore Ts'o static void ext4_set_resv_clusters(struct super_block *sb) 396127dd4385SLukas Czerner { 396227dd4385SLukas Czerner ext4_fsblk_t resv_clusters; 3963b5799018STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 396427dd4385SLukas Czerner 396527dd4385SLukas Czerner /* 396630fac0f7SJan Kara * There's no need to reserve anything when we aren't using extents. 396730fac0f7SJan Kara * The space estimates are exact, there are no unwritten extents, 396830fac0f7SJan Kara * hole punching doesn't need new metadata... This is needed especially 396930fac0f7SJan Kara * to keep ext2/3 backward compatibility. 397030fac0f7SJan Kara */ 3971e2b911c5SDarrick J. Wong if (!ext4_has_feature_extents(sb)) 3972b5799018STheodore Ts'o return; 397330fac0f7SJan Kara /* 397427dd4385SLukas Czerner * By default we reserve 2% or 4096 clusters, whichever is smaller. 397527dd4385SLukas Czerner * This should cover the situations where we can not afford to run 397627dd4385SLukas Czerner * out of space like for example punch hole, or converting 3977556615dcSLukas Czerner * unwritten extents in delalloc path. In most cases such 397827dd4385SLukas Czerner * allocation would require 1, or 2 blocks, higher numbers are 397927dd4385SLukas Czerner * very rare. 398027dd4385SLukas Czerner */ 3981b5799018STheodore Ts'o resv_clusters = (ext4_blocks_count(sbi->s_es) >> 3982b5799018STheodore Ts'o sbi->s_cluster_bits); 398327dd4385SLukas Czerner 398427dd4385SLukas Czerner do_div(resv_clusters, 50); 398527dd4385SLukas Czerner resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 398627dd4385SLukas Czerner 3987b5799018STheodore Ts'o atomic64_set(&sbi->s_resv_clusters, resv_clusters); 398827dd4385SLukas Czerner } 398927dd4385SLukas Czerner 3990ca9b404fSRoman Anufriev static const char *ext4_quota_mode(struct super_block *sb) 3991ca9b404fSRoman Anufriev { 3992ca9b404fSRoman Anufriev #ifdef CONFIG_QUOTA 3993ca9b404fSRoman Anufriev if (!ext4_quota_capable(sb)) 3994ca9b404fSRoman Anufriev return "none"; 3995ca9b404fSRoman Anufriev 3996ca9b404fSRoman Anufriev if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb)) 3997ca9b404fSRoman Anufriev return "journalled"; 3998ca9b404fSRoman Anufriev else 3999ca9b404fSRoman Anufriev return "writeback"; 4000ca9b404fSRoman Anufriev #else 4001ca9b404fSRoman Anufriev return "disabled"; 4002ca9b404fSRoman Anufriev #endif 4003ca9b404fSRoman Anufriev } 4004ca9b404fSRoman Anufriev 4005617ba13bSMingming Cao static int ext4_fill_super(struct super_block *sb, void *data, int silent) 4006ac27a0ecSDave Kleikamp { 40075e405595SDan Williams struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev); 4008d4c402d9SCurt Wohlgemuth char *orig_data = kstrdup(data, GFP_KERNEL); 40091d0c3924STheodore Ts'o struct buffer_head *bh, **group_desc; 4010617ba13bSMingming Cao struct ext4_super_block *es = NULL; 40115aee0f8aSTheodore Ts'o struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 40127c990728SSuraj Jitindar Singh struct flex_groups **flex_groups; 4013617ba13bSMingming Cao ext4_fsblk_t block; 4014617ba13bSMingming Cao ext4_fsblk_t sb_block = get_sb_block(&data); 401570bbb3e0SAndrew Morton ext4_fsblk_t logical_sb_block; 4016ac27a0ecSDave Kleikamp unsigned long offset = 0; 4017ac27a0ecSDave Kleikamp unsigned long journal_devnum = 0; 4018ac27a0ecSDave Kleikamp unsigned long def_mount_opts; 4019ac27a0ecSDave Kleikamp struct inode *root; 40200390131bSFrank Mayhar const char *descr; 4021dcc7dae3SCyrill Gorcunov int ret = -ENOMEM; 4022281b5995STheodore Ts'o int blocksize, clustersize; 40234ec11028STheodore Ts'o unsigned int db_count; 40244ec11028STheodore Ts'o unsigned int i; 4025ef5fd681SKaixu Xia int needs_recovery, has_huge_files; 4026bd81d8eeSLaurent Vivier __u64 blocks_count; 402707aa2ea1SLukas Czerner int err = 0; 4028b3881f74STheodore Ts'o unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 4029bfff6873SLukas Czerner ext4_group_t first_not_zeroed; 4030ac27a0ecSDave Kleikamp 40315aee0f8aSTheodore Ts'o if ((data && !orig_data) || !sbi) 40325aee0f8aSTheodore Ts'o goto out_free_base; 4033705895b6SPekka Enberg 4034aed9eb1bSColin Ian King sbi->s_daxdev = dax_dev; 4035705895b6SPekka Enberg sbi->s_blockgroup_lock = 4036705895b6SPekka Enberg kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 40375aee0f8aSTheodore Ts'o if (!sbi->s_blockgroup_lock) 40385aee0f8aSTheodore Ts'o goto out_free_base; 40395aee0f8aSTheodore Ts'o 4040ac27a0ecSDave Kleikamp sb->s_fs_info = sbi; 40412c0544b2STheodore Ts'o sbi->s_sb = sb; 4042240799cdSTheodore Ts'o sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 4043d9c9bef1SMiklos Szeredi sbi->s_sb_block = sb_block; 4044f613dfcbSTheodore Ts'o sbi->s_sectors_written_start = 40458446fe92SChristoph Hellwig part_stat_read(sb->s_bdev, sectors[STAT_WRITE]); 4046ac27a0ecSDave Kleikamp 40479f6200bbSTheodore Ts'o /* Cleanup superblock name */ 4048ec3904dcSRasmus Villemoes strreplace(sb->s_id, '/', '!'); 40499f6200bbSTheodore Ts'o 405007aa2ea1SLukas Czerner /* -EINVAL is default */ 4051dcc7dae3SCyrill Gorcunov ret = -EINVAL; 4052617ba13bSMingming Cao blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 4053ac27a0ecSDave Kleikamp if (!blocksize) { 4054b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 4055ac27a0ecSDave Kleikamp goto out_fail; 4056ac27a0ecSDave Kleikamp } 4057ac27a0ecSDave Kleikamp 4058ac27a0ecSDave Kleikamp /* 4059617ba13bSMingming Cao * The ext4 superblock will not be buffer aligned for other than 1kB 4060ac27a0ecSDave Kleikamp * block sizes. We need to calculate the offset from buffer start. 4061ac27a0ecSDave Kleikamp */ 4062617ba13bSMingming Cao if (blocksize != EXT4_MIN_BLOCK_SIZE) { 406370bbb3e0SAndrew Morton logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 406470bbb3e0SAndrew Morton offset = do_div(logical_sb_block, blocksize); 4065ac27a0ecSDave Kleikamp } else { 406670bbb3e0SAndrew Morton logical_sb_block = sb_block; 4067ac27a0ecSDave Kleikamp } 4068ac27a0ecSDave Kleikamp 40698394a6abSzhangyi (F) bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 40708394a6abSzhangyi (F) if (IS_ERR(bh)) { 4071b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "unable to read superblock"); 40728394a6abSzhangyi (F) ret = PTR_ERR(bh); 4073ac27a0ecSDave Kleikamp goto out_fail; 4074ac27a0ecSDave Kleikamp } 4075ac27a0ecSDave Kleikamp /* 4076ac27a0ecSDave Kleikamp * Note: s_es must be initialized as soon as possible because 4077617ba13bSMingming Cao * some ext4 macro-instructions depend on its value 4078ac27a0ecSDave Kleikamp */ 40792716b802STheodore Ts'o es = (struct ext4_super_block *) (bh->b_data + offset); 4080ac27a0ecSDave Kleikamp sbi->s_es = es; 4081ac27a0ecSDave Kleikamp sb->s_magic = le16_to_cpu(es->s_magic); 4082617ba13bSMingming Cao if (sb->s_magic != EXT4_SUPER_MAGIC) 4083617ba13bSMingming Cao goto cantfind_ext4; 4084afc32f7eSTheodore Ts'o sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 4085ac27a0ecSDave Kleikamp 4086feb0ab32SDarrick J. Wong /* Warn if metadata_csum and gdt_csum are both set. */ 4087e2b911c5SDarrick J. Wong if (ext4_has_feature_metadata_csum(sb) && 4088e2b911c5SDarrick J. Wong ext4_has_feature_gdt_csum(sb)) 4089363307e6SJakub Wilk ext4_warning(sb, "metadata_csum and uninit_bg are " 4090feb0ab32SDarrick J. Wong "redundant flags; please run fsck."); 4091feb0ab32SDarrick J. Wong 4092d25425f8SDarrick J. Wong /* Check for a known checksum algorithm */ 4093d25425f8SDarrick J. Wong if (!ext4_verify_csum_type(sb, es)) { 4094d25425f8SDarrick J. Wong ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 4095d25425f8SDarrick J. Wong "unknown checksum algorithm."); 4096d25425f8SDarrick J. Wong silent = 1; 4097d25425f8SDarrick J. Wong goto cantfind_ext4; 4098d25425f8SDarrick J. Wong } 4099d25425f8SDarrick J. Wong 41000441984aSDarrick J. Wong /* Load the checksum driver */ 41010441984aSDarrick J. Wong sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 41020441984aSDarrick J. Wong if (IS_ERR(sbi->s_chksum_driver)) { 41030441984aSDarrick J. Wong ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 41040441984aSDarrick J. Wong ret = PTR_ERR(sbi->s_chksum_driver); 41050441984aSDarrick J. Wong sbi->s_chksum_driver = NULL; 41060441984aSDarrick J. Wong goto failed_mount; 41070441984aSDarrick J. Wong } 41080441984aSDarrick J. Wong 4109a9c47317SDarrick J. Wong /* Check superblock checksum */ 4110a9c47317SDarrick J. Wong if (!ext4_superblock_csum_verify(sb, es)) { 4111a9c47317SDarrick J. Wong ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 4112a9c47317SDarrick J. Wong "invalid superblock checksum. Run e2fsck?"); 4113a9c47317SDarrick J. Wong silent = 1; 41146a797d27SDarrick J. Wong ret = -EFSBADCRC; 4115a9c47317SDarrick J. Wong goto cantfind_ext4; 4116a9c47317SDarrick J. Wong } 4117a9c47317SDarrick J. Wong 4118a9c47317SDarrick J. Wong /* Precompute checksum seed for all metadata */ 4119e2b911c5SDarrick J. Wong if (ext4_has_feature_csum_seed(sb)) 41208c81bd8fSDarrick J. Wong sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 4121dec214d0STahsin Erdogan else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) 4122a9c47317SDarrick J. Wong sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 4123a9c47317SDarrick J. Wong sizeof(es->s_uuid)); 4124a9c47317SDarrick J. Wong 4125ac27a0ecSDave Kleikamp /* Set defaults before we parse the mount options */ 4126ac27a0ecSDave Kleikamp def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 4127fd8c37ecSTheodore Ts'o set_opt(sb, INIT_INODE_TABLE); 4128617ba13bSMingming Cao if (def_mount_opts & EXT4_DEFM_DEBUG) 4129fd8c37ecSTheodore Ts'o set_opt(sb, DEBUG); 413087f26807STheodore Ts'o if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 4131fd8c37ecSTheodore Ts'o set_opt(sb, GRPID); 4132617ba13bSMingming Cao if (def_mount_opts & EXT4_DEFM_UID16) 4133fd8c37ecSTheodore Ts'o set_opt(sb, NO_UID32); 4134ea663336SEric Sandeen /* xattr user namespace & acls are now defaulted on */ 4135fd8c37ecSTheodore Ts'o set_opt(sb, XATTR_USER); 413603010a33STheodore Ts'o #ifdef CONFIG_EXT4_FS_POSIX_ACL 4137fd8c37ecSTheodore Ts'o set_opt(sb, POSIX_ACL); 41382e7842b8SHugh Dickins #endif 4139995a3ed6SHarshad Shirwadkar if (ext4_has_feature_fast_commit(sb)) 4140995a3ed6SHarshad Shirwadkar set_opt2(sb, JOURNAL_FAST_COMMIT); 414198c1a759SDarrick J. Wong /* don't forget to enable journal_csum when metadata_csum is enabled. */ 414298c1a759SDarrick J. Wong if (ext4_has_metadata_csum(sb)) 414398c1a759SDarrick J. Wong set_opt(sb, JOURNAL_CHECKSUM); 414498c1a759SDarrick J. Wong 4145617ba13bSMingming Cao if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 4146fd8c37ecSTheodore Ts'o set_opt(sb, JOURNAL_DATA); 4147617ba13bSMingming Cao else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 4148fd8c37ecSTheodore Ts'o set_opt(sb, ORDERED_DATA); 4149617ba13bSMingming Cao else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 4150fd8c37ecSTheodore Ts'o set_opt(sb, WRITEBACK_DATA); 4151ac27a0ecSDave Kleikamp 4152617ba13bSMingming Cao if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 4153fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_PANIC); 4154bb4f397aSAneesh Kumar K.V else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 4155fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_CONT); 4156bb4f397aSAneesh Kumar K.V else 4157fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_RO); 415845f1a9c3SDarrick J. Wong /* block_validity enabled by default; disable with noblock_validity */ 4159fd8c37ecSTheodore Ts'o set_opt(sb, BLOCK_VALIDITY); 41608b67f04aSTheodore Ts'o if (def_mount_opts & EXT4_DEFM_DISCARD) 4161fd8c37ecSTheodore Ts'o set_opt(sb, DISCARD); 4162ac27a0ecSDave Kleikamp 416308cefc7aSEric W. Biederman sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 416408cefc7aSEric W. Biederman sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 416530773840STheodore Ts'o sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 416630773840STheodore Ts'o sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 416730773840STheodore Ts'o sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 4168ac27a0ecSDave Kleikamp 41698b67f04aSTheodore Ts'o if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 4170fd8c37ecSTheodore Ts'o set_opt(sb, BARRIER); 4171ac27a0ecSDave Kleikamp 41721e2462f9SMingming Cao /* 4173dd919b98SAneesh Kumar K.V * enable delayed allocation by default 4174dd919b98SAneesh Kumar K.V * Use -o nodelalloc to turn it off 4175dd919b98SAneesh Kumar K.V */ 4176bc0b75f7STheodore Ts'o if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 41778b67f04aSTheodore Ts'o ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 4178fd8c37ecSTheodore Ts'o set_opt(sb, DELALLOC); 4179dd919b98SAneesh Kumar K.V 418051ce6511SLukas Czerner /* 418151ce6511SLukas Czerner * set default s_li_wait_mult for lazyinit, for the case there is 418251ce6511SLukas Czerner * no mount option specified. 418351ce6511SLukas Czerner */ 418451ce6511SLukas Czerner sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 418551ce6511SLukas Czerner 4186c9200760STheodore Ts'o if (le32_to_cpu(es->s_log_block_size) > 4187c9200760STheodore Ts'o (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 4188c9200760STheodore Ts'o ext4_msg(sb, KERN_ERR, 4189c9200760STheodore Ts'o "Invalid log block size: %u", 4190c9200760STheodore Ts'o le32_to_cpu(es->s_log_block_size)); 4191c9200760STheodore Ts'o goto failed_mount; 4192c9200760STheodore Ts'o } 4193c9200760STheodore Ts'o if (le32_to_cpu(es->s_log_cluster_size) > 4194c9200760STheodore Ts'o (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 4195c9200760STheodore Ts'o ext4_msg(sb, KERN_ERR, 4196c9200760STheodore Ts'o "Invalid log cluster size: %u", 4197c9200760STheodore Ts'o le32_to_cpu(es->s_log_cluster_size)); 4198c9200760STheodore Ts'o goto failed_mount; 4199c9200760STheodore Ts'o } 4200c9200760STheodore Ts'o 4201c9200760STheodore Ts'o blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 4202626b035bSRitesh Harjani 4203626b035bSRitesh Harjani if (blocksize == PAGE_SIZE) 4204626b035bSRitesh Harjani set_opt(sb, DIOREAD_NOLOCK); 4205626b035bSRitesh Harjani 42069803387cSTheodore Ts'o if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 42079803387cSTheodore Ts'o sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 42089803387cSTheodore Ts'o sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 42099803387cSTheodore Ts'o } else { 42109803387cSTheodore Ts'o sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 42119803387cSTheodore Ts'o sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 42129803387cSTheodore Ts'o if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 42139803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 42149803387cSTheodore Ts'o sbi->s_first_ino); 42159803387cSTheodore Ts'o goto failed_mount; 42169803387cSTheodore Ts'o } 42179803387cSTheodore Ts'o if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 42189803387cSTheodore Ts'o (!is_power_of_2(sbi->s_inode_size)) || 42199803387cSTheodore Ts'o (sbi->s_inode_size > blocksize)) { 42209803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 42219803387cSTheodore Ts'o "unsupported inode size: %d", 42229803387cSTheodore Ts'o sbi->s_inode_size); 42234f97a681STheodore Ts'o ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize); 42249803387cSTheodore Ts'o goto failed_mount; 42259803387cSTheodore Ts'o } 42269803387cSTheodore Ts'o /* 42279803387cSTheodore Ts'o * i_atime_extra is the last extra field available for 42289803387cSTheodore Ts'o * [acm]times in struct ext4_inode. Checking for that 42299803387cSTheodore Ts'o * field should suffice to ensure we have extra space 42309803387cSTheodore Ts'o * for all three. 42319803387cSTheodore Ts'o */ 42329803387cSTheodore Ts'o if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) + 42339803387cSTheodore Ts'o sizeof(((struct ext4_inode *)0)->i_atime_extra)) { 42349803387cSTheodore Ts'o sb->s_time_gran = 1; 42359803387cSTheodore Ts'o sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX; 42369803387cSTheodore Ts'o } else { 42379803387cSTheodore Ts'o sb->s_time_gran = NSEC_PER_SEC; 42389803387cSTheodore Ts'o sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX; 42399803387cSTheodore Ts'o } 42409803387cSTheodore Ts'o sb->s_time_min = EXT4_TIMESTAMP_MIN; 42419803387cSTheodore Ts'o } 42429803387cSTheodore Ts'o if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 42439803387cSTheodore Ts'o sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 42449803387cSTheodore Ts'o EXT4_GOOD_OLD_INODE_SIZE; 42459803387cSTheodore Ts'o if (ext4_has_feature_extra_isize(sb)) { 42469803387cSTheodore Ts'o unsigned v, max = (sbi->s_inode_size - 42479803387cSTheodore Ts'o EXT4_GOOD_OLD_INODE_SIZE); 42489803387cSTheodore Ts'o 42499803387cSTheodore Ts'o v = le16_to_cpu(es->s_want_extra_isize); 42509803387cSTheodore Ts'o if (v > max) { 42519803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 42529803387cSTheodore Ts'o "bad s_want_extra_isize: %d", v); 42539803387cSTheodore Ts'o goto failed_mount; 42549803387cSTheodore Ts'o } 42559803387cSTheodore Ts'o if (sbi->s_want_extra_isize < v) 42569803387cSTheodore Ts'o sbi->s_want_extra_isize = v; 42579803387cSTheodore Ts'o 42589803387cSTheodore Ts'o v = le16_to_cpu(es->s_min_extra_isize); 42599803387cSTheodore Ts'o if (v > max) { 42609803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 42619803387cSTheodore Ts'o "bad s_min_extra_isize: %d", v); 42629803387cSTheodore Ts'o goto failed_mount; 42639803387cSTheodore Ts'o } 42649803387cSTheodore Ts'o if (sbi->s_want_extra_isize < v) 42659803387cSTheodore Ts'o sbi->s_want_extra_isize = v; 42669803387cSTheodore Ts'o } 42679803387cSTheodore Ts'o } 42689803387cSTheodore Ts'o 42695aee0f8aSTheodore Ts'o if (sbi->s_es->s_mount_opts[0]) { 42705aee0f8aSTheodore Ts'o char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, 42715aee0f8aSTheodore Ts'o sizeof(sbi->s_es->s_mount_opts), 42725aee0f8aSTheodore Ts'o GFP_KERNEL); 42735aee0f8aSTheodore Ts'o if (!s_mount_opts) 42745aee0f8aSTheodore Ts'o goto failed_mount; 42755aee0f8aSTheodore Ts'o if (!parse_options(s_mount_opts, sb, &journal_devnum, 42765aee0f8aSTheodore Ts'o &journal_ioprio, 0)) { 42778b67f04aSTheodore Ts'o ext4_msg(sb, KERN_WARNING, 42788b67f04aSTheodore Ts'o "failed to parse options in superblock: %s", 42795aee0f8aSTheodore Ts'o s_mount_opts); 42805aee0f8aSTheodore Ts'o } 42815aee0f8aSTheodore Ts'o kfree(s_mount_opts); 42828b67f04aSTheodore Ts'o } 42835a916be1STheodore Ts'o sbi->s_def_mount_opt = sbi->s_mount_opt; 4284b3881f74STheodore Ts'o if (!parse_options((char *) data, sb, &journal_devnum, 4285661aa520SEric Sandeen &journal_ioprio, 0)) 4286ac27a0ecSDave Kleikamp goto failed_mount; 4287ac27a0ecSDave Kleikamp 4288c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 