1f5166768STheodore Ts'o // SPDX-License-Identifier: GPL-2.0 2ac27a0ecSDave Kleikamp /* 3617ba13bSMingming Cao * linux/fs/ext4/super.c 4ac27a0ecSDave Kleikamp * 5ac27a0ecSDave Kleikamp * Copyright (C) 1992, 1993, 1994, 1995 6ac27a0ecSDave Kleikamp * Remy Card (card@masi.ibp.fr) 7ac27a0ecSDave Kleikamp * Laboratoire MASI - Institut Blaise Pascal 8ac27a0ecSDave Kleikamp * Universite Pierre et Marie Curie (Paris VI) 9ac27a0ecSDave Kleikamp * 10ac27a0ecSDave Kleikamp * from 11ac27a0ecSDave Kleikamp * 12ac27a0ecSDave Kleikamp * linux/fs/minix/inode.c 13ac27a0ecSDave Kleikamp * 14ac27a0ecSDave Kleikamp * Copyright (C) 1991, 1992 Linus Torvalds 15ac27a0ecSDave Kleikamp * 16ac27a0ecSDave Kleikamp * Big-endian to little-endian byte-swapping/bitmaps by 17ac27a0ecSDave Kleikamp * David S. Miller (davem@caip.rutgers.edu), 1995 18ac27a0ecSDave Kleikamp */ 19ac27a0ecSDave Kleikamp 20ac27a0ecSDave Kleikamp #include <linux/module.h> 21ac27a0ecSDave Kleikamp #include <linux/string.h> 22ac27a0ecSDave Kleikamp #include <linux/fs.h> 23ac27a0ecSDave Kleikamp #include <linux/time.h> 24c5ca7c76STheodore Ts'o #include <linux/vmalloc.h> 25ac27a0ecSDave Kleikamp #include <linux/slab.h> 26ac27a0ecSDave Kleikamp #include <linux/init.h> 27ac27a0ecSDave Kleikamp #include <linux/blkdev.h> 2866114cadSTejun Heo #include <linux/backing-dev.h> 29ac27a0ecSDave Kleikamp #include <linux/parser.h> 30ac27a0ecSDave Kleikamp #include <linux/buffer_head.h> 31a5694255SChristoph Hellwig #include <linux/exportfs.h> 32ac27a0ecSDave Kleikamp #include <linux/vfs.h> 33ac27a0ecSDave Kleikamp #include <linux/random.h> 34ac27a0ecSDave Kleikamp #include <linux/mount.h> 35ac27a0ecSDave Kleikamp #include <linux/namei.h> 36ac27a0ecSDave Kleikamp #include <linux/quotaops.h> 37ac