4289f8f4acb6SDaniel Rosenberg if (ext4_has_feature_casefold(sb) && !sb->s_encoding) { 4290c83ad55eSGabriel Krisman Bertazi const struct ext4_sb_encodings *encoding_info; 4291c83ad55eSGabriel Krisman Bertazi struct unicode_map *encoding; 4292c83ad55eSGabriel Krisman Bertazi __u16 encoding_flags; 4293c83ad55eSGabriel Krisman Bertazi 4294c83ad55eSGabriel Krisman Bertazi if (ext4_has_feature_encrypt(sb)) { 4295c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 4296c83ad55eSGabriel Krisman Bertazi "Can't mount with encoding and encryption"); 4297c83ad55eSGabriel Krisman Bertazi goto failed_mount; 4298c83ad55eSGabriel Krisman Bertazi } 4299c83ad55eSGabriel Krisman Bertazi 4300c83ad55eSGabriel Krisman Bertazi if (ext4_sb_read_encoding(es, &encoding_info, 4301c83ad55eSGabriel Krisman Bertazi &encoding_flags)) { 4302c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 4303c83ad55eSGabriel Krisman Bertazi "Encoding requested by superblock is unknown"); 4304c83ad55eSGabriel Krisman Bertazi goto failed_mount; 4305c83ad55eSGabriel Krisman Bertazi } 4306c83ad55eSGabriel Krisman Bertazi 4307c83ad55eSGabriel Krisman Bertazi encoding = utf8_load(encoding_info->version); 4308c83ad55eSGabriel Krisman Bertazi if (IS_ERR(encoding)) { 4309c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 4310c83ad55eSGabriel Krisman Bertazi "can't mount with superblock charset: %s-%s " 4311c83ad55eSGabriel Krisman Bertazi "not supported by the kernel. flags: 0x%x.", 4312c83ad55eSGabriel Krisman Bertazi encoding_info->name, encoding_info->version, 4313c83ad55eSGabriel Krisman Bertazi encoding_flags); 4314c83ad55eSGabriel Krisman Bertazi goto failed_mount; 4315c83ad55eSGabriel Krisman Bertazi } 4316c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " 4317c83ad55eSGabriel Krisman Bertazi "%s-%s with flags 0x%hx", encoding_info->name, 4318c83ad55eSGabriel Krisman Bertazi encoding_info->version?:"\b", encoding_flags); 4319c83ad55eSGabriel Krisman Bertazi 4320f8f4acb6SDaniel Rosenberg sb->s_encoding = encoding; 4321f8f4acb6SDaniel Rosenberg sb->s_encoding_flags = encoding_flags; 4322c83ad55eSGabriel Krisman Bertazi } 4323c83ad55eSGabriel Krisman Bertazi #endif 4324c83ad55eSGabriel Krisman Bertazi 432556889787STheodore Ts'o if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 4326556e0319SHarshad Shirwadkar printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n"); 4327781c036bSKaixu Xia /* can't mount with both data=journal and dioread_nolock. */ 4328244adf64STheodore Ts'o clear_opt(sb, DIOREAD_NOLOCK); 4329556e0319SHarshad Shirwadkar clear_opt2(sb, JOURNAL_FAST_COMMIT); 433056889787STheodore Ts'o if (test_opt2(sb, EXPLICIT_DELALLOC)) { 433156889787STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't mount with " 433256889787STheodore Ts'o "both data=journal and delalloc"); 433356889787STheodore Ts'o goto failed_mount; 433456889787STheodore Ts'o } 4335fc626fe3SIra Weiny if (test_opt(sb, DAX_ALWAYS)) { 4336923ae0ffSRoss Zwisler ext4_msg(sb, KERN_ERR, "can't mount with " 4337923ae0ffSRoss Zwisler "both data=journal and dax"); 4338923ae0ffSRoss Zwisler goto failed_mount; 4339923ae0ffSRoss Zwisler } 434073b92a2aSSergey Karamov if (ext4_has_feature_encrypt(sb)) { 434173b92a2aSSergey Karamov ext4_msg(sb, KERN_WARNING, 434273b92a2aSSergey Karamov "encrypted files will use data=ordered " 434373b92a2aSSergey Karamov "instead of data journaling mode"); 434473b92a2aSSergey Karamov } 434556889787STheodore Ts'o if (test_opt(sb, DELALLOC)) 434656889787STheodore Ts'o clear_opt(sb, DELALLOC); 4347001e4a87STejun Heo } else { 4348001e4a87STejun Heo sb->s_iflags |= SB_I_CGROUPWB; 434956889787STheodore Ts'o } 435056889787STheodore Ts'o 43511751e8a6SLinus Torvalds sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 43521751e8a6SLinus Torvalds (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 4353ac27a0ecSDave Kleikamp 4354617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 4355e2b911c5SDarrick J. Wong (ext4_has_compat_features(sb) || 4356e2b911c5SDarrick J. Wong ext4_has_ro_compat_features(sb) || 4357e2b911c5SDarrick J. Wong ext4_has_incompat_features(sb))) 4358b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 4359b31e1552SEric Sandeen "feature flags set on rev 0 fs, " 4360b31e1552SEric Sandeen "running e2fsck is recommended"); 4361469108ffSTheodore Tso 4362ed3654ebSTheodore Ts'o if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 4363ed3654ebSTheodore Ts'o set_opt2(sb, HURD_COMPAT); 4364e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb)) { 4365ed3654ebSTheodore Ts'o ext4_msg(sb, KERN_ERR, 4366ed3654ebSTheodore Ts'o "The Hurd can't support 64-bit file systems"); 4367ed3654ebSTheodore Ts'o goto failed_mount; 4368ed3654ebSTheodore Ts'o } 4369dec214d0STahsin Erdogan 4370dec214d0STahsin Erdogan /* 4371dec214d0STahsin Erdogan * ea_inode feature uses l_i_version field which is not 4372dec214d0STahsin Erdogan * available in HURD_COMPAT mode. 4373dec214d0STahsin Erdogan */ 4374dec214d0STahsin Erdogan if (ext4_has_feature_ea_inode(sb)) { 4375dec214d0STahsin Erdogan ext4_msg(sb, KERN_ERR, 4376dec214d0STahsin Erdogan "ea_inode feature is not supported for Hurd"); 4377dec214d0STahsin Erdogan goto failed_mount; 4378dec214d0STahsin Erdogan } 4379ed3654ebSTheodore Ts'o } 4380ed3654ebSTheodore Ts'o 43812035e776STheodore Ts'o if (IS_EXT2_SB(sb)) { 43822035e776STheodore Ts'o if (ext2_feature_set_ok(sb)) 43832035e776STheodore Ts'o ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 43842035e776STheodore Ts'o "using the ext4 subsystem"); 43852035e776STheodore Ts'o else { 43860d9366d6SEric Sandeen /* 43870d9366d6SEric Sandeen * If we're probing be silent, if this looks like 43880d9366d6SEric Sandeen * it's actually an ext[34] filesystem. 43890d9366d6SEric Sandeen */ 43900d9366d6SEric Sandeen if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 43910d9366d6SEric Sandeen goto failed_mount; 43922035e776STheodore Ts'o ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 43932035e776STheodore Ts'o "to feature incompatibilities"); 43942035e776STheodore Ts'o goto failed_mount; 43952035e776STheodore Ts'o } 43962035e776STheodore Ts'o } 43972035e776STheodore Ts'o 43982035e776STheodore Ts'o if (IS_EXT3_SB(sb)) { 43992035e776STheodore Ts'o if (ext3_feature_set_ok(sb)) 44002035e776STheodore Ts'o ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 44012035e776STheodore Ts'o "using the ext4 subsystem"); 44022035e776STheodore Ts'o else { 44030d9366d6SEric Sandeen /* 44040d9366d6SEric Sandeen * If we're probing be silent, if this looks like 44050d9366d6SEric Sandeen * it's actually an ext4 filesystem. 44060d9366d6SEric Sandeen */ 44070d9366d6SEric Sandeen if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 44080d9366d6SEric Sandeen goto failed_mount; 44092035e776STheodore Ts'o ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 44102035e776STheodore Ts'o "to feature incompatibilities"); 44112035e776STheodore Ts'o goto failed_mount; 44122035e776STheodore Ts'o } 44132035e776STheodore Ts'o } 44142035e776STheodore Ts'o 4415469108ffSTheodore Tso /* 4416ac27a0ecSDave Kleikamp * Check feature flags regardless of the revision level, since we 4417ac27a0ecSDave Kleikamp * previously didn't change the revision level when setting the flags, 4418ac27a0ecSDave Kleikamp * so there is a chance incompat flags are set on a rev 0 filesystem. 4419ac27a0ecSDave Kleikamp */ 4420bc98a42cSDavid Howells if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 4421ac27a0ecSDave Kleikamp goto failed_mount; 4422a13fb1a4SEric Sandeen 44235b9554dcSTheodore Ts'o if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { 44245b9554dcSTheodore Ts'o ext4_msg(sb, KERN_ERR, 44255b9554dcSTheodore Ts'o "Number of reserved GDT blocks insanely large: %d", 44265b9554dcSTheodore Ts'o le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 44275b9554dcSTheodore Ts'o goto failed_mount; 44285b9554dcSTheodore Ts'o } 44295b9554dcSTheodore Ts'o 4430a8ab6d38SIra Weiny if (bdev_dax_supported(sb->s_bdev, blocksize)) 4431a8ab6d38SIra Weiny set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags); 4432a8ab6d38SIra Weiny 4433fc626fe3SIra Weiny if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) { 4434559db4c6SRoss Zwisler if (ext4_has_feature_inline_data(sb)) { 4435559db4c6SRoss Zwisler ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 4436559db4c6SRoss Zwisler " that may contain inline data"); 4437361d24d4SEric Sandeen goto failed_mount; 4438559db4c6SRoss Zwisler } 4439a8ab6d38SIra Weiny if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) { 444024f3478dSDan Williams ext4_msg(sb, KERN_ERR, 4441361d24d4SEric Sandeen "DAX unsupported by block device."); 4442361d24d4SEric Sandeen goto failed_mount; 444324f3478dSDan Williams } 4444923ae0ffSRoss Zwisler } 4445923ae0ffSRoss Zwisler 4446e2b911c5SDarrick J. Wong if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 44476ddb2447STheodore Ts'o ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 44486ddb2447STheodore Ts'o es->s_encryption_level); 44496ddb2447STheodore Ts'o goto failed_mount; 44506ddb2447STheodore Ts'o } 44516ddb2447STheodore Ts'o 4452ac27a0ecSDave Kleikamp if (sb->s_blocksize != blocksize) { 4453ce40733cSAneesh Kumar K.V /* Validate the filesystem blocksize */ 4454ce40733cSAneesh Kumar K.V if (!sb_set_blocksize(sb, blocksize)) { 4455b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "bad block size %d", 4456ce40733cSAneesh Kumar K.V blocksize); 4457ac27a0ecSDave Kleikamp goto failed_mount; 4458ac27a0ecSDave Kleikamp } 4459ac27a0ecSDave Kleikamp 4460ac27a0ecSDave Kleikamp brelse(bh); 446170bbb3e0SAndrew Morton logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 446270bbb3e0SAndrew Morton offset = do_div(logical_sb_block, blocksize); 44638394a6abSzhangyi (F) bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 44648394a6abSzhangyi (F) if (IS_ERR(bh)) { 4465b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4466b31e1552SEric Sandeen "Can't read superblock on 2nd try"); 44678394a6abSzhangyi (F) ret = PTR_ERR(bh); 44688394a6abSzhangyi (F) bh = NULL; 4469ac27a0ecSDave Kleikamp goto failed_mount; 4470ac27a0ecSDave Kleikamp } 44712716b802STheodore Ts'o es = (struct ext4_super_block *)(bh->b_data + offset); 4472ac27a0ecSDave Kleikamp sbi->s_es = es; 4473617ba13bSMingming Cao if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 4474b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4475b31e1552SEric Sandeen "Magic mismatch, very weird!"); 4476ac27a0ecSDave Kleikamp goto failed_mount; 4477ac27a0ecSDave Kleikamp } 4478ac27a0ecSDave Kleikamp } 4479ac27a0ecSDave Kleikamp 4480e2b911c5SDarrick J. Wong has_huge_files = ext4_has_feature_huge_file(sb); 4481f287a1a5STheodore Ts'o sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 4482f287a1a5STheodore Ts'o has_huge_files); 4483f287a1a5STheodore Ts'o sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 4484ac27a0ecSDave Kleikamp 44850d1ee42fSAlexandre Ratchov sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 4486e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb)) { 44878fadc143SAlexandre Ratchov if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 44880d1ee42fSAlexandre Ratchov sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 4489d8ea6cf8Svignesh babu !is_power_of_2(sbi->s_desc_size)) { 4490b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4491b31e1552SEric Sandeen "unsupported descriptor size %lu", 44920d1ee42fSAlexandre Ratchov sbi->s_desc_size); 44930d1ee42fSAlexandre Ratchov goto failed_mount; 44940d1ee42fSAlexandre Ratchov } 44950d1ee42fSAlexandre Ratchov } else 44960d1ee42fSAlexandre Ratchov sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 44970b8e58a1SAndreas Dilger 4498ac27a0ecSDave Kleikamp sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 4499ac27a0ecSDave Kleikamp sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 45000b8e58a1SAndreas Dilger 4501617ba13bSMingming Cao sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 4502ac27a0ecSDave Kleikamp if (sbi->s_inodes_per_block == 0) 4503617ba13bSMingming Cao goto cantfind_ext4; 4504cd6bb35bSTheodore Ts'o if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 4505cd6bb35bSTheodore Ts'o sbi->s_inodes_per_group > blocksize * 8) { 4506cd6bb35bSTheodore Ts'o ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 4507b9c538daSJosh Triplett sbi->s_inodes_per_group); 4508cd6bb35bSTheodore Ts'o goto failed_mount; 4509cd6bb35bSTheodore Ts'o } 4510ac27a0ecSDave Kleikamp sbi->s_itb_per_group = sbi->s_inodes_per_group / 4511ac27a0ecSDave Kleikamp sbi->s_inodes_per_block; 45120d1ee42fSAlexandre Ratchov sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 4513ac27a0ecSDave Kleikamp sbi->s_sbh = bh; 4514ac27a0ecSDave Kleikamp sbi->s_mount_state = le16_to_cpu(es->s_state); 4515e57aa839SFengguang Wu sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 4516e57aa839SFengguang Wu sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 45170b8e58a1SAndreas Dilger 4518ac27a0ecSDave Kleikamp for (i = 0; i < 4; i++) 4519ac27a0ecSDave Kleikamp sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 4520ac27a0ecSDave Kleikamp sbi->s_def_hash_version = es->s_def_hash_version; 4521e2b911c5SDarrick J. Wong if (ext4_has_feature_dir_index(sb)) { 4522f99b2589STheodore Ts'o i = le32_to_cpu(es->s_flags); 4523f99b2589STheodore Ts'o if (i & EXT2_FLAGS_UNSIGNED_HASH) 4524f99b2589STheodore Ts'o sbi->s_hash_unsigned = 3; 4525f99b2589STheodore Ts'o else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 4526f99b2589STheodore Ts'o #ifdef __CHAR_UNSIGNED__ 4527bc98a42cSDavid Howells if (!sb_rdonly(sb)) 452823301410STheodore Ts'o es->s_flags |= 452923301410STheodore Ts'o cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 4530f99b2589STheodore Ts'o sbi->s_hash_unsigned = 3; 4531f99b2589STheodore Ts'o #else 4532bc98a42cSDavid Howells if (!sb_rdonly(sb)) 453323301410STheodore Ts'o es->s_flags |= 453423301410STheodore Ts'o cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 4535f99b2589STheodore Ts'o #endif 4536f99b2589STheodore Ts'o } 453723301410STheodore Ts'o } 4538ac27a0ecSDave Kleikamp 4539281b5995STheodore Ts'o /* Handle clustersize */ 4540281b5995STheodore Ts'o clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 4541ef5fd681SKaixu Xia if (ext4_has_feature_bigalloc(sb)) { 4542281b5995STheodore Ts'o if (clustersize < blocksize) { 4543281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, 4544281b5995STheodore Ts'o "cluster size (%d) smaller than " 4545281b5995STheodore Ts'o "block size (%d)", clustersize, blocksize); 4546281b5995STheodore Ts'o goto failed_mount; 4547281b5995STheodore Ts'o } 4548281b5995STheodore Ts'o sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 4549281b5995STheodore Ts'o le32_to_cpu(es->s_log_block_size); 4550281b5995STheodore Ts'o sbi->s_clusters_per_group = 4551281b5995STheodore Ts'o le32_to_cpu(es->s_clusters_per_group); 4552281b5995STheodore Ts'o if (sbi->s_clusters_per_group > blocksize * 8) { 4553281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, 4554281b5995STheodore Ts'o "#clusters per group too big: %lu", 4555281b5995STheodore Ts'o sbi->s_clusters_per_group); 4556281b5995STheodore Ts'o goto failed_mount; 4557281b5995STheodore Ts'o } 4558281b5995STheodore Ts'o if (sbi->s_blocks_per_group != 4559281b5995STheodore Ts'o (sbi->s_clusters_per_group * (clustersize / blocksize))) { 4560281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 4561281b5995STheodore Ts'o "clusters per group (%lu) inconsistent", 4562281b5995STheodore Ts'o sbi->s_blocks_per_group, 4563281b5995STheodore Ts'o sbi->s_clusters_per_group); 4564281b5995STheodore Ts'o goto failed_mount; 4565281b5995STheodore Ts'o } 4566281b5995STheodore Ts'o } else { 4567281b5995STheodore Ts'o if (clustersize != blocksize) { 4568bfe0a5f4STheodore Ts'o ext4_msg(sb, KERN_ERR, 4569bfe0a5f4STheodore Ts'o "fragment/cluster size (%d) != " 4570bfe0a5f4STheodore Ts'o "block size (%d)", clustersize, blocksize); 4571bfe0a5f4STheodore Ts'o goto failed_mount; 4572281b5995STheodore Ts'o } 4573ac27a0ecSDave Kleikamp if (sbi->s_blocks_per_group > blocksize * 8) { 4574b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4575b31e1552SEric Sandeen "#blocks per group too big: %lu", 4576ac27a0ecSDave Kleikamp sbi->s_blocks_per_group); 4577ac27a0ecSDave Kleikamp goto failed_mount; 4578ac27a0ecSDave Kleikamp } 4579281b5995STheodore Ts'o sbi->s_clusters_per_group = sbi->s_blocks_per_group; 4580281b5995STheodore Ts'o sbi->s_cluster_bits = 0; 4581281b5995STheodore Ts'o } 4582281b5995STheodore Ts'o sbi->s_cluster_ratio = clustersize / blocksize; 4583281b5995STheodore Ts'o 4584960fd856STheodore Ts'o /* Do we have standard group size of clustersize * 8 blocks ? */ 4585960fd856STheodore Ts'o if (sbi->s_blocks_per_group == clustersize << 3) 4586960fd856STheodore Ts'o set_opt2(sb, STD_GROUP_SIZE); 4587960fd856STheodore Ts'o 4588bf43d84bSEric Sandeen /* 4589bf43d84bSEric Sandeen * Test whether we have more sectors than will fit in sector_t, 4590bf43d84bSEric Sandeen * and whether the max offset is addressable by the page cache. 4591bf43d84bSEric Sandeen */ 45925a9ae68aSDarrick J. Wong err = generic_check_addressable(sb->s_blocksize_bits, 459330ca22c7SPatrick J. LoPresti ext4_blocks_count(es)); 45945a9ae68aSDarrick J. Wong if (err) { 4595b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "filesystem" 4596bf43d84bSEric Sandeen " too large to mount safely on this system"); 4597ac27a0ecSDave Kleikamp goto failed_mount; 4598ac27a0ecSDave Kleikamp } 4599ac27a0ecSDave Kleikamp 4600617ba13bSMingming Cao if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 4601617ba13bSMingming Cao goto cantfind_ext4; 4602e7c95593SEric Sandeen 46030f2ddca6SFrom: Thiemo Nagel /* check blocks count against device size */ 46040f2ddca6SFrom: Thiemo Nagel blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 46050f2ddca6SFrom: Thiemo Nagel if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4606b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4607b31e1552SEric Sandeen "exceeds size of device (%llu blocks)", 46080f2ddca6SFrom: Thiemo Nagel ext4_blocks_count(es), blocks_count); 46090f2ddca6SFrom: Thiemo Nagel goto failed_mount; 46100f2ddca6SFrom: Thiemo Nagel } 46110f2ddca6SFrom: Thiemo Nagel 46124ec11028STheodore Ts'o /* 46134ec11028STheodore Ts'o * It makes no sense for the first data block to be beyond the end 46144ec11028STheodore Ts'o * of the filesystem. 46154ec11028STheodore Ts'o */ 46164ec11028STheodore Ts'o if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4617b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4618b31e1552SEric Sandeen "block %u is beyond end of filesystem (%llu)", 4619e7c95593SEric Sandeen le32_to_cpu(es->s_first_data_block), 46204ec11028STheodore Ts'o ext4_blocks_count(es)); 4621e7c95593SEric Sandeen goto failed_mount; 4622e7c95593SEric Sandeen } 4623bfe0a5f4STheodore Ts'o if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4624bfe0a5f4STheodore Ts'o (sbi->s_cluster_ratio == 1)) { 4625bfe0a5f4STheodore Ts'o ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4626bfe0a5f4STheodore Ts'o "block is 0 with a 1k block and cluster size"); 4627bfe0a5f4STheodore Ts'o goto failed_mount; 4628bfe0a5f4STheodore Ts'o } 4629bfe0a5f4STheodore Ts'o 4630bd81d8eeSLaurent Vivier blocks_count = (ext4_blocks_count(es) - 4631bd81d8eeSLaurent Vivier le32_to_cpu(es->s_first_data_block) + 4632bd81d8eeSLaurent Vivier EXT4_BLOCKS_PER_GROUP(sb) - 1); 4633bd81d8eeSLaurent Vivier do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 46344ec11028STheodore Ts'o if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4635df41460aSJosh Triplett ext4_msg(sb, KERN_WARNING, "groups count too large: %llu " 46364ec11028STheodore Ts'o "(block count %llu, first data block %u, " 4637df41460aSJosh Triplett "blocks per group %lu)", blocks_count, 46384ec11028STheodore Ts'o ext4_blocks_count(es), 46394ec11028STheodore Ts'o le32_to_cpu(es->s_first_data_block), 46404ec11028STheodore Ts'o EXT4_BLOCKS_PER_GROUP(sb)); 46414ec11028STheodore Ts'o goto failed_mount; 46424ec11028STheodore Ts'o } 4643bd81d8eeSLaurent Vivier sbi->s_groups_count = blocks_count; 4644fb0a387dSEric Sandeen sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 4645fb0a387dSEric Sandeen (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 46469e463084STheodore Ts'o if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 46479e463084STheodore Ts'o le32_to_cpu(es->s_inodes_count)) { 46489e463084STheodore Ts'o ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 46499e463084STheodore Ts'o le32_to_cpu(es->s_inodes_count), 46509e463084STheodore Ts'o ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 46519e463084STheodore Ts'o ret = -EINVAL; 46529e463084STheodore Ts'o goto failed_mount; 46539e463084STheodore Ts'o } 4654617ba13bSMingming Cao db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4655617ba13bSMingming Cao EXT4_DESC_PER_BLOCK(sb); 46563a4b77cdSEryu Guan if (ext4_has_feature_meta_bg(sb)) { 46572ba3e6e8STheodore Ts'o if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 46583a4b77cdSEryu Guan ext4_msg(sb, KERN_WARNING, 46593a4b77cdSEryu Guan "first meta block group too large: %u " 46603a4b77cdSEryu Guan "(group descriptor block count %u)", 46613a4b77cdSEryu Guan le32_to_cpu(es->s_first_meta_bg), db_count); 46623a4b77cdSEryu Guan goto failed_mount; 46633a4b77cdSEryu Guan } 46643a4b77cdSEryu Guan } 46651d0c3924STheodore Ts'o rcu_assign_pointer(sbi->s_group_desc, 46661d0c3924STheodore Ts'o kvmalloc_array(db_count, 4667f18a5f21STheodore Ts'o sizeof(struct buffer_head *), 46681d0c3924STheodore Ts'o GFP_KERNEL)); 4669ac27a0ecSDave Kleikamp if (sbi->s_group_desc == NULL) { 4670b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "not enough memory"); 46712cde417dSTheodore Ts'o ret = -ENOMEM; 4672ac27a0ecSDave Kleikamp goto failed_mount; 4673ac27a0ecSDave Kleikamp } 4674ac27a0ecSDave Kleikamp 4675705895b6SPekka Enberg bgl_lock_init(sbi->s_blockgroup_lock); 4676ac27a0ecSDave Kleikamp 467785c8f176SAndrew Perepechko /* Pre-read the descriptors into the buffer cache */ 467885c8f176SAndrew Perepechko for (i = 0; i < db_count; i++) { 467985c8f176SAndrew Perepechko block = descriptor_loc(sb, logical_sb_block, i); 46805df1d412Szhangyi (F) ext4_sb_breadahead_unmovable(sb, block); 468185c8f176SAndrew Perepechko } 468285c8f176SAndrew Perepechko 4683ac27a0ecSDave Kleikamp for (i = 0; i < db_count; i++) { 46841d0c3924STheodore Ts'o struct buffer_head *bh; 46851d0c3924STheodore Ts'o 468670bbb3e0SAndrew Morton block = descriptor_loc(sb, logical_sb_block, i); 46878394a6abSzhangyi (F) bh = ext4_sb_bread_unmovable(sb, block); 46888394a6abSzhangyi (F) if (IS_ERR(bh)) { 4689b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4690b31e1552SEric Sandeen "can't read group descriptor %d", i); 4691ac27a0ecSDave Kleikamp db_count = i; 46928394a6abSzhangyi (F) ret = PTR_ERR(bh); 4693ac27a0ecSDave Kleikamp goto failed_mount2; 4694ac27a0ecSDave Kleikamp } 46951d0c3924STheodore Ts'o rcu_read_lock(); 46961d0c3924STheodore Ts'o rcu_dereference(sbi->s_group_desc)[i] = bh; 46971d0c3924STheodore Ts'o rcu_read_unlock(); 4698ac27a0ecSDave Kleikamp } 469944de022cSTheodore Ts'o sbi->s_gdb_count = db_count; 4700829fa70dSTheodore Ts'o if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { 4701b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 47026a797d27SDarrick J. Wong ret = -EFSCORRUPTED; 4703f9ae9cf5STheodore Ts'o goto failed_mount2; 4704ac27a0ecSDave Kleikamp } 4705772cb7c8SJose R. Santos 4706235699a8SKees Cook timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 4707c92dc856SJan Kara spin_lock_init(&sbi->s_error_lock); 4708c92dc856SJan Kara INIT_WORK(&sbi->s_error_work, flush_stashed_error_work); 470904496411STao Ma 4710a75ae78fSDmitry Monakhov /* Register extent status tree shrinker */ 4711eb68d0e2SZheng Liu if (ext4_es_register_shrinker(sbi)) 4712ce7e010aSTheodore Ts'o goto failed_mount3; 4713ce7e010aSTheodore Ts'o 4714c9de560dSAlex Tomas sbi->s_stripe = ext4_get_stripe_size(sbi); 471567a5da56SZheng Liu sbi->s_extent_max_zeroout_kb = 32; 4716c9de560dSAlex Tomas 4717f9ae9cf5STheodore Ts'o /* 4718f9ae9cf5STheodore Ts'o * set up enough so that it can read an inode 4719f9ae9cf5STheodore Ts'o */ 4720f9ae9cf5STheodore Ts'o sb->s_op = &ext4_sops; 4721617ba13bSMingming Cao sb->s_export_op = &ext4_export_ops; 4722617ba13bSMingming Cao sb->s_xattr = ext4_xattr_handlers; 4723643fa961SChandan Rajendra #ifdef CONFIG_FS_ENCRYPTION 4724a7550b30SJaegeuk Kim sb->s_cop = &ext4_cryptops; 4725ffcc4182SEric Biggers #endif 4726c93d8f88SEric Biggers #ifdef CONFIG_FS_VERITY 4727c93d8f88SEric Biggers sb->s_vop = &ext4_verityops; 4728c93d8f88SEric Biggers #endif 4729ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4730617ba13bSMingming Cao sb->dq_op = &ext4_quota_operations; 4731e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) 47321fa5efe3SJan Kara sb->s_qcop = &dquot_quotactl_sysfile_ops; 4733262b4662SJan Kara else 4734262b4662SJan Kara sb->s_qcop = &ext4_qctl_operations; 4735689c958cSLi Xi sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4736ac27a0ecSDave Kleikamp #endif 473785787090SChristoph Hellwig memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 4738f2fa2ffcSAneesh Kumar K.V 4739ac27a0ecSDave Kleikamp INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 47403b9d4ed2STheodore Ts'o mutex_init(&sbi->s_orphan_lock); 4741ac27a0ecSDave Kleikamp 4742aa75f4d3SHarshad Shirwadkar /* Initialize fast commit stuff */ 4743aa75f4d3SHarshad Shirwadkar atomic_set(&sbi->s_fc_subtid, 0); 4744aa75f4d3SHarshad Shirwadkar atomic_set(&sbi->s_fc_ineligible_updates, 0); 4745aa75f4d3SHarshad Shirwadkar INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]); 4746aa75f4d3SHarshad Shirwadkar INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]); 4747aa75f4d3SHarshad Shirwadkar INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]); 4748aa75f4d3SHarshad Shirwadkar INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]); 4749aa75f4d3SHarshad Shirwadkar sbi->s_fc_bytes = 0; 47509b5f6c9bSHarshad Shirwadkar ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE); 47519b5f6c9bSHarshad Shirwadkar ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING); 4752aa75f4d3SHarshad Shirwadkar spin_lock_init(&sbi->s_fc_lock); 4753aa75f4d3SHarshad Shirwadkar memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats)); 47548016e29fSHarshad Shirwadkar sbi->s_fc_replay_state.fc_regions = NULL; 47558016e29fSHarshad Shirwadkar sbi->s_fc_replay_state.fc_regions_size = 0; 47568016e29fSHarshad Shirwadkar sbi->s_fc_replay_state.fc_regions_used = 0; 47578016e29fSHarshad Shirwadkar sbi->s_fc_replay_state.fc_regions_valid = 0; 47588016e29fSHarshad Shirwadkar sbi->s_fc_replay_state.fc_modified_inodes = NULL; 47598016e29fSHarshad Shirwadkar sbi->s_fc_replay_state.fc_modified_inodes_size = 0; 47608016e29fSHarshad Shirwadkar sbi->s_fc_replay_state.fc_modified_inodes_used = 0; 4761aa75f4d3SHarshad Shirwadkar 4762ac27a0ecSDave Kleikamp sb->s_root = NULL; 4763ac27a0ecSDave Kleikamp 4764ac27a0ecSDave Kleikamp needs_recovery = (es->s_last_orphan != 0 || 4765e2b911c5SDarrick J. Wong ext4_has_feature_journal_needs_recovery(sb)); 4766ac27a0ecSDave Kleikamp 4767bc98a42cSDavid Howells if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) 4768c5e06d10SJohann Lombardi if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 476950460fe8SDarrick J. Wong goto failed_mount3a; 4770c5e06d10SJohann Lombardi 4771ac27a0ecSDave Kleikamp /* 4772ac27a0ecSDave Kleikamp * The first inode we look at is the journal inode. Don't try 4773ac27a0ecSDave Kleikamp * root first: it may be modified in the journal! 4774ac27a0ecSDave Kleikamp */ 4775e2b911c5SDarrick J. Wong if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 47764753d8a2STheodore Ts'o err = ext4_load_journal(sb, es, journal_devnum); 47774753d8a2STheodore Ts'o if (err) 477850460fe8SDarrick J. Wong goto failed_mount3a; 4779bc98a42cSDavid Howells } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 4780e2b911c5SDarrick J. Wong ext4_has_feature_journal_needs_recovery(sb)) { 4781b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "required journal recovery " 4782b31e1552SEric Sandeen "suppressed and not mounted read-only"); 4783744692dcSJiaying Zhang goto failed_mount_wq; 4784ac27a0ecSDave Kleikamp } else { 47851e381f60SDmitry Monakhov /* Nojournal mode, all journal mount options are illegal */ 47861e381f60SDmitry Monakhov if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 47871e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 47881e381f60SDmitry Monakhov "journal_checksum, fs mounted w/o journal"); 47891e381f60SDmitry Monakhov goto failed_mount_wq; 47901e381f60SDmitry Monakhov } 47911e381f60SDmitry Monakhov if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 47921e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 47931e381f60SDmitry Monakhov "journal_async_commit, fs mounted w/o journal"); 47941e381f60SDmitry Monakhov goto failed_mount_wq; 47951e381f60SDmitry Monakhov } 47961e381f60SDmitry Monakhov if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 47971e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 47981e381f60SDmitry Monakhov "commit=%lu, fs mounted w/o journal", 47991e381f60SDmitry Monakhov sbi->s_commit_interval / HZ); 48001e381f60SDmitry Monakhov goto failed_mount_wq; 48011e381f60SDmitry Monakhov } 48021e381f60SDmitry Monakhov if (EXT4_MOUNT_DATA_FLAGS & 48031e381f60SDmitry Monakhov (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 48041e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 48051e381f60SDmitry Monakhov "data=, fs mounted w/o journal"); 48061e381f60SDmitry Monakhov goto failed_mount_wq; 48071e381f60SDmitry Monakhov } 480850b29d8fSDebabrata Banerjee sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; 48091e381f60SDmitry Monakhov clear_opt(sb, JOURNAL_CHECKSUM); 4810fd8c37ecSTheodore Ts'o clear_opt(sb, DATA_FLAGS); 4811995a3ed6SHarshad Shirwadkar clear_opt2(sb, JOURNAL_FAST_COMMIT); 48120390131bSFrank Mayhar sbi->s_journal = NULL; 48130390131bSFrank Mayhar needs_recovery = 0; 48140390131bSFrank Mayhar goto no_journal; 4815ac27a0ecSDave Kleikamp } 4816ac27a0ecSDave Kleikamp 4817e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb) && 4818eb40a09cSJose R. Santos !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4819eb40a09cSJose R. Santos JBD2_FEATURE_INCOMPAT_64BIT)) { 4820b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 4821744692dcSJiaying Zhang goto failed_mount_wq; 4822eb40a09cSJose R. Santos } 4823eb40a09cSJose R. Santos 482425ed6e8aSDarrick J. Wong if (!set_journal_csum_feature_set(sb)) { 482525ed6e8aSDarrick J. Wong ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 482625ed6e8aSDarrick J. Wong "feature set"); 482725ed6e8aSDarrick J. Wong goto failed_mount_wq; 4828d4da6c9cSLinus Torvalds } 4829818d276cSGirish Shilamkar 4830a1e5e465SHarshad Shirwadkar if (test_opt2(sb, JOURNAL_FAST_COMMIT) && 4831a1e5e465SHarshad Shirwadkar !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4832a1e5e465SHarshad Shirwadkar JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) { 4833a1e5e465SHarshad Shirwadkar ext4_msg(sb, KERN_ERR, 4834a1e5e465SHarshad Shirwadkar "Failed to set fast commit journal feature"); 4835a1e5e465SHarshad Shirwadkar goto failed_mount_wq; 4836a1e5e465SHarshad Shirwadkar } 4837a1e5e465SHarshad Shirwadkar 4838ac27a0ecSDave Kleikamp /* We have now updated the journal if required, so we can 4839ac27a0ecSDave Kleikamp * validate the data journaling mode. */ 4840ac27a0ecSDave Kleikamp switch (test_opt(sb, DATA_FLAGS)) { 4841ac27a0ecSDave Kleikamp case 0: 4842ac27a0ecSDave Kleikamp /* No mode set, assume a default based on the journal 484363f57933SAndrew Morton * capabilities: ORDERED_DATA if the journal can 484463f57933SAndrew Morton * cope, else JOURNAL_DATA 484563f57933SAndrew Morton */ 4846dab291afSMingming Cao if (jbd2_journal_check_available_features 484727f394a7STyson Nottingham (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4848fd8c37ecSTheodore Ts'o set_opt(sb, ORDERED_DATA); 484927f394a7STyson Nottingham sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 485027f394a7STyson Nottingham } else { 4851fd8c37ecSTheodore Ts'o set_opt(sb, JOURNAL_DATA); 485227f394a7STyson Nottingham sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 485327f394a7STyson Nottingham } 4854ac27a0ecSDave Kleikamp break; 4855ac27a0ecSDave Kleikamp 4856617ba13bSMingming Cao case EXT4_MOUNT_ORDERED_DATA: 4857617ba13bSMingming Cao case EXT4_MOUNT_WRITEBACK_DATA: 4858dab291afSMingming Cao if (!jbd2_journal_check_available_features 4859dab291afSMingming Cao (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4860b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Journal does not support " 4861b31e1552SEric Sandeen "requested data journaling mode"); 4862744692dcSJiaying Zhang goto failed_mount_wq; 4863ac27a0ecSDave Kleikamp } 48645a150bdeSGustavo A. R. Silva break; 4865ac27a0ecSDave Kleikamp default: 4866ac27a0ecSDave Kleikamp break; 4867ac27a0ecSDave Kleikamp } 4868ab04df78SJan Kara 4869ab04df78SJan Kara if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 4870ab04df78SJan Kara test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4871ab04df78SJan Kara ext4_msg(sb, KERN_ERR, "can't mount with " 4872ab04df78SJan Kara "journal_async_commit in data=ordered mode"); 4873ab04df78SJan Kara goto failed_mount_wq; 4874ab04df78SJan Kara } 4875ab04df78SJan Kara 4876b3881f74STheodore Ts'o set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4877ac27a0ecSDave Kleikamp 4878342af94eSMauricio Faria de Oliveira sbi->s_journal->j_submit_inode_data_buffers = 4879afb585a9SMauricio Faria de Oliveira ext4_journal_submit_inode_data_buffers; 4880342af94eSMauricio Faria de Oliveira sbi->s_journal->j_finish_inode_data_buffers = 4881afb585a9SMauricio Faria de Oliveira ext4_journal_finish_inode_data_buffers; 488218aadd47SBobi Jam 4883ce7e010aSTheodore Ts'o no_journal: 4884cdb7ee4cSTahsin Erdogan if (!test_opt(sb, NO_MBCACHE)) { 488547387409STahsin Erdogan sbi->s_ea_block_cache = ext4_xattr_create_cache(); 488647387409STahsin Erdogan if (!sbi->s_ea_block_cache) { 4887cdb7ee4cSTahsin Erdogan ext4_msg(sb, KERN_ERR, 4888cdb7ee4cSTahsin Erdogan "Failed to create ea_block_cache"); 48899c191f70ST Makphaibulchoke goto failed_mount_wq; 48909c191f70ST Makphaibulchoke } 48919c191f70ST Makphaibulchoke 4892dec214d0STahsin Erdogan if (ext4_has_feature_ea_inode(sb)) { 4893dec214d0STahsin Erdogan sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 4894dec214d0STahsin Erdogan if (!sbi->s_ea_inode_cache) { 4895dec214d0STahsin Erdogan ext4_msg(sb, KERN_ERR, 4896dec214d0STahsin Erdogan "Failed to create ea_inode_cache"); 4897dec214d0STahsin Erdogan goto failed_mount_wq; 4898dec214d0STahsin Erdogan } 4899dec214d0STahsin Erdogan } 4900cdb7ee4cSTahsin Erdogan } 4901dec214d0STahsin Erdogan 4902c93d8f88SEric Biggers if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) { 4903c93d8f88SEric Biggers ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity"); 4904c93d8f88SEric Biggers goto failed_mount_wq; 4905c93d8f88SEric Biggers } 4906c93d8f88SEric Biggers 4907bc98a42cSDavid Howells if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) && 4908e2b911c5SDarrick J. Wong !ext4_has_feature_encrypt(sb)) { 4909e2b911c5SDarrick J. Wong ext4_set_feature_encrypt(sb); 49104392fbc4SJan Kara ext4_commit_super(sb); 49116ddb2447STheodore Ts'o } 49126ddb2447STheodore Ts'o 4913fd89d5f2STejun Heo /* 4914952fc18eSTheodore Ts'o * Get the # of file system overhead blocks from the 4915952fc18eSTheodore Ts'o * superblock if present. 4916952fc18eSTheodore Ts'o */ 4917952fc18eSTheodore Ts'o if (es->s_overhead_clusters) 4918952fc18eSTheodore Ts'o sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 4919952fc18eSTheodore Ts'o else { 492007aa2ea1SLukas Czerner err = ext4_calculate_overhead(sb); 492107aa2ea1SLukas Czerner if (err) 4922952fc18eSTheodore Ts'o goto failed_mount_wq; 4923952fc18eSTheodore Ts'o } 4924952fc18eSTheodore Ts'o 4925952fc18eSTheodore Ts'o /* 4926fd89d5f2STejun Heo * The maximum number of concurrent works can be high and 4927fd89d5f2STejun Heo * concurrency isn't really necessary. Limit it to 1. 4928fd89d5f2STejun Heo */ 49292e8fa54eSJan Kara EXT4_SB(sb)->rsv_conversion_wq = 49302e8fa54eSJan Kara alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 49312e8fa54eSJan Kara if (!EXT4_SB(sb)->rsv_conversion_wq) { 49322e8fa54eSJan Kara printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 493307aa2ea1SLukas Czerner ret = -ENOMEM; 49342e8fa54eSJan Kara goto failed_mount4; 49352e8fa54eSJan Kara } 49362e8fa54eSJan Kara 4937ac27a0ecSDave Kleikamp /* 4938dab291afSMingming Cao * The jbd2_journal_load will have done any necessary log recovery, 4939ac27a0ecSDave Kleikamp * so we can safely mount the rest of the filesystem now. 4940ac27a0ecSDave Kleikamp */ 4941ac27a0ecSDave Kleikamp 49428a363970STheodore Ts'o root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 49431d1fe1eeSDavid Howells if (IS_ERR(root)) { 4944b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "get root inode failed"); 49451d1fe1eeSDavid Howells ret = PTR_ERR(root); 494632a9bb57SManish Katiyar root = NULL; 4947ac27a0ecSDave Kleikamp goto failed_mount4; 4948ac27a0ecSDave Kleikamp } 4949ac27a0ecSDave Kleikamp if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 4950b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 495194bf608aSAl Viro iput(root); 4952ac27a0ecSDave Kleikamp goto failed_mount4; 4953ac27a0ecSDave Kleikamp } 4954b886ee3eSGabriel Krisman Bertazi 495548fde701SAl Viro sb->s_root = d_make_root(root); 49561d1fe1eeSDavid Howells if (!sb->s_root) { 4957b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "get root dentry failed"); 49581d1fe1eeSDavid Howells ret = -ENOMEM; 49591d1fe1eeSDavid Howells goto failed_mount4; 49601d1fe1eeSDavid Howells } 4961ac27a0ecSDave Kleikamp 4962c89128a0SJaegeuk Kim ret = ext4_setup_super(sb, es, sb_rdonly(sb)); 4963c89128a0SJaegeuk Kim if (ret == -EROFS) { 49641751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 4965c89128a0SJaegeuk Kim ret = 0; 4966c89128a0SJaegeuk Kim } else if (ret) 4967c89128a0SJaegeuk Kim goto failed_mount4a; 4968ef7f3835SKalpak Shah 4969b5799018STheodore Ts'o ext4_set_resv_clusters(sb); 497027dd4385SLukas Czerner 49710f5bde1dSJan Kara if (test_opt(sb, BLOCK_VALIDITY)) { 49726fd058f7STheodore Ts'o err = ext4_setup_system_zone(sb); 49736fd058f7STheodore Ts'o if (err) { 4974b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "failed to initialize system " 4975fbe845ddSCurt Wohlgemuth "zone (%d)", err); 4976f9ae9cf5STheodore Ts'o goto failed_mount4a; 4977f9ae9cf5STheodore Ts'o } 49780f5bde1dSJan Kara } 49798016e29fSHarshad Shirwadkar ext4_fc_replay_cleanup(sb); 4980f9ae9cf5STheodore Ts'o 4981f9ae9cf5STheodore Ts'o ext4_ext_init(sb); 4982f9ae9cf5STheodore Ts'o err = ext4_mb_init(sb); 4983f9ae9cf5STheodore Ts'o if (err) { 4984f9ae9cf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 4985f9ae9cf5STheodore Ts'o err); 4986dcf2d804STao Ma goto failed_mount5; 4987c2774d84SAneesh Kumar K.V } 4988c2774d84SAneesh Kumar K.V 4989027f14f5STheodore Ts'o /* 4990027f14f5STheodore Ts'o * We can only set up the journal commit callback once 4991027f14f5STheodore Ts'o * mballoc is initialized 4992027f14f5STheodore Ts'o */ 4993027f14f5STheodore Ts'o if (sbi->s_journal) 4994027f14f5STheodore Ts'o sbi->s_journal->j_commit_callback = 4995027f14f5STheodore Ts'o ext4_journal_commit_callback; 4996027f14f5STheodore Ts'o 4997d5e03cbbSTheodore Ts'o block = ext4_count_free_clusters(sb); 4998d5e03cbbSTheodore Ts'o ext4_free_blocks_count_set(sbi->s_es, 4999d5e03cbbSTheodore Ts'o EXT4_C2B(sbi, block)); 5000908c7f19STejun Heo err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 5001908c7f19STejun Heo GFP_KERNEL); 5002d5e03cbbSTheodore Ts'o if (!err) { 5003d5e03cbbSTheodore Ts'o unsigned long freei = ext4_count_free_inodes(sb); 5004d5e03cbbSTheodore Ts'o sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 5005908c7f19STejun Heo err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 5006908c7f19STejun Heo GFP_KERNEL); 5007d5e03cbbSTheodore Ts'o } 5008d5e03cbbSTheodore Ts'o if (!err) 5009d5e03cbbSTheodore Ts'o err = percpu_counter_init(&sbi->s_dirs_counter, 5010908c7f19STejun Heo ext4_count_dirs(sb), GFP_KERNEL); 5011d5e03cbbSTheodore Ts'o if (!err) 5012908c7f19STejun Heo err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 5013908c7f19STejun Heo GFP_KERNEL); 5014c8585c6fSDaeho Jeong if (!err) 5015bbd55937SEric Biggers err = percpu_init_rwsem(&sbi->s_writepages_rwsem); 5016c8585c6fSDaeho Jeong 5017d5e03cbbSTheodore Ts'o if (err) { 5018d5e03cbbSTheodore Ts'o ext4_msg(sb, KERN_ERR, "insufficient memory"); 5019d5e03cbbSTheodore Ts'o goto failed_mount6; 5020d5e03cbbSTheodore Ts'o } 5021d5e03cbbSTheodore Ts'o 5022e2b911c5SDarrick J. Wong if (ext4_has_feature_flex_bg(sb)) 5023d5e03cbbSTheodore Ts'o if (!ext4_fill_flex_info(sb)) { 5024d5e03cbbSTheodore Ts'o ext4_msg(sb, KERN_ERR, 5025d5e03cbbSTheodore Ts'o "unable to initialize " 5026d5e03cbbSTheodore Ts'o "flex_bg meta info!"); 5027d5e03cbbSTheodore Ts'o goto failed_mount6; 5028d5e03cbbSTheodore Ts'o } 5029d5e03cbbSTheodore Ts'o 5030bfff6873SLukas Czerner err = ext4_register_li_request(sb, first_not_zeroed); 5031bfff6873SLukas Czerner if (err) 5032dcf2d804STao Ma goto failed_mount6; 5033bfff6873SLukas Czerner 5034b5799018STheodore Ts'o err = ext4_register_sysfs(sb); 5035dcf2d804STao Ma if (err) 5036dcf2d804STao Ma goto failed_mount7; 50373197ebdbSTheodore Ts'o 50389b2ff357SJan Kara #ifdef CONFIG_QUOTA 50399b2ff357SJan Kara /* Enable quota usage during mount. */ 5040bc98a42cSDavid Howells if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 50419b2ff357SJan Kara err = ext4_enable_quotas(sb); 50429b2ff357SJan Kara if (err) 50439b2ff357SJan Kara goto failed_mount8; 50449b2ff357SJan Kara } 50459b2ff357SJan Kara #endif /* CONFIG_QUOTA */ 50469b2ff357SJan Kara 5047bc71726cSzhangyi (F) /* 5048bc71726cSzhangyi (F) * Save the original bdev mapping's wb_err value which could be 5049bc71726cSzhangyi (F) * used to detect the metadata async write error. 5050bc71726cSzhangyi (F) */ 5051bc71726cSzhangyi (F) spin_lock_init(&sbi->s_bdev_wb_lock); 5052bc71726cSzhangyi (F) errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err, 5053bc71726cSzhangyi (F) &sbi->s_bdev_wb_err); 5054bc71726cSzhangyi (F) sb->s_bdev->bd_super = sb; 5055617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 5056617ba13bSMingming Cao ext4_orphan_cleanup(sb, es); 5057617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 50580390131bSFrank Mayhar if (needs_recovery) { 5059b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "recovery complete"); 506011215630SJan Kara err = ext4_mark_recovery_complete(sb, es); 506111215630SJan Kara if (err) 506211215630SJan Kara goto failed_mount8; 50630390131bSFrank Mayhar } 50640390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal) { 50650390131bSFrank Mayhar if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 50660390131bSFrank Mayhar descr = " journalled data mode"; 50670390131bSFrank Mayhar else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 50680390131bSFrank Mayhar descr = " ordered data mode"; 50690390131bSFrank Mayhar else 50700390131bSFrank Mayhar descr = " writeback data mode"; 50710390131bSFrank Mayhar } else 50720390131bSFrank Mayhar descr = "out journal"; 50730390131bSFrank Mayhar 507479add3a3SLukas Czerner if (test_opt(sb, DISCARD)) { 507579add3a3SLukas Czerner struct request_queue *q = bdev_get_queue(sb->s_bdev); 507679add3a3SLukas Czerner if (!blk_queue_discard(q)) 507779add3a3SLukas Czerner ext4_msg(sb, KERN_WARNING, 507879add3a3SLukas Czerner "mounting with \"discard\" option, but " 507979add3a3SLukas Czerner "the device does not support discard"); 508079add3a3SLukas Czerner } 508179add3a3SLukas Czerner 5082e294a537STheodore Ts'o if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 5083d4c402d9SCurt Wohlgemuth ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 5084ca9b404fSRoman Anufriev "Opts: %.*s%s%s. Quota mode: %s.", descr, 50855aee0f8aSTheodore Ts'o (int) sizeof(sbi->s_es->s_mount_opts), 50865aee0f8aSTheodore Ts'o sbi->s_es->s_mount_opts, 5087ca9b404fSRoman Anufriev *sbi->s_es->s_mount_opts ? "; " : "", orig_data, 5088ca9b404fSRoman Anufriev ext4_quota_mode(sb)); 5089ac27a0ecSDave Kleikamp 509066e61a9eSTheodore Ts'o if (es->s_error_count) 509166e61a9eSTheodore Ts'o mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 5092ac27a0ecSDave Kleikamp 5093efbed4dcSTheodore Ts'o /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 5094efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 5095efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 5096efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 50971cf006edSDmitry Monakhov atomic_set(&sbi->s_warning_count, 0); 50981cf006edSDmitry Monakhov atomic_set(&sbi->s_msg_count, 0); 5099efbed4dcSTheodore Ts'o 5100d4c402d9SCurt Wohlgemuth kfree(orig_data); 5101ac27a0ecSDave Kleikamp return 0; 5102ac27a0ecSDave Kleikamp 5103617ba13bSMingming Cao cantfind_ext4: 5104ac27a0ecSDave Kleikamp if (!silent) 5105b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 5106ac27a0ecSDave Kleikamp goto failed_mount; 5107ac27a0ecSDave Kleikamp 510872ba7450STheodore Ts'o failed_mount8: 5109ebd173beSTheodore Ts'o ext4_unregister_sysfs(sb); 5110cb8d53d2SEric Biggers kobject_put(&sbi->s_kobj); 5111dcf2d804STao Ma failed_mount7: 5112dcf2d804STao Ma ext4_unregister_li_request(sb); 5113dcf2d804STao Ma failed_mount6: 5114f9ae9cf5STheodore Ts'o ext4_mb_release(sb); 51157c990728SSuraj Jitindar Singh rcu_read_lock(); 51167c990728SSuraj Jitindar Singh flex_groups = rcu_dereference(sbi->s_flex_groups); 51177c990728SSuraj Jitindar Singh if (flex_groups) { 51187c990728SSuraj Jitindar Singh for (i = 0; i < sbi->s_flex_groups_allocated; i++) 51197c990728SSuraj Jitindar Singh kvfree(flex_groups[i]); 51207c990728SSuraj Jitindar Singh kvfree(flex_groups); 51217c990728SSuraj Jitindar Singh } 51227c990728SSuraj Jitindar Singh rcu_read_unlock(); 5123d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_freeclusters_counter); 5124d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_freeinodes_counter); 5125d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_dirs_counter); 5126d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 5127bbd55937SEric Biggers percpu_free_rwsem(&sbi->s_writepages_rwsem); 512800764937SAzat Khuzhin failed_mount5: 5129f9ae9cf5STheodore Ts'o ext4_ext_release(sb); 5130f9ae9cf5STheodore Ts'o ext4_release_system_zone(sb); 5131f9ae9cf5STheodore Ts'o failed_mount4a: 513294bf608aSAl Viro dput(sb->s_root); 513332a9bb57SManish Katiyar sb->s_root = NULL; 513494bf608aSAl Viro failed_mount4: 5135b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "mount failed"); 51362e8fa54eSJan Kara if (EXT4_SB(sb)->rsv_conversion_wq) 51372e8fa54eSJan Kara destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 51384c0425ffSMingming Cao failed_mount_wq: 5139dec214d0STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 5140dec214d0STahsin Erdogan sbi->s_ea_inode_cache = NULL; 514150c15df6SChengguang Xu 514247387409STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 514347387409STahsin Erdogan sbi->s_ea_block_cache = NULL; 514450c15df6SChengguang Xu 51450390131bSFrank Mayhar if (sbi->s_journal) { 5146dab291afSMingming Cao jbd2_journal_destroy(sbi->s_journal); 514747b4a50bSJan Kara sbi->s_journal = NULL; 51480390131bSFrank Mayhar } 514950460fe8SDarrick J. Wong failed_mount3a: 5150d3922a77SZheng Liu ext4_es_unregister_shrinker(sbi); 5151eb68d0e2SZheng Liu failed_mount3: 51529105bb14SAl Viro del_timer_sync(&sbi->s_err_report); 5153c92dc856SJan Kara flush_work(&sbi->s_error_work); 5154c5e06d10SJohann Lombardi if (sbi->s_mmp_tsk) 5155c5e06d10SJohann Lombardi kthread_stop(sbi->s_mmp_tsk); 5156ac27a0ecSDave Kleikamp failed_mount2: 51571d0c3924STheodore Ts'o rcu_read_lock(); 51581d0c3924STheodore Ts'o group_desc = rcu_dereference(sbi->s_group_desc); 5159ac27a0ecSDave Kleikamp for (i = 0; i < db_count; i++) 51601d0c3924STheodore Ts'o brelse(group_desc[i]); 51611d0c3924STheodore Ts'o kvfree(group_desc); 51621d0c3924STheodore Ts'o rcu_read_unlock(); 5163ac27a0ecSDave Kleikamp failed_mount: 51640441984aSDarrick J. Wong if (sbi->s_chksum_driver) 51650441984aSDarrick J. Wong crypto_free_shash(sbi->s_chksum_driver); 5166c83ad55eSGabriel Krisman Bertazi 5167c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 5168f8f4acb6SDaniel Rosenberg utf8_unload(sb->s_encoding); 5169c83ad55eSGabriel Krisman Bertazi #endif 5170c83ad55eSGabriel Krisman Bertazi 5171ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 5172a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 51730ba33facSTheodore Ts'o kfree(get_qf_name(sb, sbi, i)); 5174ac27a0ecSDave Kleikamp #endif 5175ac4acb1fSEric Biggers fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy); 5176617ba13bSMingming Cao ext4_blkdev_remove(sbi); 5177ac27a0ecSDave Kleikamp brelse(bh); 5178ac27a0ecSDave Kleikamp out_fail: 5179ac27a0ecSDave Kleikamp sb->s_fs_info = NULL; 5180f6830165SManish Katiyar kfree(sbi->s_blockgroup_lock); 51815aee0f8aSTheodore Ts'o out_free_base: 5182ac27a0ecSDave Kleikamp kfree(sbi); 5183d4c402d9SCurt Wohlgemuth kfree(orig_data); 51845e405595SDan Williams fs_put_dax(dax_dev); 518507aa2ea1SLukas Czerner return err ? err : ret; 5186ac27a0ecSDave Kleikamp } 5187ac27a0ecSDave Kleikamp 5188ac27a0ecSDave Kleikamp /* 5189ac27a0ecSDave Kleikamp * Setup any per-fs journal parameters now. We'll do this both on 5190ac27a0ecSDave Kleikamp * initial mount, once the journal has been initialised but before we've 5191ac27a0ecSDave Kleikamp * done any recovery; and again on any subsequent remount. 5192ac27a0ecSDave Kleikamp */ 5193617ba13bSMingming Cao static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 5194ac27a0ecSDave Kleikamp { 5195617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 5196ac27a0ecSDave Kleikamp 5197ac27a0ecSDave Kleikamp journal->j_commit_interval = sbi->s_commit_interval; 519830773840STheodore Ts'o journal->j_min_batch_time = sbi->s_min_batch_time; 519930773840STheodore Ts'o journal->j_max_batch_time = sbi->s_max_batch_time; 52006866d7b3SHarshad Shirwadkar ext4_fc_init(sb, journal); 5201ac27a0ecSDave Kleikamp 5202a931da6aSTheodore Ts'o write_lock(&journal->j_state_lock); 5203ac27a0ecSDave Kleikamp if (test_opt(sb, BARRIER)) 5204dab291afSMingming Cao journal->j_flags |= JBD2_BARRIER; 5205ac27a0ecSDave Kleikamp else 5206dab291afSMingming Cao journal->j_flags &= ~JBD2_BARRIER; 52075bf5683aSHidehiro Kawai if (test_opt(sb, DATA_ERR_ABORT)) 52085bf5683aSHidehiro Kawai journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 52095bf5683aSHidehiro Kawai else 52105bf5683aSHidehiro Kawai journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 5211a931da6aSTheodore Ts'o write_unlock(&journal->j_state_lock); 5212ac27a0ecSDave Kleikamp } 5213ac27a0ecSDave Kleikamp 5214c6cb7e77SEric Whitney static struct inode *ext4_get_journal_inode(struct super_block *sb, 5215ac27a0ecSDave Kleikamp unsigned int journal_inum) 5216ac27a0ecSDave Kleikamp { 5217ac27a0ecSDave Kleikamp struct inode *journal_inode; 5218ac27a0ecSDave Kleikamp 5219c6cb7e77SEric Whitney /* 5220c6cb7e77SEric Whitney * Test for the existence of a valid inode on disk. Bad things 5221c6cb7e77SEric Whitney * happen if we iget() an unused inode, as the subsequent iput() 5222c6cb7e77SEric Whitney * will try to delete it. 5223c6cb7e77SEric Whitney */ 52248a363970STheodore Ts'o journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 52251d1fe1eeSDavid Howells if (IS_ERR(journal_inode)) { 5226b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "no journal found"); 5227ac27a0ecSDave Kleikamp return NULL; 5228ac27a0ecSDave Kleikamp } 5229ac27a0ecSDave Kleikamp if (!journal_inode->i_nlink) { 5230ac27a0ecSDave Kleikamp make_bad_inode(journal_inode); 5231ac27a0ecSDave Kleikamp iput(journal_inode); 5232b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 5233ac27a0ecSDave Kleikamp return NULL; 5234ac27a0ecSDave Kleikamp } 5235ac27a0ecSDave Kleikamp 5236e5f8eab8STheodore Ts'o jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 5237ac27a0ecSDave Kleikamp journal_inode, journal_inode->i_size); 52381d1fe1eeSDavid Howells if (!S_ISREG(journal_inode->i_mode)) { 5239b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "invalid journal inode"); 5240ac27a0ecSDave Kleikamp iput(journal_inode); 5241ac27a0ecSDave Kleikamp return NULL; 5242ac27a0ecSDave Kleikamp } 5243c6cb7e77SEric Whitney return journal_inode; 5244c6cb7e77SEric Whitney } 5245c6cb7e77SEric Whitney 5246c6cb7e77SEric Whitney static journal_t *ext4_get_journal(struct super_block *sb, 5247c6cb7e77SEric Whitney unsigned int journal_inum) 5248c6cb7e77SEric Whitney { 5249c6cb7e77SEric Whitney struct inode *journal_inode; 5250c6cb7e77SEric Whitney journal_t *journal; 5251c6cb7e77SEric Whitney 525211215630SJan Kara if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 525311215630SJan Kara return NULL; 5254c6cb7e77SEric Whitney 5255c6cb7e77SEric Whitney journal_inode = ext4_get_journal_inode(sb, journal_inum); 5256c6cb7e77SEric Whitney if (!journal_inode) 5257c6cb7e77SEric Whitney return NULL; 5258ac27a0ecSDave Kleikamp 5259dab291afSMingming Cao journal = jbd2_journal_init_inode(journal_inode); 5260ac27a0ecSDave Kleikamp if (!journal) { 5261b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 5262ac27a0ecSDave Kleikamp iput(journal_inode); 5263ac27a0ecSDave Kleikamp return NULL; 5264ac27a0ecSDave Kleikamp } 5265ac27a0ecSDave Kleikamp journal->j_private = sb; 5266617ba13bSMingming Cao ext4_init_journal_params(sb, journal); 5267ac27a0ecSDave Kleikamp return journal; 5268ac27a0ecSDave Kleikamp } 5269ac27a0ecSDave Kleikamp 5270617ba13bSMingming Cao static journal_t *ext4_get_dev_journal(struct super_block *sb, 5271ac27a0ecSDave Kleikamp dev_t j_dev) 5272ac27a0ecSDave Kleikamp { 5273ac27a0ecSDave Kleikamp struct buffer_head *bh; 5274ac27a0ecSDave Kleikamp journal_t *journal; 5275617ba13bSMingming Cao ext4_fsblk_t start; 5276617ba13bSMingming Cao ext4_fsblk_t len; 5277ac27a0ecSDave Kleikamp int hblock, blocksize; 5278617ba13bSMingming Cao ext4_fsblk_t sb_block; 5279ac27a0ecSDave Kleikamp unsigned long offset; 5280617ba13bSMingming Cao struct ext4_super_block *es; 5281ac27a0ecSDave Kleikamp struct block_device *bdev; 5282ac27a0ecSDave Kleikamp 528311215630SJan Kara if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 528411215630SJan Kara return NULL; 52850390131bSFrank Mayhar 5286b31e1552SEric Sandeen bdev = ext4_blkdev_get(j_dev, sb); 5287ac27a0ecSDave Kleikamp if (bdev == NULL) 5288ac27a0ecSDave Kleikamp return NULL; 5289ac27a0ecSDave Kleikamp 5290ac27a0ecSDave Kleikamp blocksize = sb->s_blocksize; 5291e1defc4fSMartin K. Petersen hblock = bdev_logical_block_size(bdev); 5292ac27a0ecSDave Kleikamp if (blocksize < hblock) { 5293b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 5294b31e1552SEric Sandeen "blocksize too small for journal device"); 5295ac27a0ecSDave Kleikamp goto out_bdev; 5296ac27a0ecSDave Kleikamp } 5297ac27a0ecSDave Kleikamp 5298617ba13bSMingming Cao sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 5299617ba13bSMingming Cao offset = EXT4_MIN_BLOCK_SIZE % blocksize; 5300ac27a0ecSDave Kleikamp set_blocksize(bdev, blocksize); 5301ac27a0ecSDave Kleikamp if (!(bh = __bread(bdev, sb_block, blocksize))) { 5302b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 5303b31e1552SEric Sandeen "external journal"); 5304ac27a0ecSDave Kleikamp goto out_bdev; 5305ac27a0ecSDave Kleikamp } 5306ac27a0ecSDave Kleikamp 53072716b802STheodore Ts'o es = (struct ext4_super_block *) (bh->b_data + offset); 5308617ba13bSMingming Cao if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 5309ac27a0ecSDave Kleikamp !(le32_to_cpu(es->s_feature_incompat) & 5310617ba13bSMingming Cao EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 5311b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "external journal has " 5312b31e1552SEric Sandeen "bad superblock"); 5313ac27a0ecSDave Kleikamp brelse(bh); 5314ac27a0ecSDave Kleikamp goto out_bdev; 5315ac27a0ecSDave Kleikamp } 5316ac27a0ecSDave Kleikamp 5317df4763beSDarrick J. Wong if ((le32_to_cpu(es->s_feature_ro_compat) & 5318df4763beSDarrick J. Wong EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 5319df4763beSDarrick J. Wong es->s_checksum != ext4_superblock_csum(sb, es)) { 5320df4763beSDarrick J. Wong ext4_msg(sb, KERN_ERR, "external journal has " 5321df4763beSDarrick J. Wong "corrupt superblock"); 5322df4763beSDarrick J. Wong brelse(bh); 5323df4763beSDarrick J. Wong goto out_bdev; 5324df4763beSDarrick J. Wong } 5325df4763beSDarrick J. Wong 5326617ba13bSMingming Cao if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 5327b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 5328ac27a0ecSDave Kleikamp brelse(bh); 5329ac27a0ecSDave Kleikamp goto out_bdev; 5330ac27a0ecSDave Kleikamp } 5331ac27a0ecSDave Kleikamp 5332bd81d8eeSLaurent Vivier len = ext4_blocks_count(es); 5333ac27a0ecSDave Kleikamp start = sb_block + 1; 5334ac27a0ecSDave Kleikamp brelse(bh); /* we're done with the superblock */ 5335ac27a0ecSDave Kleikamp 5336dab291afSMingming Cao journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 5337ac27a0ecSDave Kleikamp start, len, blocksize); 5338ac27a0ecSDave Kleikamp if (!journal) { 5339b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "failed to create device journal"); 5340ac27a0ecSDave Kleikamp goto out_bdev; 5341ac27a0ecSDave Kleikamp } 5342ac27a0ecSDave Kleikamp journal->j_private = sb; 53432d069c08Szhangyi (F) if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) { 5344b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 5345ac27a0ecSDave Kleikamp goto out_journal; 5346ac27a0ecSDave Kleikamp } 5347ac27a0ecSDave Kleikamp if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 5348b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "External journal has more than one " 5349b31e1552SEric Sandeen "user (unsupported) - %d", 5350ac27a0ecSDave Kleikamp be32_to_cpu(journal->j_superblock->s_nr_users)); 5351ac27a0ecSDave Kleikamp goto out_journal; 5352ac27a0ecSDave Kleikamp } 5353ee7ed3aaSChunguang Xu EXT4_SB(sb)->s_journal_bdev = bdev; 5354617ba13bSMingming Cao ext4_init_journal_params(sb, journal); 5355ac27a0ecSDave Kleikamp return journal; 53560b8e58a1SAndreas Dilger 5357ac27a0ecSDave Kleikamp out_journal: 5358dab291afSMingming Cao jbd2_journal_destroy(journal); 5359ac27a0ecSDave Kleikamp out_bdev: 5360617ba13bSMingming Cao ext4_blkdev_put(bdev); 5361ac27a0ecSDave Kleikamp return NULL; 5362ac27a0ecSDave Kleikamp } 5363ac27a0ecSDave Kleikamp 5364617ba13bSMingming Cao static int ext4_load_journal(struct super_block *sb, 5365617ba13bSMingming Cao struct ext4_super_block *es, 5366ac27a0ecSDave Kleikamp unsigned long journal_devnum) 5367ac27a0ecSDave Kleikamp { 5368ac27a0ecSDave Kleikamp journal_t *journal; 5369ac27a0ecSDave Kleikamp unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 5370ac27a0ecSDave Kleikamp dev_t journal_dev; 5371ac27a0ecSDave Kleikamp int err = 0; 5372ac27a0ecSDave Kleikamp int really_read_only; 5373273108faSLukas Czerner int journal_dev_ro; 5374ac27a0ecSDave Kleikamp 537511215630SJan Kara if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 537611215630SJan Kara return -EFSCORRUPTED; 53770390131bSFrank Mayhar 5378ac27a0ecSDave Kleikamp if (journal_devnum && 5379ac27a0ecSDave Kleikamp journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5380b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "external journal device major/minor " 5381b31e1552SEric Sandeen "numbers have changed"); 5382ac27a0ecSDave Kleikamp journal_dev = new_decode_dev(journal_devnum); 5383ac27a0ecSDave Kleikamp } else 5384ac27a0ecSDave Kleikamp journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 5385ac27a0ecSDave Kleikamp 5386273108faSLukas Czerner if (journal_inum && journal_dev) { 5387273108faSLukas Czerner ext4_msg(sb, KERN_ERR, 5388273108faSLukas Czerner "filesystem has both journal inode and journal device!"); 5389273108faSLukas Czerner return -EINVAL; 5390273108faSLukas Czerner } 5391273108faSLukas Czerner 5392273108faSLukas Czerner if (journal_inum) { 5393273108faSLukas Czerner journal = ext4_get_journal(sb, journal_inum); 5394273108faSLukas Czerner if (!journal) 5395273108faSLukas Czerner return -EINVAL; 5396273108faSLukas Czerner } else { 5397273108faSLukas Czerner journal = ext4_get_dev_journal(sb, journal_dev); 5398273108faSLukas Czerner if (!journal) 5399273108faSLukas Czerner return -EINVAL; 5400273108faSLukas Czerner } 5401273108faSLukas Czerner 5402273108faSLukas Czerner journal_dev_ro = bdev_read_only(journal->j_dev); 5403273108faSLukas Czerner really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro; 5404273108faSLukas Czerner 5405273108faSLukas Czerner if (journal_dev_ro && !sb_rdonly(sb)) { 5406273108faSLukas Czerner ext4_msg(sb, KERN_ERR, 5407273108faSLukas Czerner "journal device read-only, try mounting with '-o ro'"); 5408273108faSLukas Czerner err = -EROFS; 5409273108faSLukas Czerner goto err_out; 5410273108faSLukas Czerner } 5411ac27a0ecSDave Kleikamp 5412ac27a0ecSDave Kleikamp /* 5413ac27a0ecSDave Kleikamp * Are we loading a blank journal or performing recovery after a 5414ac27a0ecSDave Kleikamp * crash? For recovery, we need to check in advance whether we 5415ac27a0ecSDave Kleikamp * can get read-write access to the device. 5416ac27a0ecSDave Kleikamp */ 5417e2b911c5SDarrick J. Wong if (ext4_has_feature_journal_needs_recovery(sb)) { 5418bc98a42cSDavid Howells if (sb_rdonly(sb)) { 5419b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "INFO: recovery " 5420b31e1552SEric Sandeen "required on readonly filesystem"); 5421ac27a0ecSDave Kleikamp if (really_read_only) { 5422b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "write access " 5423d98bf8cdSSimon Ruderich "unavailable, cannot proceed " 5424d98bf8cdSSimon Ruderich "(try mounting with noload)"); 5425273108faSLukas Czerner err = -EROFS; 5426273108faSLukas Czerner goto err_out; 5427ac27a0ecSDave Kleikamp } 5428b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "write access will " 5429b31e1552SEric Sandeen "be enabled during recovery"); 5430ac27a0ecSDave Kleikamp } 5431ac27a0ecSDave Kleikamp } 5432ac27a0ecSDave Kleikamp 543390576c0bSTheodore Ts'o if (!(journal->j_flags & JBD2_BARRIER)) 5434b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "barriers disabled"); 54354776004fSTheodore Ts'o 5436e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal_needs_recovery(sb)) 5437dab291afSMingming Cao err = jbd2_journal_wipe(journal, !really_read_only); 54381c13d5c0STheodore Ts'o if (!err) { 54391c13d5c0STheodore Ts'o char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 54401c13d5c0STheodore Ts'o if (save) 54411c13d5c0STheodore Ts'o memcpy(save, ((char *) es) + 54421c13d5c0STheodore Ts'o EXT4_S_ERR_START, EXT4_S_ERR_LEN); 5443dab291afSMingming Cao err = jbd2_journal_load(journal); 54441c13d5c0STheodore Ts'o if (save) 54451c13d5c0STheodore Ts'o memcpy(((char *) es) + EXT4_S_ERR_START, 54461c13d5c0STheodore Ts'o save, EXT4_S_ERR_LEN); 54471c13d5c0STheodore Ts'o kfree(save); 54481c13d5c0STheodore Ts'o } 5449ac27a0ecSDave Kleikamp 5450ac27a0ecSDave Kleikamp if (err) { 5451b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "error loading journal"); 5452273108faSLukas Czerner goto err_out; 5453ac27a0ecSDave Kleikamp } 5454ac27a0ecSDave Kleikamp 5455617ba13bSMingming Cao EXT4_SB(sb)->s_journal = journal; 545611215630SJan Kara err = ext4_clear_journal_err(sb, es); 545711215630SJan Kara if (err) { 545811215630SJan Kara EXT4_SB(sb)->s_journal = NULL; 545911215630SJan Kara jbd2_journal_destroy(journal); 546011215630SJan Kara return err; 546111215630SJan Kara } 5462ac27a0ecSDave Kleikamp 5463c41303ceSMaciej Żenczykowski if (!really_read_only && journal_devnum && 5464ac27a0ecSDave Kleikamp journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5465ac27a0ecSDave Kleikamp es->s_journal_dev = cpu_to_le32(journal_devnum); 5466ac27a0ecSDave Kleikamp 5467ac27a0ecSDave Kleikamp /* Make sure we flush the recovery flag to disk. */ 54684392fbc4SJan Kara ext4_commit_super(sb); 5469ac27a0ecSDave Kleikamp } 5470ac27a0ecSDave Kleikamp 5471ac27a0ecSDave Kleikamp return 0; 5472273108faSLukas Czerner 5473273108faSLukas Czerner err_out: 5474273108faSLukas Czerner jbd2_journal_destroy(journal); 5475273108faSLukas Czerner return err; 5476ac27a0ecSDave Kleikamp } 5477ac27a0ecSDave Kleikamp 54782d01ddc8SJan Kara /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */ 54792d01ddc8SJan Kara static void ext4_update_super(struct super_block *sb) 5480ac27a0ecSDave Kleikamp { 5481c92dc856SJan Kara struct ext4_sb_info *sbi = EXT4_SB(sb); 5482e92ad03fSJan Kara struct ext4_super_block *es = sbi->s_es; 5483e92ad03fSJan Kara struct buffer_head *sbh = sbi->s_sbh; 5484ac27a0ecSDave Kleikamp 548505c2c00fSJan Kara lock_buffer(sbh); 5486a17712c8SJon Derrick /* 548771290b36STheodore Ts'o * If the file system is mounted read-only, don't update the 548871290b36STheodore Ts'o * superblock write time. This avoids updating the superblock 548971290b36STheodore Ts'o * write time when we are mounting the root file system 549071290b36STheodore Ts'o * read/only but we need to replay the journal; at that point, 549171290b36STheodore Ts'o * for people who are east of GMT and who make their clock 549271290b36STheodore Ts'o * tick in localtime for Windows bug-for-bug compatibility, 549371290b36STheodore Ts'o * the clock is set in the future, and this will cause e2fsck 549471290b36STheodore Ts'o * to complain and force a full file system check. 549571290b36STheodore Ts'o */ 54961751e8a6SLinus Torvalds if (!(sb->s_flags & SB_RDONLY)) 54976a0678a7SArnd Bergmann ext4_update_tstamp(es, s_wtime); 5498afc32f7eSTheodore Ts'o es->s_kbytes_written = 5499e92ad03fSJan Kara cpu_to_le64(sbi->s_kbytes_written + 55008446fe92SChristoph Hellwig ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) - 5501e92ad03fSJan Kara sbi->s_sectors_written_start) >> 1)); 5502e92ad03fSJan Kara if (percpu_counter_initialized(&sbi->s_freeclusters_counter)) 550357042651STheodore Ts'o ext4_free_blocks_count_set(es, 5504e92ad03fSJan Kara EXT4_C2B(sbi, percpu_counter_sum_positive( 5505e92ad03fSJan Kara &sbi->s_freeclusters_counter))); 5506e92ad03fSJan Kara if (percpu_counter_initialized(&sbi->s_freeinodes_counter)) 55077f93cff9STheodore Ts'o es->s_free_inodes_count = 55087f93cff9STheodore Ts'o cpu_to_le32(percpu_counter_sum_positive( 5509e92ad03fSJan Kara &sbi->s_freeinodes_counter)); 5510c92dc856SJan Kara /* Copy error information to the on-disk superblock */ 5511c92dc856SJan Kara spin_lock(&sbi->s_error_lock); 5512c92dc856SJan Kara if (sbi->s_add_error_count > 0) { 5513c92dc856SJan Kara es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 5514c92dc856SJan Kara if (!es->s_first_error_time && !es->s_first_error_time_hi) { 5515c92dc856SJan Kara __ext4_update_tstamp(&es->s_first_error_time, 5516c92dc856SJan Kara &es->s_first_error_time_hi, 5517c92dc856SJan Kara sbi->s_first_error_time); 5518c92dc856SJan Kara strncpy(es->s_first_error_func, sbi->s_first_error_func, 5519c92dc856SJan Kara sizeof(es->s_first_error_func)); 5520c92dc856SJan Kara es->s_first_error_line = 5521c92dc856SJan Kara cpu_to_le32(sbi->s_first_error_line); 5522c92dc856SJan Kara es->s_first_error_ino = 5523c92dc856SJan Kara cpu_to_le32(sbi->s_first_error_ino); 5524c92dc856SJan Kara es->s_first_error_block = 5525c92dc856SJan Kara cpu_to_le64(sbi->s_first_error_block); 5526c92dc856SJan Kara es->s_first_error_errcode = 5527c92dc856SJan Kara ext4_errno_to_code(sbi->s_first_error_code); 5528c92dc856SJan Kara } 5529c92dc856SJan Kara __ext4_update_tstamp(&es->s_last_error_time, 5530c92dc856SJan Kara &es->s_last_error_time_hi, 5531c92dc856SJan Kara sbi->s_last_error_time); 5532c92dc856SJan Kara strncpy(es->s_last_error_func, sbi->s_last_error_func, 5533c92dc856SJan Kara sizeof(es->s_last_error_func)); 5534c92dc856SJan Kara es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line); 5535c92dc856SJan Kara es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino); 5536c92dc856SJan Kara es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block); 5537c92dc856SJan Kara es->s_last_error_errcode = 5538c92dc856SJan Kara ext4_errno_to_code(sbi->s_last_error_code); 5539c92dc856SJan Kara /* 5540c92dc856SJan Kara * Start the daily error reporting function if it hasn't been 5541c92dc856SJan Kara * started already 5542c92dc856SJan Kara */ 5543c92dc856SJan Kara if (!es->s_error_count) 5544c92dc856SJan Kara mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); 5545c92dc856SJan Kara le32_add_cpu(&es->s_error_count, sbi->s_add_error_count); 5546c92dc856SJan Kara sbi->s_add_error_count = 0; 5547c92dc856SJan Kara } 5548c92dc856SJan Kara spin_unlock(&sbi->s_error_lock); 5549c92dc856SJan Kara 555006db49e6STheodore Ts'o ext4_superblock_csum_set(sb); 55512d01ddc8SJan Kara unlock_buffer(sbh); 55522d01ddc8SJan Kara } 55532d01ddc8SJan Kara 55542d01ddc8SJan Kara static int ext4_commit_super(struct super_block *sb) 55552d01ddc8SJan Kara { 55562d01ddc8SJan Kara struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 55572d01ddc8SJan Kara int error = 0; 55582d01ddc8SJan Kara 55592d01ddc8SJan Kara if (!sbh || block_device_ejected(sb)) 55602d01ddc8SJan Kara return error; 55612d01ddc8SJan Kara 55622d01ddc8SJan Kara ext4_update_super(sb); 55632d01ddc8SJan Kara 5564e8680786STheodore Ts'o if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 55654743f839SPranay Kr. Srivastava /* 55664743f839SPranay Kr. Srivastava * Oh, dear. A previous attempt to write the 55674743f839SPranay Kr. Srivastava * superblock failed. This could happen because the 55684743f839SPranay Kr. Srivastava * USB device was yanked out. Or it could happen to 55694743f839SPranay Kr. Srivastava * be a transient write error and maybe the block will 55704743f839SPranay Kr. Srivastava * be remapped. Nothing we can do but to retry the 55714743f839SPranay Kr. Srivastava * write and hope for the best. 55724743f839SPranay Kr. Srivastava */ 55734743f839SPranay Kr. Srivastava ext4_msg(sb, KERN_ERR, "previous I/O error to " 55744743f839SPranay Kr. Srivastava "superblock detected"); 55754743f839SPranay Kr. Srivastava clear_buffer_write_io_error(sbh); 55764743f839SPranay Kr. Srivastava set_buffer_uptodate(sbh); 55774743f839SPranay Kr. Srivastava } 55782d01ddc8SJan Kara BUFFER_TRACE(sbh, "marking dirty"); 5579ac27a0ecSDave Kleikamp mark_buffer_dirty(sbh); 5580564bc402SDaeho Jeong error = __sync_dirty_buffer(sbh, 558100473374SJan Kara REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0)); 5582c89128a0SJaegeuk Kim if (buffer_write_io_error(sbh)) { 5583b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "I/O error while writing " 5584b31e1552SEric Sandeen "superblock"); 5585914258bfSTheodore Ts'o clear_buffer_write_io_error(sbh); 5586914258bfSTheodore Ts'o set_buffer_uptodate(sbh); 5587914258bfSTheodore Ts'o } 5588c4be0c1dSTakashi Sato return error; 5589ac27a0ecSDave Kleikamp } 5590ac27a0ecSDave Kleikamp 5591ac27a0ecSDave Kleikamp /* 5592ac27a0ecSDave Kleikamp * Have we just finished recovery? If so, and if we are mounting (or 5593ac27a0ecSDave Kleikamp * remounting) the filesystem readonly, then we will end up with a 5594ac27a0ecSDave Kleikamp * consistent fs on disk. Record that fact. 5595ac27a0ecSDave Kleikamp */ 559611215630SJan Kara static int ext4_mark_recovery_complete(struct super_block *sb, 5597617ba13bSMingming Cao struct ext4_super_block *es) 5598ac27a0ecSDave Kleikamp { 559911215630SJan Kara int err; 5600617ba13bSMingming Cao journal_t *journal = EXT4_SB(sb)->s_journal; 5601ac27a0ecSDave Kleikamp 5602e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal(sb)) { 560311215630SJan Kara if (journal != NULL) { 560411215630SJan Kara ext4_error(sb, "Journal got removed while the fs was " 560511215630SJan Kara "mounted!"); 560611215630SJan Kara return -EFSCORRUPTED; 560711215630SJan Kara } 560811215630SJan Kara return 0; 56090390131bSFrank Mayhar } 5610dab291afSMingming Cao jbd2_journal_lock_updates(journal); 561111215630SJan Kara err = jbd2_journal_flush(journal); 561211215630SJan Kara if (err < 0) 56137ffe1ea8SHidehiro Kawai goto out; 56147ffe1ea8SHidehiro Kawai 5615bc98a42cSDavid Howells if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) { 5616e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 56174392fbc4SJan Kara ext4_commit_super(sb); 5618ac27a0ecSDave Kleikamp } 56197ffe1ea8SHidehiro Kawai out: 5620dab291afSMingming Cao jbd2_journal_unlock_updates(journal); 562111215630SJan Kara return err; 5622ac27a0ecSDave Kleikamp } 5623ac27a0ecSDave Kleikamp 5624ac27a0ecSDave Kleikamp /* 5625ac27a0ecSDave Kleikamp * If we are mounting (or read-write remounting) a filesystem whose journal 5626ac27a0ecSDave Kleikamp * has recorded an error from a previous lifetime, move that error to the 5627ac27a0ecSDave Kleikamp * main filesystem now. 5628ac27a0ecSDave Kleikamp */ 562911215630SJan Kara static int ext4_clear_journal_err(struct super_block *sb, 5630617ba13bSMingming Cao struct ext4_super_block *es) 5631ac27a0ecSDave Kleikamp { 5632ac27a0ecSDave Kleikamp journal_t *journal; 5633ac27a0ecSDave Kleikamp int j_errno; 5634ac27a0ecSDave Kleikamp const char *errstr; 5635ac27a0ecSDave Kleikamp 563611215630SJan Kara if (!ext4_has_feature_journal(sb)) { 563711215630SJan Kara ext4_error(sb, "Journal got removed while the fs was mounted!"); 563811215630SJan Kara return -EFSCORRUPTED; 563911215630SJan Kara } 56400390131bSFrank Mayhar 5641617ba13bSMingming Cao journal = EXT4_SB(sb)->s_journal; 5642ac27a0ecSDave Kleikamp 5643ac27a0ecSDave Kleikamp /* 5644ac27a0ecSDave Kleikamp * Now check for any error status which may have been recorded in the 5645617ba13bSMingming Cao * journal by a prior ext4_error() or ext4_abort() 5646ac27a0ecSDave Kleikamp */ 5647ac27a0ecSDave Kleikamp 5648dab291afSMingming Cao j_errno = jbd2_journal_errno(journal); 5649ac27a0ecSDave Kleikamp if (j_errno) { 5650ac27a0ecSDave Kleikamp char nbuf[16]; 5651ac27a0ecSDave Kleikamp 5652617ba13bSMingming Cao errstr = ext4_decode_error(sb, j_errno, nbuf); 565312062dddSEric Sandeen ext4_warning(sb, "Filesystem error recorded " 5654ac27a0ecSDave Kleikamp "from previous mount: %s", errstr); 565512062dddSEric Sandeen ext4_warning(sb, "Marking fs in need of filesystem check."); 5656ac27a0ecSDave Kleikamp 5657617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 5658617ba13bSMingming Cao es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 56594392fbc4SJan Kara ext4_commit_super(sb); 5660ac27a0ecSDave Kleikamp 5661dab291afSMingming Cao jbd2_journal_clear_err(journal); 5662d796c52eSTheodore Ts'o jbd2_journal_update_sb_errno(journal); 5663ac27a0ecSDave Kleikamp } 566411215630SJan Kara return 0; 5665ac27a0ecSDave Kleikamp } 5666ac27a0ecSDave Kleikamp 5667ac27a0ecSDave Kleikamp /* 5668ac27a0ecSDave Kleikamp * Force the running and committing transactions to commit, 5669ac27a0ecSDave Kleikamp * and wait on the commit. 5670ac27a0ecSDave Kleikamp */ 5671617ba13bSMingming Cao int ext4_force_commit(struct super_block *sb) 5672ac27a0ecSDave Kleikamp { 5673ac27a0ecSDave Kleikamp journal_t *journal; 5674ac27a0ecSDave Kleikamp 5675bc98a42cSDavid Howells if (sb_rdonly(sb)) 5676ac27a0ecSDave Kleikamp return 0; 5677ac27a0ecSDave Kleikamp 5678617ba13bSMingming Cao journal = EXT4_SB(sb)->s_journal; 5679b1deefc9SGuo Chao return ext4_journal_force_commit(journal); 5680ac27a0ecSDave Kleikamp } 5681ac27a0ecSDave Kleikamp 5682617ba13bSMingming Cao static int ext4_sync_fs(struct super_block *sb, int wait) 5683ac27a0ecSDave Kleikamp { 568414ce0cb4STheodore Ts'o int ret = 0; 56859eddacf9SJan Kara tid_t target; 568606a407f1SDmitry Monakhov bool needs_barrier = false; 56878d5d02e6SMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 5688ac27a0ecSDave Kleikamp 568949598e04SJun Piao if (unlikely(ext4_forced_shutdown(sbi))) 56900db1ff22STheodore Ts'o return 0; 56910db1ff22STheodore Ts'o 56929bffad1eSTheodore Ts'o trace_ext4_sync_fs(sb, wait); 56932e8fa54eSJan Kara flush_workqueue(sbi->rsv_conversion_wq); 5694a1177825SJan Kara /* 5695a1177825SJan Kara * Writeback quota in non-journalled quota case - journalled quota has 5696a1177825SJan Kara * no dirty dquots 5697a1177825SJan Kara */ 5698a1177825SJan Kara dquot_writeback_dquots(sb, -1); 569906a407f1SDmitry Monakhov /* 570006a407f1SDmitry Monakhov * Data writeback is possible w/o journal transaction, so barrier must 570106a407f1SDmitry Monakhov * being sent at the end of the function. But we can skip it if 570206a407f1SDmitry Monakhov * transaction_commit will do it for us. 570306a407f1SDmitry Monakhov */ 5704bda32530STheodore Ts'o if (sbi->s_journal) { 570506a407f1SDmitry Monakhov target = jbd2_get_latest_transaction(sbi->s_journal); 570606a407f1SDmitry Monakhov if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 570706a407f1SDmitry Monakhov !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 570806a407f1SDmitry Monakhov needs_barrier = true; 570906a407f1SDmitry Monakhov 57108d5d02e6SMingming Cao if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 5711ac27a0ecSDave Kleikamp if (wait) 5712bda32530STheodore Ts'o ret = jbd2_log_wait_commit(sbi->s_journal, 5713bda32530STheodore Ts'o target); 57140390131bSFrank Mayhar } 5715bda32530STheodore Ts'o } else if (wait && test_opt(sb, BARRIER)) 5716bda32530STheodore Ts'o needs_barrier = true; 571706a407f1SDmitry Monakhov if (needs_barrier) { 571806a407f1SDmitry Monakhov int err; 5719c6bf3f0eSChristoph Hellwig err = blkdev_issue_flush(sb->s_bdev); 572006a407f1SDmitry Monakhov if (!ret) 572106a407f1SDmitry Monakhov ret = err; 572206a407f1SDmitry Monakhov } 572306a407f1SDmitry Monakhov 572406a407f1SDmitry Monakhov return ret; 572506a407f1SDmitry Monakhov } 572606a407f1SDmitry Monakhov 5727ac27a0ecSDave Kleikamp /* 5728ac27a0ecSDave Kleikamp * LVM calls this function before a (read-only) snapshot is created. This 5729ac27a0ecSDave Kleikamp * gives us a chance to flush the journal completely and mark the fs clean. 5730be4f27d3SYongqiang Yang * 5731be4f27d3SYongqiang Yang * Note that only this function cannot bring a filesystem to be in a clean 57328e8ad8a5SJan Kara * state independently. It relies on upper layer to stop all data & metadata 57338e8ad8a5SJan Kara * modifications. 5734ac27a0ecSDave Kleikamp */ 5735c4be0c1dSTakashi Sato static int ext4_freeze(struct super_block *sb) 5736ac27a0ecSDave Kleikamp { 5737c4be0c1dSTakashi Sato int error = 0; 5738c4be0c1dSTakashi Sato journal_t *journal; 5739ac27a0ecSDave Kleikamp 5740bc98a42cSDavid Howells if (sb_rdonly(sb)) 57419ca92389STheodore Ts'o return 0; 57429ca92389STheodore Ts'o 5743c4be0c1dSTakashi Sato journal = EXT4_SB(sb)->s_journal; 5744ac27a0ecSDave Kleikamp 5745bb044576STheodore Ts'o if (journal) { 5746ac27a0ecSDave Kleikamp /* Now we set up the journal barrier. */ 5747dab291afSMingming Cao jbd2_journal_lock_updates(journal); 57487ffe1ea8SHidehiro Kawai 57497ffe1ea8SHidehiro Kawai /* 5750bb044576STheodore Ts'o * Don't clear the needs_recovery flag if we failed to 5751bb044576STheodore Ts'o * flush the journal. 57527ffe1ea8SHidehiro Kawai */ 5753c4be0c1dSTakashi Sato error = jbd2_journal_flush(journal); 57546b0310fbSEric Sandeen if (error < 0) 57556b0310fbSEric Sandeen goto out; 5756ac27a0ecSDave Kleikamp 5757ac27a0ecSDave Kleikamp /* Journal blocked and flushed, clear needs_recovery flag. */ 5758e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 5759c642dc9eSEric Sandeen } 5760c642dc9eSEric Sandeen 57614392fbc4SJan Kara error = ext4_commit_super(sb); 57626b0310fbSEric Sandeen out: 5763bb044576STheodore Ts'o if (journal) 57648e8ad8a5SJan Kara /* we rely on upper layer to stop further updates */ 5765bb044576STheodore Ts'o jbd2_journal_unlock_updates(journal); 57666b0310fbSEric Sandeen return error; 5767ac27a0ecSDave Kleikamp } 5768ac27a0ecSDave Kleikamp 5769ac27a0ecSDave Kleikamp /* 5770ac27a0ecSDave Kleikamp * Called by LVM after the snapshot is done. We need to reset the RECOVER 5771ac27a0ecSDave Kleikamp * flag here, even though the filesystem is not technically dirty yet. 5772ac27a0ecSDave Kleikamp */ 5773c4be0c1dSTakashi Sato static int ext4_unfreeze(struct super_block *sb) 5774ac27a0ecSDave Kleikamp { 5775bc98a42cSDavid Howells if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) 57769ca92389STheodore Ts'o return 0; 57779ca92389STheodore Ts'o 5778c642dc9eSEric Sandeen if (EXT4_SB(sb)->s_journal) { 57799ca92389STheodore Ts'o /* Reset the needs_recovery flag before the fs is unlocked. */ 5780e2b911c5SDarrick J. Wong ext4_set_feature_journal_needs_recovery(sb); 5781c642dc9eSEric Sandeen } 5782c642dc9eSEric Sandeen 57834392fbc4SJan Kara ext4_commit_super(sb); 5784c4be0c1dSTakashi Sato return 0; 5785ac27a0ecSDave Kleikamp } 5786ac27a0ecSDave Kleikamp 5787673c6100STheodore Ts'o /* 5788673c6100STheodore Ts'o * Structure to save mount options for ext4_remount's benefit 5789673c6100STheodore Ts'o */ 5790673c6100STheodore Ts'o struct ext4_mount_options { 5791673c6100STheodore Ts'o unsigned long s_mount_opt; 5792a2595b8aSTheodore Ts'o unsigned long s_mount_opt2; 579308cefc7aSEric W. Biederman kuid_t s_resuid; 579408cefc7aSEric W. Biederman kgid_t s_resgid; 5795673c6100STheodore Ts'o unsigned long s_commit_interval; 5796673c6100STheodore Ts'o u32 s_min_batch_time, s_max_batch_time; 5797673c6100STheodore Ts'o #ifdef CONFIG_QUOTA 5798673c6100STheodore Ts'o int s_jquota_fmt; 5799a2d4a646SJan Kara char *s_qf_names[EXT4_MAXQUOTAS]; 5800673c6100STheodore Ts'o #endif 5801673c6100STheodore Ts'o }; 5802673c6100STheodore Ts'o 5803617ba13bSMingming Cao static int ext4_remount(struct super_block *sb, int *flags, char *data) 5804ac27a0ecSDave Kleikamp { 5805617ba13bSMingming Cao struct ext4_super_block *es; 5806617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 5807f25391ebSLukas Czerner unsigned long old_sb_flags, vfs_flags; 5808617ba13bSMingming Cao struct ext4_mount_options old_opts; 5809c79d967dSChristoph Hellwig int enable_quota = 0; 58108a266467STheodore Ts'o ext4_group_t g; 5811b3881f74STheodore Ts'o unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 5812c5e06d10SJohann Lombardi int err = 0; 5813ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 581403dafb5fSChen Gang int i, j; 581533458eabSTheodore Ts'o char *to_free[EXT4_MAXQUOTAS]; 5816ac27a0ecSDave Kleikamp #endif 5817d4c402d9SCurt Wohlgemuth char *orig_data = kstrdup(data, GFP_KERNEL); 5818ac27a0ecSDave Kleikamp 581921ac738eSChengguang Xu if (data && !orig_data) 582021ac738eSChengguang Xu return -ENOMEM; 582121ac738eSChengguang Xu 5822ac27a0ecSDave Kleikamp /* Store the original options */ 5823ac27a0ecSDave Kleikamp old_sb_flags = sb->s_flags; 5824ac27a0ecSDave Kleikamp old_opts.s_mount_opt = sbi->s_mount_opt; 5825a2595b8aSTheodore Ts'o old_opts.s_mount_opt2 = sbi->s_mount_opt2; 5826ac27a0ecSDave Kleikamp old_opts.s_resuid = sbi->s_resuid; 5827ac27a0ecSDave Kleikamp old_opts.s_resgid = sbi->s_resgid; 5828ac27a0ecSDave Kleikamp old_opts.s_commit_interval = sbi->s_commit_interval; 582930773840STheodore Ts'o old_opts.s_min_batch_time = sbi->s_min_batch_time; 583030773840STheodore Ts'o old_opts.s_max_batch_time = sbi->s_max_batch_time; 5831ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 5832ac27a0ecSDave Kleikamp old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 5833a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 583403dafb5fSChen Gang if (sbi->s_qf_names[i]) { 583533458eabSTheodore Ts'o char *qf_name = get_qf_name(sb, sbi, i); 583633458eabSTheodore Ts'o 583733458eabSTheodore Ts'o old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 583803dafb5fSChen Gang if (!old_opts.s_qf_names[i]) { 583903dafb5fSChen Gang for (j = 0; j < i; j++) 584003dafb5fSChen Gang kfree(old_opts.s_qf_names[j]); 58413e36a163SWei Yongjun kfree(orig_data); 584203dafb5fSChen Gang return -ENOMEM; 584303dafb5fSChen Gang } 584403dafb5fSChen Gang } else 584503dafb5fSChen Gang old_opts.s_qf_names[i] = NULL; 5846ac27a0ecSDave Kleikamp #endif 5847b3881f74STheodore Ts'o if (sbi->s_journal && sbi->s_journal->j_task->io_context) 5848b3881f74STheodore Ts'o journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 5849ac27a0ecSDave Kleikamp 5850f25391ebSLukas Czerner /* 5851f25391ebSLukas Czerner * Some options can be enabled by ext4 and/or by VFS mount flag 5852f25391ebSLukas Czerner * either way we need to make sure it matches in both *flags and 5853f25391ebSLukas Czerner * s_flags. Copy those selected flags from *flags to s_flags 5854f25391ebSLukas Czerner */ 5855f25391ebSLukas Czerner vfs_flags = SB_LAZYTIME | SB_I_VERSION; 5856f25391ebSLukas Czerner sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags); 5857f25391ebSLukas Czerner 5858661aa520SEric Sandeen if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 5859ac27a0ecSDave Kleikamp err = -EINVAL; 5860ac27a0ecSDave Kleikamp goto restore_opts; 5861ac27a0ecSDave Kleikamp } 5862ac27a0ecSDave Kleikamp 58636b992ff2SDarrick J. Wong if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 58646b992ff2SDarrick J. Wong test_opt(sb, JOURNAL_CHECKSUM)) { 58656b992ff2SDarrick J. Wong ext4_msg(sb, KERN_ERR, "changing journal_checksum " 58662d5b86e0SEric Sandeen "during remount not supported; ignoring"); 58672d5b86e0SEric Sandeen sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 5868c6d3d56dSDarrick J. Wong } 5869c6d3d56dSDarrick J. Wong 58706ae6514bSPiotr Sarna if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 58716ae6514bSPiotr Sarna if (test_opt2(sb, EXPLICIT_DELALLOC)) { 58726ae6514bSPiotr Sarna ext4_msg(sb, KERN_ERR, "can't mount with " 58736ae6514bSPiotr Sarna "both data=journal and delalloc"); 58746ae6514bSPiotr Sarna err = -EINVAL; 58756ae6514bSPiotr Sarna goto restore_opts; 58766ae6514bSPiotr Sarna } 58776ae6514bSPiotr Sarna if (test_opt(sb, DIOREAD_NOLOCK)) { 58786ae6514bSPiotr Sarna ext4_msg(sb, KERN_ERR, "can't mount with " 58796ae6514bSPiotr Sarna "both data=journal and dioread_nolock"); 58806ae6514bSPiotr Sarna err = -EINVAL; 58816ae6514bSPiotr Sarna goto restore_opts; 58826ae6514bSPiotr Sarna } 5883ab04df78SJan Kara } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 5884ab04df78SJan Kara if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5885ab04df78SJan Kara ext4_msg(sb, KERN_ERR, "can't mount with " 5886ab04df78SJan Kara "journal_async_commit in data=ordered mode"); 5887ab04df78SJan Kara err = -EINVAL; 5888ab04df78SJan Kara goto restore_opts; 5889ab04df78SJan Kara } 5890923ae0ffSRoss Zwisler } 5891923ae0ffSRoss Zwisler 5892cdb7ee4cSTahsin Erdogan if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 5893cdb7ee4cSTahsin Erdogan ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 5894cdb7ee4cSTahsin Erdogan err = -EINVAL; 5895cdb7ee4cSTahsin Erdogan goto restore_opts; 5896cdb7ee4cSTahsin Erdogan } 5897cdb7ee4cSTahsin Erdogan 58989b5f6c9bSHarshad Shirwadkar if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) 589954d3adbcSTheodore Ts'o ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user"); 5900ac27a0ecSDave Kleikamp 59011751e8a6SLinus Torvalds sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 59021751e8a6SLinus Torvalds (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 5903ac27a0ecSDave Kleikamp 5904ac27a0ecSDave Kleikamp es = sbi->s_es; 5905ac27a0ecSDave Kleikamp 5906b3881f74STheodore Ts'o if (sbi->s_journal) { 5907617ba13bSMingming Cao ext4_init_journal_params(sb, sbi->s_journal); 5908b3881f74STheodore Ts'o set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 5909b3881f74STheodore Ts'o } 5910ac27a0ecSDave Kleikamp 5911c92dc856SJan Kara /* Flush outstanding errors before changing fs state */ 5912c92dc856SJan Kara flush_work(&sbi->s_error_work); 5913c92dc856SJan Kara 59141751e8a6SLinus Torvalds if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) { 59159b5f6c9bSHarshad Shirwadkar if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) { 5916ac27a0ecSDave Kleikamp err = -EROFS; 5917ac27a0ecSDave Kleikamp goto restore_opts; 5918ac27a0ecSDave Kleikamp } 5919ac27a0ecSDave Kleikamp 59201751e8a6SLinus Torvalds if (*flags & SB_RDONLY) { 592138c03b34STheodore Ts'o err = sync_filesystem(sb); 592238c03b34STheodore Ts'o if (err < 0) 592338c03b34STheodore Ts'o goto restore_opts; 59240f0dd62fSChristoph Hellwig err = dquot_suspend(sb, -1); 59250f0dd62fSChristoph Hellwig if (err < 0) 5926c79d967dSChristoph Hellwig goto restore_opts; 5927c79d967dSChristoph Hellwig 5928ac27a0ecSDave Kleikamp /* 5929ac27a0ecSDave Kleikamp * First of all, the unconditional stuff we have to do 5930ac27a0ecSDave Kleikamp * to disable replay of the journal when we next remount 5931ac27a0ecSDave Kleikamp */ 59321751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 5933ac27a0ecSDave Kleikamp 5934ac27a0ecSDave Kleikamp /* 5935ac27a0ecSDave Kleikamp * OK, test if we are remounting a valid rw partition 5936ac27a0ecSDave Kleikamp * readonly, and if so set the rdonly flag and then 5937ac27a0ecSDave Kleikamp * mark the partition as valid again. 5938ac27a0ecSDave Kleikamp */ 5939617ba13bSMingming Cao if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 5940617ba13bSMingming Cao (sbi->s_mount_state & EXT4_VALID_FS)) 5941ac27a0ecSDave Kleikamp es->s_state = cpu_to_le16(sbi->s_mount_state); 5942ac27a0ecSDave Kleikamp 594311215630SJan Kara if (sbi->s_journal) { 594411215630SJan Kara /* 594511215630SJan Kara * We let remount-ro finish even if marking fs 594611215630SJan Kara * as clean failed... 594711215630SJan Kara */ 5948617ba13bSMingming Cao ext4_mark_recovery_complete(sb, es); 594911215630SJan Kara } 59502dca60d9STheodore Ts'o if (sbi->s_mmp_tsk) 59512dca60d9STheodore Ts'o kthread_stop(sbi->s_mmp_tsk); 5952ac27a0ecSDave Kleikamp } else { 5953a13fb1a4SEric Sandeen /* Make sure we can mount this feature set readwrite */ 5954e2b911c5SDarrick J. Wong if (ext4_has_feature_readonly(sb) || 59552cb5cc8bSDarrick J. Wong !ext4_feature_set_ok(sb, 0)) { 5956ac27a0ecSDave Kleikamp err = -EROFS; 5957ac27a0ecSDave Kleikamp goto restore_opts; 5958ac27a0ecSDave Kleikamp } 5959ead6596bSEric Sandeen /* 59608a266467STheodore Ts'o * Make sure the group descriptor checksums 59610b8e58a1SAndreas Dilger * are sane. If they aren't, refuse to remount r/w. 59628a266467STheodore Ts'o */ 59638a266467STheodore Ts'o for (g = 0; g < sbi->s_groups_count; g++) { 59648a266467STheodore Ts'o struct ext4_group_desc *gdp = 59658a266467STheodore Ts'o ext4_get_group_desc(sb, g, NULL); 59668a266467STheodore Ts'o 5967feb0ab32SDarrick J. Wong if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 5968b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 5969b31e1552SEric Sandeen "ext4_remount: Checksum for group %u failed (%u!=%u)", 5970e2b911c5SDarrick J. Wong g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 59718a266467STheodore Ts'o le16_to_cpu(gdp->bg_checksum)); 59726a797d27SDarrick J. Wong err = -EFSBADCRC; 59738a266467STheodore Ts'o goto restore_opts; 59748a266467STheodore Ts'o } 59758a266467STheodore Ts'o } 59768a266467STheodore Ts'o 59778a266467STheodore Ts'o /* 5978ead6596bSEric Sandeen * If we have an unprocessed orphan list hanging 5979ead6596bSEric Sandeen * around from a previously readonly bdev mount, 5980ead6596bSEric Sandeen * require a full umount/remount for now. 5981ead6596bSEric Sandeen */ 5982ead6596bSEric Sandeen if (es->s_last_orphan) { 5983b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "Couldn't " 5984ead6596bSEric Sandeen "remount RDWR because of unprocessed " 5985ead6596bSEric Sandeen "orphan inode list. Please " 5986b31e1552SEric Sandeen "umount/remount instead"); 5987ead6596bSEric Sandeen err = -EINVAL; 5988ead6596bSEric Sandeen goto restore_opts; 5989ead6596bSEric Sandeen } 5990ead6596bSEric Sandeen 5991ac27a0ecSDave Kleikamp /* 5992ac27a0ecSDave Kleikamp * Mounting a RDONLY partition read-write, so reread 5993ac27a0ecSDave Kleikamp * and store the current valid flag. (It may have 5994ac27a0ecSDave Kleikamp * been changed by e2fsck since we originally mounted 5995ac27a0ecSDave Kleikamp * the partition.) 5996ac27a0ecSDave Kleikamp */ 599711215630SJan Kara if (sbi->s_journal) { 599811215630SJan Kara err = ext4_clear_journal_err(sb, es); 599911215630SJan Kara if (err) 600011215630SJan Kara goto restore_opts; 600111215630SJan Kara } 6002ac27a0ecSDave Kleikamp sbi->s_mount_state = le16_to_cpu(es->s_state); 6003c89128a0SJaegeuk Kim 6004c89128a0SJaegeuk Kim err = ext4_setup_super(sb, es, 0); 6005c89128a0SJaegeuk Kim if (err) 6006c89128a0SJaegeuk Kim goto restore_opts; 6007c89128a0SJaegeuk Kim 60081751e8a6SLinus Torvalds sb->s_flags &= ~SB_RDONLY; 6009e2b911c5SDarrick J. Wong if (ext4_has_feature_mmp(sb)) 6010c5e06d10SJohann Lombardi if (ext4_multi_mount_protect(sb, 6011c5e06d10SJohann Lombardi le64_to_cpu(es->s_mmp_block))) { 6012c5e06d10SJohann Lombardi err = -EROFS; 6013c5e06d10SJohann Lombardi goto restore_opts; 6014c5e06d10SJohann Lombardi } 6015c79d967dSChristoph Hellwig enable_quota = 1; 6016ac27a0ecSDave Kleikamp } 6017ac27a0ecSDave Kleikamp } 6018bfff6873SLukas Czerner 6019bfff6873SLukas Czerner /* 6020bfff6873SLukas Czerner * Reinitialize lazy itable initialization thread based on 6021bfff6873SLukas Czerner * current settings 6022bfff6873SLukas Czerner */ 6023bc98a42cSDavid Howells if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 6024bfff6873SLukas Czerner ext4_unregister_li_request(sb); 6025bfff6873SLukas Czerner else { 6026bfff6873SLukas Czerner ext4_group_t first_not_zeroed; 6027bfff6873SLukas Czerner first_not_zeroed = ext4_has_uninit_itable(sb); 6028bfff6873SLukas Czerner ext4_register_li_request(sb, first_not_zeroed); 6029bfff6873SLukas Czerner } 6030bfff6873SLukas Czerner 60310f5bde1dSJan Kara /* 60320f5bde1dSJan Kara * Handle creation of system zone data early because it can fail. 60330f5bde1dSJan Kara * Releasing of existing data is done when we are sure remount will 60340f5bde1dSJan Kara * succeed. 60350f5bde1dSJan Kara */ 6036dd0db94fSChunguang Xu if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) { 6037d176b1f6SJan Kara err = ext4_setup_system_zone(sb); 6038d176b1f6SJan Kara if (err) 6039d176b1f6SJan Kara goto restore_opts; 60400f5bde1dSJan Kara } 6041d176b1f6SJan Kara 6042c89128a0SJaegeuk Kim if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 60434392fbc4SJan Kara err = ext4_commit_super(sb); 6044c89128a0SJaegeuk Kim if (err) 6045c89128a0SJaegeuk Kim goto restore_opts; 6046c89128a0SJaegeuk Kim } 60470390131bSFrank Mayhar 6048ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 6049ac27a0ecSDave Kleikamp /* Release old quota file names */ 6050a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 6051ac27a0ecSDave Kleikamp kfree(old_opts.s_qf_names[i]); 60527c319d32SAditya Kali if (enable_quota) { 60537c319d32SAditya Kali if (sb_any_quota_suspended(sb)) 60540f0dd62fSChristoph Hellwig dquot_resume(sb, -1); 6055e2b911c5SDarrick J. Wong else if (ext4_has_feature_quota(sb)) { 60567c319d32SAditya Kali err = ext4_enable_quotas(sb); 605707724f98STheodore Ts'o if (err) 60587c319d32SAditya Kali goto restore_opts; 60597c319d32SAditya Kali } 60607c319d32SAditya Kali } 60617c319d32SAditya Kali #endif 6062dd0db94fSChunguang Xu if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 60630f5bde1dSJan Kara ext4_release_system_zone(sb); 6064d4c402d9SCurt Wohlgemuth 6065f25391ebSLukas Czerner /* 6066f25391ebSLukas Czerner * Some options can be enabled by ext4 and/or by VFS mount flag 6067f25391ebSLukas Czerner * either way we need to make sure it matches in both *flags and 6068f25391ebSLukas Czerner * s_flags. Copy those selected flags from s_flags to *flags 6069f25391ebSLukas Czerner */ 6070f25391ebSLukas Czerner *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags); 6071f25391ebSLukas Czerner 6072ca9b404fSRoman Anufriev ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s. Quota mode: %s.", 6073ca9b404fSRoman Anufriev orig_data, ext4_quota_mode(sb)); 6074d4c402d9SCurt Wohlgemuth kfree(orig_data); 6075ac27a0ecSDave Kleikamp return 0; 60760b8e58a1SAndreas Dilger 6077ac27a0ecSDave Kleikamp restore_opts: 6078ac27a0ecSDave Kleikamp sb->s_flags = old_sb_flags; 6079ac27a0ecSDave Kleikamp sbi->s_mount_opt = old_opts.s_mount_opt; 6080a2595b8aSTheodore Ts'o sbi->s_mount_opt2 = old_opts.s_mount_opt2; 6081ac27a0ecSDave Kleikamp sbi->s_resuid = old_opts.s_resuid; 6082ac27a0ecSDave Kleikamp sbi->s_resgid = old_opts.s_resgid; 6083ac27a0ecSDave Kleikamp sbi->s_commit_interval = old_opts.s_commit_interval; 608430773840STheodore Ts'o sbi->s_min_batch_time = old_opts.s_min_batch_time; 608530773840STheodore Ts'o sbi->s_max_batch_time = old_opts.s_max_batch_time; 6086dd0db94fSChunguang Xu if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 60870f5bde1dSJan Kara ext4_release_system_zone(sb); 6088ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 6089ac27a0ecSDave Kleikamp sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 6090a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 609133458eabSTheodore Ts'o to_free[i] = get_qf_name(sb, sbi, i); 609233458eabSTheodore Ts'o rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 6093ac27a0ecSDave Kleikamp } 609433458eabSTheodore Ts'o synchronize_rcu(); 609533458eabSTheodore Ts'o for (i = 0; i < EXT4_MAXQUOTAS; i++) 609633458eabSTheodore Ts'o kfree(to_free[i]); 6097ac27a0ecSDave Kleikamp #endif 6098d4c402d9SCurt Wohlgemuth kfree(orig_data); 6099ac27a0ecSDave Kleikamp return err; 6100ac27a0ecSDave Kleikamp } 6101ac27a0ecSDave Kleikamp 6102689c958cSLi Xi #ifdef CONFIG_QUOTA 6103689c958cSLi Xi static int ext4_statfs_project(struct super_block *sb, 6104689c958cSLi Xi kprojid_t projid, struct kstatfs *buf) 6105689c958cSLi Xi { 6106689c958cSLi Xi struct kqid qid; 6107689c958cSLi Xi struct dquot *dquot; 6108689c958cSLi Xi u64 limit; 6109689c958cSLi Xi u64 curblock; 6110689c958cSLi Xi 6111689c958cSLi Xi qid = make_kqid_projid(projid); 6112689c958cSLi Xi dquot = dqget(sb, qid); 6113689c958cSLi Xi if (IS_ERR(dquot)) 6114689c958cSLi Xi return PTR_ERR(dquot); 61157b9ca4c6SJan Kara spin_lock(&dquot->dq_dqb_lock); 6116689c958cSLi Xi 6117a08fe66eSChengguang Xu limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 6118a08fe66eSChengguang Xu dquot->dq_dqb.dqb_bhardlimit); 611957c32ea4SChengguang Xu limit >>= sb->s_blocksize_bits; 612057c32ea4SChengguang Xu 6121689c958cSLi Xi if (limit && buf->f_blocks > limit) { 6122f06925c7SKonstantin Khlebnikov curblock = (dquot->dq_dqb.dqb_curspace + 6123f06925c7SKonstantin Khlebnikov dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 6124689c958cSLi Xi buf->f_blocks = limit; 6125689c958cSLi Xi buf->f_bfree = buf->f_bavail = 6126689c958cSLi Xi (buf->f_blocks > curblock) ? 6127689c958cSLi Xi (buf->f_blocks - curblock) : 0; 6128689c958cSLi Xi } 6129689c958cSLi Xi 6130a08fe66eSChengguang Xu limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 6131a08fe66eSChengguang Xu dquot->dq_dqb.dqb_ihardlimit); 6132689c958cSLi Xi if (limit && buf->f_files > limit) { 6133689c958cSLi Xi buf->f_files = limit; 6134689c958cSLi Xi buf->f_ffree = 6135689c958cSLi Xi (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 6136689c958cSLi Xi (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 6137689c958cSLi Xi } 6138689c958cSLi Xi 61397b9ca4c6SJan Kara spin_unlock(&dquot->dq_dqb_lock); 6140689c958cSLi Xi dqput(dquot); 6141689c958cSLi Xi return 0; 6142689c958cSLi Xi } 6143689c958cSLi Xi #endif 6144689c958cSLi Xi 6145617ba13bSMingming Cao static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 6146ac27a0ecSDave Kleikamp { 6147ac27a0ecSDave Kleikamp struct super_block *sb = dentry->d_sb; 6148617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 6149617ba13bSMingming Cao struct ext4_super_block *es = sbi->s_es; 615027dd4385SLukas Czerner ext4_fsblk_t overhead = 0, resv_blocks; 6151d02a9391SKazuya Mio s64 bfree; 615227dd4385SLukas Czerner resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 6153ac27a0ecSDave Kleikamp 6154952fc18eSTheodore Ts'o if (!test_opt(sb, MINIX_DF)) 6155952fc18eSTheodore Ts'o overhead = sbi->s_overhead; 6156ac27a0ecSDave Kleikamp 6157617ba13bSMingming Cao buf->f_type = EXT4_SUPER_MAGIC; 6158ac27a0ecSDave Kleikamp buf->f_bsize = sb->s_blocksize; 6159b72f78cbSEric Sandeen buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 616057042651STheodore Ts'o bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 616157042651STheodore Ts'o percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 6162d02a9391SKazuya Mio /* prevent underflow in case that few free space is available */ 616357042651STheodore Ts'o buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 616427dd4385SLukas Czerner buf->f_bavail = buf->f_bfree - 616527dd4385SLukas Czerner (ext4_r_blocks_count(es) + resv_blocks); 616627dd4385SLukas Czerner if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 6167ac27a0ecSDave Kleikamp buf->f_bavail = 0; 6168ac27a0ecSDave Kleikamp buf->f_files = le32_to_cpu(es->s_inodes_count); 616952d9f3b4SPeter Zijlstra buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 6170617ba13bSMingming Cao buf->f_namelen = EXT4_NAME_LEN; 6171*9591c3a3SAmir Goldstein buf->f_fsid = uuid_to_fsid(es->s_uuid); 61720b8e58a1SAndreas Dilger 6173689c958cSLi Xi #ifdef CONFIG_QUOTA 6174689c958cSLi Xi if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 6175689c958cSLi Xi sb_has_quota_limits_enabled(sb, PRJQUOTA)) 6176689c958cSLi Xi ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 6177689c958cSLi Xi #endif 6178ac27a0ecSDave Kleikamp return 0; 6179ac27a0ecSDave Kleikamp } 6180ac27a0ecSDave Kleikamp 6181ac27a0ecSDave Kleikamp 6182ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 6183ac27a0ecSDave Kleikamp 6184bc8230eeSJan Kara /* 6185bc8230eeSJan Kara * Helper functions so that transaction is started before we acquire dqio_sem 6186bc8230eeSJan Kara * to keep correct lock ordering of transaction > dqio_sem 6187bc8230eeSJan Kara */ 6188ac27a0ecSDave Kleikamp static inline struct inode *dquot_to_inode(struct dquot *dquot) 6189ac27a0ecSDave Kleikamp { 61904c376dcaSEric W. Biederman return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 6191ac27a0ecSDave Kleikamp } 6192ac27a0ecSDave Kleikamp 6193617ba13bSMingming Cao static int ext4_write_dquot(struct dquot *dquot) 6194ac27a0ecSDave Kleikamp { 6195ac27a0ecSDave Kleikamp int ret, err; 6196ac27a0ecSDave Kleikamp handle_t *handle; 6197ac27a0ecSDave Kleikamp struct inode *inode; 6198ac27a0ecSDave Kleikamp 6199ac27a0ecSDave Kleikamp inode = dquot_to_inode(dquot); 62009924a92aSTheodore Ts'o handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 6201617ba13bSMingming Cao EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 6202ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 6203ac27a0ecSDave Kleikamp return PTR_ERR(handle); 6204ac27a0ecSDave Kleikamp ret = dquot_commit(dquot); 6205617ba13bSMingming Cao err = ext4_journal_stop(handle); 6206ac27a0ecSDave Kleikamp if (!ret) 6207ac27a0ecSDave Kleikamp ret = err; 6208ac27a0ecSDave Kleikamp return ret; 6209ac27a0ecSDave Kleikamp } 6210ac27a0ecSDave Kleikamp 6211617ba13bSMingming Cao static int ext4_acquire_dquot(struct dquot *dquot) 6212ac27a0ecSDave Kleikamp { 6213ac27a0ecSDave Kleikamp int ret, err; 6214ac27a0ecSDave Kleikamp handle_t *handle; 6215ac27a0ecSDave Kleikamp 62169924a92aSTheodore Ts'o handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 6217617ba13bSMingming Cao EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 6218ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 6219ac27a0ecSDave Kleikamp return PTR_ERR(handle); 6220ac27a0ecSDave Kleikamp ret = dquot_acquire(dquot); 6221617ba13bSMingming Cao err = ext4_journal_stop(handle); 6222ac27a0ecSDave Kleikamp if (!ret) 6223ac27a0ecSDave Kleikamp ret = err; 6224ac27a0ecSDave Kleikamp return ret; 6225ac27a0ecSDave Kleikamp } 6226ac27a0ecSDave Kleikamp 6227617ba13bSMingming Cao static int ext4_release_dquot(struct dquot *dquot) 6228ac27a0ecSDave Kleikamp { 6229ac27a0ecSDave Kleikamp int ret, err; 6230ac27a0ecSDave Kleikamp handle_t *handle; 6231ac27a0ecSDave Kleikamp 62329924a92aSTheodore Ts'o handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 6233617ba13bSMingming Cao EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 62349c3013e9SJan Kara if (IS_ERR(handle)) { 62359c3013e9SJan Kara /* Release dquot anyway to avoid endless cycle in dqput() */ 62369c3013e9SJan Kara dquot_release(dquot); 6237ac27a0ecSDave Kleikamp return PTR_ERR(handle); 62389c3013e9SJan Kara } 6239ac27a0ecSDave Kleikamp ret = dquot_release(dquot); 6240617ba13bSMingming Cao err = ext4_journal_stop(handle); 6241ac27a0ecSDave Kleikamp if (!ret) 6242ac27a0ecSDave Kleikamp ret = err; 6243ac27a0ecSDave Kleikamp return ret; 6244ac27a0ecSDave Kleikamp } 6245ac27a0ecSDave Kleikamp 6246617ba13bSMingming Cao static int ext4_mark_dquot_dirty(struct dquot *dquot) 6247ac27a0ecSDave Kleikamp { 6248262b4662SJan Kara struct super_block *sb = dquot->dq_sb; 6249262b4662SJan Kara 6250f177ee08SRoman Anufriev if (ext4_is_quota_journalled(sb)) { 6251ac27a0ecSDave Kleikamp dquot_mark_dquot_dirty(dquot); 6252617ba13bSMingming Cao return ext4_write_dquot(dquot); 6253ac27a0ecSDave Kleikamp } else { 6254ac27a0ecSDave Kleikamp return dquot_mark_dquot_dirty(dquot); 6255ac27a0ecSDave Kleikamp } 6256ac27a0ecSDave Kleikamp } 6257ac27a0ecSDave Kleikamp 6258617ba13bSMingming Cao static int ext4_write_info(struct super_block *sb, int type) 6259ac27a0ecSDave Kleikamp { 6260ac27a0ecSDave Kleikamp int ret, err; 6261ac27a0ecSDave Kleikamp handle_t *handle; 6262ac27a0ecSDave Kleikamp 6263ac27a0ecSDave Kleikamp /* Data block + inode block */ 62642b0143b5SDavid Howells handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); 6265ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 6266ac27a0ecSDave Kleikamp return PTR_ERR(handle); 6267ac27a0ecSDave Kleikamp ret = dquot_commit_info(sb, type); 6268617ba13bSMingming Cao err = ext4_journal_stop(handle); 6269ac27a0ecSDave Kleikamp if (!ret) 6270ac27a0ecSDave Kleikamp ret = err; 6271ac27a0ecSDave Kleikamp return ret; 6272ac27a0ecSDave Kleikamp } 6273ac27a0ecSDave Kleikamp 6274ac27a0ecSDave Kleikamp /* 6275ac27a0ecSDave Kleikamp * Turn on quotas during mount time - we need to find 6276ac27a0ecSDave Kleikamp * the quota file and such... 6277ac27a0ecSDave Kleikamp */ 6278617ba13bSMingming Cao static int ext4_quota_on_mount(struct super_block *sb, int type) 6279ac27a0ecSDave Kleikamp { 628033458eabSTheodore Ts'o return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type), 6281617ba13bSMingming Cao EXT4_SB(sb)->s_jquota_fmt, type); 6282ac27a0ecSDave Kleikamp } 6283ac27a0ecSDave Kleikamp 6284daf647d2STheodore Ts'o static void lockdep_set_quota_inode(struct inode *inode, int subclass) 6285daf647d2STheodore Ts'o { 6286daf647d2STheodore Ts'o struct ext4_inode_info *ei = EXT4_I(inode); 6287daf647d2STheodore Ts'o 6288daf647d2STheodore Ts'o /* The first argument of lockdep_set_subclass has to be 6289daf647d2STheodore Ts'o * *exactly* the same as the argument to init_rwsem() --- in 6290daf647d2STheodore Ts'o * this case, in init_once() --- or lockdep gets unhappy 6291daf647d2STheodore Ts'o * because the name of the lock is set using the 6292daf647d2STheodore Ts'o * stringification of the argument to init_rwsem(). 6293daf647d2STheodore Ts'o */ 6294daf647d2STheodore Ts'o (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 6295daf647d2STheodore Ts'o lockdep_set_subclass(&ei->i_data_sem, subclass); 6296daf647d2STheodore Ts'o } 6297daf647d2STheodore Ts'o 6298ac27a0ecSDave Kleikamp /* 6299ac27a0ecSDave Kleikamp * Standard function to be called on quota_on 6300ac27a0ecSDave Kleikamp */ 6301617ba13bSMingming Cao static int ext4_quota_on(struct super_block *sb, int type, int format_id, 63028c54ca9cSAl Viro const struct path *path) 6303ac27a0ecSDave Kleikamp { 6304ac27a0ecSDave Kleikamp int err; 6305ac27a0ecSDave Kleikamp 6306ac27a0ecSDave Kleikamp if (!test_opt(sb, QUOTA)) 6307ac27a0ecSDave Kleikamp return -EINVAL; 63080623543bSJan Kara 6309ac27a0ecSDave Kleikamp /* Quotafile not on the same filesystem? */ 6310d8c9584eSAl Viro if (path->dentry->d_sb != sb) 6311ac27a0ecSDave Kleikamp return -EXDEV; 6312e0770e91SJan Kara 6313e0770e91SJan Kara /* Quota already enabled for this file? */ 6314e0770e91SJan Kara if (IS_NOQUOTA(d_inode(path->dentry))) 6315e0770e91SJan Kara return -EBUSY; 6316e0770e91SJan Kara 63170623543bSJan Kara /* Journaling quota? */ 63180623543bSJan Kara if (EXT4_SB(sb)->s_qf_names[type]) { 63192b2d6d01STheodore Ts'o /* Quotafile not in fs root? */ 6320f00c9e44SJan Kara if (path->dentry->d_parent != sb->s_root) 6321b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 6322b31e1552SEric Sandeen "Quota file not on filesystem root. " 6323b31e1552SEric Sandeen "Journaled quota will not work"); 632491389240SJan Kara sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 632591389240SJan Kara } else { 632691389240SJan Kara /* 632791389240SJan Kara * Clear the flag just in case mount options changed since 632891389240SJan Kara * last time. 632991389240SJan Kara */ 633091389240SJan Kara sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 63310623543bSJan Kara } 63320623543bSJan Kara 63330623543bSJan Kara /* 63340623543bSJan Kara * When we journal data on quota file, we have to flush journal to see 63350623543bSJan Kara * all updates to the file when we bypass pagecache... 63360623543bSJan Kara */ 63370390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal && 63382b0143b5SDavid Howells ext4_should_journal_data(d_inode(path->dentry))) { 63390623543bSJan Kara /* 63400623543bSJan Kara * We don't need to lock updates but journal_flush() could 63410623543bSJan Kara * otherwise be livelocked... 63420623543bSJan Kara */ 63430623543bSJan Kara jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 63447ffe1ea8SHidehiro Kawai err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 63450623543bSJan Kara jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 6346f00c9e44SJan Kara if (err) 63477ffe1ea8SHidehiro Kawai return err; 63487ffe1ea8SHidehiro Kawai } 6349957153fcSJan Kara 6350daf647d2STheodore Ts'o lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 6351daf647d2STheodore Ts'o err = dquot_quota_on(sb, type, format_id, path); 6352957153fcSJan Kara if (err) { 6353daf647d2STheodore Ts'o lockdep_set_quota_inode(path->dentry->d_inode, 6354daf647d2STheodore Ts'o I_DATA_SEM_NORMAL); 6355957153fcSJan Kara } else { 6356957153fcSJan Kara struct inode *inode = d_inode(path->dentry); 6357957153fcSJan Kara handle_t *handle; 6358957153fcSJan Kara 635961a92987SJan Kara /* 636061a92987SJan Kara * Set inode flags to prevent userspace from messing with quota 636161a92987SJan Kara * files. If this fails, we return success anyway since quotas 636261a92987SJan Kara * are already enabled and this is not a hard failure. 636361a92987SJan Kara */ 6364957153fcSJan Kara inode_lock(inode); 6365957153fcSJan Kara handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 6366957153fcSJan Kara if (IS_ERR(handle)) 6367957153fcSJan Kara goto unlock_inode; 6368957153fcSJan Kara EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 6369957153fcSJan Kara inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 6370957153fcSJan Kara S_NOATIME | S_IMMUTABLE); 63714209ae12SHarshad Shirwadkar err = ext4_mark_inode_dirty(handle, inode); 6372957153fcSJan Kara ext4_journal_stop(handle); 6373957153fcSJan Kara unlock_inode: 6374957153fcSJan Kara inode_unlock(inode); 6375957153fcSJan Kara } 6376daf647d2STheodore Ts'o return err; 6377ac27a0ecSDave Kleikamp } 6378ac27a0ecSDave Kleikamp 63797c319d32SAditya Kali static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 63807c319d32SAditya Kali unsigned int flags) 63817c319d32SAditya Kali { 63827c319d32SAditya Kali int err; 63837c319d32SAditya Kali struct inode *qf_inode; 6384a2d4a646SJan Kara unsigned long qf_inums[EXT4_MAXQUOTAS] = { 63857c319d32SAditya Kali le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 6386689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 6387689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 63887c319d32SAditya Kali }; 63897c319d32SAditya Kali 6390e2b911c5SDarrick J. Wong BUG_ON(!ext4_has_feature_quota(sb)); 63917c319d32SAditya Kali 63927c319d32SAditya Kali if (!qf_inums[type]) 63937c319d32SAditya Kali return -EPERM; 63947c319d32SAditya Kali 63958a363970STheodore Ts'o qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 63967c319d32SAditya Kali if (IS_ERR(qf_inode)) { 63977c319d32SAditya Kali ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 63987c319d32SAditya Kali return PTR_ERR(qf_inode); 63997c319d32SAditya Kali } 64007c319d32SAditya Kali 6401bcb13850SJan Kara /* Don't account quota for quota files to avoid recursion */ 6402bcb13850SJan Kara qf_inode->i_flags |= S_NOQUOTA; 6403daf647d2STheodore Ts'o lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 64047212b95eSJan Kara err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 6405daf647d2STheodore Ts'o if (err) 6406daf647d2STheodore Ts'o lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 640761157b24SPan Bian iput(qf_inode); 64087c319d32SAditya Kali 64097c319d32SAditya Kali return err; 64107c319d32SAditya Kali } 64117c319d32SAditya Kali 64127c319d32SAditya Kali /* Enable usage tracking for all quota types. */ 64137c319d32SAditya Kali static int ext4_enable_quotas(struct super_block *sb) 64147c319d32SAditya Kali { 64157c319d32SAditya Kali int type, err = 0; 6416a2d4a646SJan Kara unsigned long qf_inums[EXT4_MAXQUOTAS] = { 64177c319d32SAditya Kali le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 6418689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 6419689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 64207c319d32SAditya Kali }; 642149da9392SJan Kara bool quota_mopt[EXT4_MAXQUOTAS] = { 642249da9392SJan Kara test_opt(sb, USRQUOTA), 642349da9392SJan Kara test_opt(sb, GRPQUOTA), 642449da9392SJan Kara test_opt(sb, PRJQUOTA), 642549da9392SJan Kara }; 64267c319d32SAditya Kali 642791389240SJan Kara sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 6428a2d4a646SJan Kara for (type = 0; type < EXT4_MAXQUOTAS; type++) { 64297c319d32SAditya Kali if (qf_inums[type]) { 64307c319d32SAditya Kali err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 643149da9392SJan Kara DQUOT_USAGE_ENABLED | 643249da9392SJan Kara (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 64337c319d32SAditya Kali if (err) { 64347c319d32SAditya Kali ext4_warning(sb, 643572ba7450STheodore Ts'o "Failed to enable quota tracking " 643672ba7450STheodore Ts'o "(type=%d, err=%d). Please run " 643772ba7450STheodore Ts'o "e2fsck to fix.", type, err); 64387f144fd0SJunichi Uekawa for (type--; type >= 0; type--) 64397f144fd0SJunichi Uekawa dquot_quota_off(sb, type); 64407f144fd0SJunichi Uekawa 64417c319d32SAditya Kali return err; 64427c319d32SAditya Kali } 64437c319d32SAditya Kali } 64447c319d32SAditya Kali } 64457c319d32SAditya Kali return 0; 64467c319d32SAditya Kali } 64477c319d32SAditya Kali 6448ca0e05e4SDmitry Monakhov static int ext4_quota_off(struct super_block *sb, int type) 6449ca0e05e4SDmitry Monakhov { 645021f97697SJan Kara struct inode *inode = sb_dqopt(sb)->files[type]; 645121f97697SJan Kara handle_t *handle; 6452957153fcSJan Kara int err; 645321f97697SJan Kara 645487009d86SDmitry Monakhov /* Force all delayed allocation blocks to be allocated. 645587009d86SDmitry Monakhov * Caller already holds s_umount sem */ 645687009d86SDmitry Monakhov if (test_opt(sb, DELALLOC)) 6457ca0e05e4SDmitry Monakhov sync_filesystem(sb); 6458ca0e05e4SDmitry Monakhov 6459957153fcSJan Kara if (!inode || !igrab(inode)) 64600b268590SAmir Goldstein goto out; 64610b268590SAmir Goldstein 6462957153fcSJan Kara err = dquot_quota_off(sb, type); 6463964edf66SJan Kara if (err || ext4_has_feature_quota(sb)) 6464957153fcSJan Kara goto out_put; 6465957153fcSJan Kara 6466957153fcSJan Kara inode_lock(inode); 646761a92987SJan Kara /* 646861a92987SJan Kara * Update modification times of quota files when userspace can 646961a92987SJan Kara * start looking at them. If we fail, we return success anyway since 647061a92987SJan Kara * this is not a hard failure and quotas are already disabled. 647161a92987SJan Kara */ 64729924a92aSTheodore Ts'o handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 64734209ae12SHarshad Shirwadkar if (IS_ERR(handle)) { 64744209ae12SHarshad Shirwadkar err = PTR_ERR(handle); 6475957153fcSJan Kara goto out_unlock; 64764209ae12SHarshad Shirwadkar } 6477957153fcSJan Kara EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 6478957153fcSJan Kara inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 6479eeca7ea1SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 64804209ae12SHarshad Shirwadkar err = ext4_mark_inode_dirty(handle, inode); 648121f97697SJan Kara ext4_journal_stop(handle); 6482957153fcSJan Kara out_unlock: 6483957153fcSJan Kara inode_unlock(inode); 6484957153fcSJan Kara out_put: 6485964edf66SJan Kara lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 6486957153fcSJan Kara iput(inode); 6487957153fcSJan Kara return err; 648821f97697SJan Kara out: 6489ca0e05e4SDmitry Monakhov return dquot_quota_off(sb, type); 6490ca0e05e4SDmitry Monakhov } 6491ca0e05e4SDmitry Monakhov 6492ac27a0ecSDave Kleikamp /* Read data from quotafile - avoid pagecache and such because we cannot afford 6493ac27a0ecSDave Kleikamp * acquiring the locks... As quota files are never truncated and quota code 6494ac27a0ecSDave Kleikamp * itself serializes the operations (and no one else should touch the files) 6495ac27a0ecSDave Kleikamp * we don't have to be afraid of races */ 6496617ba13bSMingming Cao static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 6497ac27a0ecSDave Kleikamp size_t len, loff_t off) 6498ac27a0ecSDave Kleikamp { 6499ac27a0ecSDave Kleikamp struct inode *inode = sb_dqopt(sb)->files[type]; 6500725d26d3SAneesh Kumar K.V ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 6501ac27a0ecSDave Kleikamp int offset = off & (sb->s_blocksize - 1); 6502ac27a0ecSDave Kleikamp int tocopy; 6503ac27a0ecSDave Kleikamp size_t toread; 6504ac27a0ecSDave Kleikamp struct buffer_head *bh; 6505ac27a0ecSDave Kleikamp loff_t i_size = i_size_read(inode); 6506ac27a0ecSDave Kleikamp 6507ac27a0ecSDave Kleikamp if (off > i_size) 6508ac27a0ecSDave Kleikamp return 0; 6509ac27a0ecSDave Kleikamp if (off+len > i_size) 6510ac27a0ecSDave Kleikamp len = i_size-off; 6511ac27a0ecSDave Kleikamp toread = len; 6512ac27a0ecSDave Kleikamp while (toread > 0) { 6513ac27a0ecSDave Kleikamp tocopy = sb->s_blocksize - offset < toread ? 6514ac27a0ecSDave Kleikamp sb->s_blocksize - offset : toread; 65151c215028STheodore Ts'o bh = ext4_bread(NULL, inode, blk, 0); 65161c215028STheodore Ts'o if (IS_ERR(bh)) 65171c215028STheodore Ts'o return PTR_ERR(bh); 6518ac27a0ecSDave Kleikamp if (!bh) /* A hole? */ 6519ac27a0ecSDave Kleikamp memset(data, 0, tocopy); 6520ac27a0ecSDave Kleikamp else 6521ac27a0ecSDave Kleikamp memcpy(data, bh->b_data+offset, tocopy); 6522ac27a0ecSDave Kleikamp brelse(bh); 6523ac27a0ecSDave Kleikamp offset = 0; 6524ac27a0ecSDave Kleikamp toread -= tocopy; 6525ac27a0ecSDave Kleikamp data += tocopy; 6526ac27a0ecSDave Kleikamp blk++; 6527ac27a0ecSDave Kleikamp } 6528ac27a0ecSDave Kleikamp return len; 6529ac27a0ecSDave Kleikamp } 6530ac27a0ecSDave Kleikamp 6531ac27a0ecSDave Kleikamp /* Write to quotafile (we know the transaction is already started and has 6532ac27a0ecSDave Kleikamp * enough credits) */ 6533617ba13bSMingming Cao static ssize_t ext4_quota_write(struct super_block *sb, int type, 6534ac27a0ecSDave Kleikamp const char *data, size_t len, loff_t off) 6535ac27a0ecSDave Kleikamp { 6536ac27a0ecSDave Kleikamp struct inode *inode = sb_dqopt(sb)->files[type]; 6537725d26d3SAneesh Kumar K.V ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 65384209ae12SHarshad Shirwadkar int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1); 6539c5e298aeSTheodore Ts'o int retries = 0; 6540ac27a0ecSDave Kleikamp struct buffer_head *bh; 6541ac27a0ecSDave Kleikamp handle_t *handle = journal_current_handle(); 6542ac27a0ecSDave Kleikamp 65430390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal && !handle) { 6544b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 6545b31e1552SEric Sandeen " cancelled because transaction is not started", 65469c3013e9SJan Kara (unsigned long long)off, (unsigned long long)len); 65479c3013e9SJan Kara return -EIO; 65489c3013e9SJan Kara } 654967eeb568SDmitry Monakhov /* 655067eeb568SDmitry Monakhov * Since we account only one data block in transaction credits, 655167eeb568SDmitry Monakhov * then it is impossible to cross a block boundary. 655267eeb568SDmitry Monakhov */ 655367eeb568SDmitry Monakhov if (sb->s_blocksize - offset < len) { 655467eeb568SDmitry Monakhov ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 655567eeb568SDmitry Monakhov " cancelled because not block aligned", 655667eeb568SDmitry Monakhov (unsigned long long)off, (unsigned long long)len); 655767eeb568SDmitry Monakhov return -EIO; 655867eeb568SDmitry Monakhov } 655967eeb568SDmitry Monakhov 6560c5e298aeSTheodore Ts'o do { 6561c5e298aeSTheodore Ts'o bh = ext4_bread(handle, inode, blk, 6562c5e298aeSTheodore Ts'o EXT4_GET_BLOCKS_CREATE | 6563c5e298aeSTheodore Ts'o EXT4_GET_BLOCKS_METADATA_NOFAIL); 656445586c70SMasahiro Yamada } while (PTR_ERR(bh) == -ENOSPC && 6565c5e298aeSTheodore Ts'o ext4_should_retry_alloc(inode->i_sb, &retries)); 65661c215028STheodore Ts'o if (IS_ERR(bh)) 65671c215028STheodore Ts'o return PTR_ERR(bh); 6568ac27a0ecSDave Kleikamp if (!bh) 6569ac27a0ecSDave Kleikamp goto out; 65705d601255Sliang xie BUFFER_TRACE(bh, "get write access"); 6571617ba13bSMingming Cao err = ext4_journal_get_write_access(handle, bh); 6572ac27a0ecSDave Kleikamp if (err) { 6573ac27a0ecSDave Kleikamp brelse(bh); 65741c215028STheodore Ts'o return err; 6575ac27a0ecSDave Kleikamp } 6576ac27a0ecSDave Kleikamp lock_buffer(bh); 657767eeb568SDmitry Monakhov memcpy(bh->b_data+offset, data, len); 6578ac27a0ecSDave Kleikamp flush_dcache_page(bh->b_page); 6579ac27a0ecSDave Kleikamp unlock_buffer(bh); 65800390131bSFrank Mayhar err = ext4_handle_dirty_metadata(handle, NULL, bh); 6581ac27a0ecSDave Kleikamp brelse(bh); 6582ac27a0ecSDave Kleikamp out: 658367eeb568SDmitry Monakhov if (inode->i_size < off + len) { 658467eeb568SDmitry Monakhov i_size_write(inode, off + len); 6585617ba13bSMingming Cao EXT4_I(inode)->i_disksize = inode->i_size; 65864209ae12SHarshad Shirwadkar err2 = ext4_mark_inode_dirty(handle, inode); 65874209ae12SHarshad Shirwadkar if (unlikely(err2 && !err)) 65884209ae12SHarshad Shirwadkar err = err2; 658921f97697SJan Kara } 65904209ae12SHarshad Shirwadkar return err ? err : len; 6591ac27a0ecSDave Kleikamp } 6592ac27a0ecSDave Kleikamp #endif 6593ac27a0ecSDave Kleikamp 6594152a0836SAl Viro static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 6595152a0836SAl Viro const char *dev_name, void *data) 6596ac27a0ecSDave Kleikamp { 6597152a0836SAl Viro return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 6598ac27a0ecSDave Kleikamp } 6599ac27a0ecSDave Kleikamp 6600c290ea01SJan Kara #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 660124b58424STheodore Ts'o static inline void register_as_ext2(void) 660224b58424STheodore Ts'o { 660324b58424STheodore Ts'o int err = register_filesystem(&ext2_fs_type); 660424b58424STheodore Ts'o if (err) 660524b58424STheodore Ts'o printk(KERN_WARNING 660624b58424STheodore Ts'o "EXT4-fs: Unable to register as ext2 (%d)\n", err); 660724b58424STheodore Ts'o } 660824b58424STheodore Ts'o 660924b58424STheodore Ts'o static inline void unregister_as_ext2(void) 661024b58424STheodore Ts'o { 661124b58424STheodore Ts'o unregister_filesystem(&ext2_fs_type); 661224b58424STheodore Ts'o } 66132035e776STheodore Ts'o 66142035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb) 66152035e776STheodore Ts'o { 6616e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext2_incompat_features(sb)) 66172035e776STheodore Ts'o return 0; 6618bc98a42cSDavid Howells if (sb_rdonly(sb)) 66192035e776STheodore Ts'o return 1; 6620e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext2_ro_compat_features(sb)) 66212035e776STheodore Ts'o return 0; 66222035e776STheodore Ts'o return 1; 66232035e776STheodore Ts'o } 662424b58424STheodore Ts'o #else 662524b58424STheodore Ts'o static inline void register_as_ext2(void) { } 662624b58424STheodore Ts'o static inline void unregister_as_ext2(void) { } 66272035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 662824b58424STheodore Ts'o #endif 662924b58424STheodore Ts'o 663024b58424STheodore Ts'o static inline void register_as_ext3(void) 663124b58424STheodore Ts'o { 663224b58424STheodore Ts'o int err = register_filesystem(&ext3_fs_type); 663324b58424STheodore Ts'o if (err) 663424b58424STheodore Ts'o printk(KERN_WARNING 663524b58424STheodore Ts'o "EXT4-fs: Unable to register as ext3 (%d)\n", err); 663624b58424STheodore Ts'o } 663724b58424STheodore Ts'o 663824b58424STheodore Ts'o static inline void unregister_as_ext3(void) 663924b58424STheodore Ts'o { 664024b58424STheodore Ts'o unregister_filesystem(&ext3_fs_type); 664124b58424STheodore Ts'o } 66422035e776STheodore Ts'o 66432035e776STheodore Ts'o static inline int ext3_feature_set_ok(struct super_block *sb) 66442035e776STheodore Ts'o { 6645e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext3_incompat_features(sb)) 66462035e776STheodore Ts'o return 0; 6647e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal(sb)) 66482035e776STheodore Ts'o return 0; 6649bc98a42cSDavid Howells if (sb_rdonly(sb)) 66502035e776STheodore Ts'o return 1; 6651e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext3_ro_compat_features(sb)) 66522035e776STheodore Ts'o return 0; 66532035e776STheodore Ts'o return 1; 66542035e776STheodore Ts'o } 665524b58424STheodore Ts'o 665603010a33STheodore Ts'o static struct file_system_type ext4_fs_type = { 6657ac27a0ecSDave Kleikamp .owner = THIS_MODULE, 665803010a33STheodore Ts'o .name = "ext4", 6659152a0836SAl Viro .mount = ext4_mount, 6660ac27a0ecSDave Kleikamp .kill_sb = kill_block_super, 666114f3db55SChristian Brauner .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, 6662ac27a0ecSDave Kleikamp }; 66637f78e035SEric W. Biederman MODULE_ALIAS_FS("ext4"); 6664ac27a0ecSDave Kleikamp 6665e9e3bcecSEric Sandeen /* Shared across all ext4 file systems */ 6666e9e3bcecSEric Sandeen wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 6667e9e3bcecSEric Sandeen 66685dabfc78STheodore Ts'o static int __init ext4_init_fs(void) 6669ac27a0ecSDave Kleikamp { 6670e9e3bcecSEric Sandeen int i, err; 6671c9de560dSAlex Tomas 6672e294a537STheodore Ts'o ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 667307c0c5d8SAl Viro ext4_li_info = NULL; 667407c0c5d8SAl Viro 66759a4c8019SCarlos Maiolino /* Build-time check for flags consistency */ 667612e9b892SDmitry Monakhov ext4_check_flag_values(); 6677e9e3bcecSEric Sandeen 6678e142d052SJan Kara for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 6679e9e3bcecSEric Sandeen init_waitqueue_head(&ext4__ioend_wq[i]); 6680e9e3bcecSEric Sandeen 668151865fdaSZheng Liu err = ext4_init_es(); 66826fd058f7STheodore Ts'o if (err) 66836fd058f7STheodore Ts'o return err; 668451865fdaSZheng Liu 66851dc0aa46SEric Whitney err = ext4_init_pending(); 66861dc0aa46SEric Whitney if (err) 668722cfe4b4SEric Biggers goto out7; 668822cfe4b4SEric Biggers 668922cfe4b4SEric Biggers err = ext4_init_post_read_processing(); 669022cfe4b4SEric Biggers if (err) 66911dc0aa46SEric Whitney goto out6; 66921dc0aa46SEric Whitney 669351865fdaSZheng Liu err = ext4_init_pageio(); 669451865fdaSZheng Liu if (err) 6695b5799018STheodore Ts'o goto out5; 669651865fdaSZheng Liu 66975dabfc78STheodore Ts'o err = ext4_init_system_zone(); 6698bd2d0210STheodore Ts'o if (err) 6699b5799018STheodore Ts'o goto out4; 6700857ac889SLukas Czerner 6701b5799018STheodore Ts'o err = ext4_init_sysfs(); 6702dd68314cSTheodore Ts'o if (err) 6703b5799018STheodore Ts'o goto out3; 6704857ac889SLukas Czerner 67055dabfc78STheodore Ts'o err = ext4_init_mballoc(); 6706ac27a0ecSDave Kleikamp if (err) 6707c9de560dSAlex Tomas goto out2; 6708ac27a0ecSDave Kleikamp err = init_inodecache(); 6709ac27a0ecSDave Kleikamp if (err) 6710ac27a0ecSDave Kleikamp goto out1; 6711aa75f4d3SHarshad Shirwadkar 6712aa75f4d3SHarshad Shirwadkar err = ext4_fc_init_dentry_cache(); 6713aa75f4d3SHarshad Shirwadkar if (err) 6714aa75f4d3SHarshad Shirwadkar goto out05; 6715aa75f4d3SHarshad Shirwadkar 671624b58424STheodore Ts'o register_as_ext3(); 67172035e776STheodore Ts'o register_as_ext2(); 671803010a33STheodore Ts'o err = register_filesystem(&ext4_fs_type); 6719ac27a0ecSDave Kleikamp if (err) 6720ac27a0ecSDave Kleikamp goto out; 6721bfff6873SLukas Czerner 6722ac27a0ecSDave Kleikamp return 0; 6723ac27a0ecSDave Kleikamp out: 672424b58424STheodore Ts'o unregister_as_ext2(); 672524b58424STheodore Ts'o unregister_as_ext3(); 6726aa75f4d3SHarshad Shirwadkar out05: 6727ac27a0ecSDave Kleikamp destroy_inodecache(); 6728ac27a0ecSDave Kleikamp out1: 67295dabfc78STheodore Ts'o ext4_exit_mballoc(); 67309c191f70ST Makphaibulchoke out2: 6731b5799018STheodore Ts'o ext4_exit_sysfs(); 6732b5799018STheodore Ts'o out3: 6733dd68314cSTheodore Ts'o ext4_exit_system_zone(); 6734b5799018STheodore Ts'o out4: 67355dabfc78STheodore Ts'o ext4_exit_pageio(); 6736b5799018STheodore Ts'o out5: 673722cfe4b4SEric Biggers ext4_exit_post_read_processing(); 67381dc0aa46SEric Whitney out6: 673922cfe4b4SEric Biggers ext4_exit_pending(); 674022cfe4b4SEric Biggers out7: 674151865fdaSZheng Liu ext4_exit_es(); 674251865fdaSZheng Liu 6743ac27a0ecSDave Kleikamp return err; 6744ac27a0ecSDave Kleikamp } 6745ac27a0ecSDave Kleikamp 67465dabfc78STheodore Ts'o static void __exit ext4_exit_fs(void) 6747ac27a0ecSDave Kleikamp { 6748bfff6873SLukas Czerner ext4_destroy_lazyinit_thread(); 674924b58424STheodore Ts'o unregister_as_ext2(); 675024b58424STheodore Ts'o unregister_as_ext3(); 675103010a33STheodore Ts'o unregister_filesystem(&ext4_fs_type); 6752ac27a0ecSDave Kleikamp destroy_inodecache(); 67535dabfc78STheodore Ts'o ext4_exit_mballoc(); 6754b5799018STheodore Ts'o ext4_exit_sysfs(); 67555dabfc78STheodore Ts'o ext4_exit_system_zone(); 67565dabfc78STheodore Ts'o ext4_exit_pageio(); 675722cfe4b4SEric Biggers ext4_exit_post_read_processing(); 6758dd12ed14SEric Sandeen ext4_exit_es(); 67591dc0aa46SEric Whitney ext4_exit_pending(); 6760ac27a0ecSDave Kleikamp } 6761ac27a0ecSDave Kleikamp 6762ac27a0ecSDave Kleikamp MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 676383982b6fSTheodore Ts'o MODULE_DESCRIPTION("Fourth Extended Filesystem"); 6764ac27a0ecSDave Kleikamp MODULE_LICENSE("GPL"); 67657ef79ad5STheodore Ts'o MODULE_SOFTDEP("pre: crc32c"); 67665dabfc78STheodore Ts'o module_init(ext4_init_fs) 67675dabfc78STheodore Ts'o module_exit(ext4_exit_fs) 6768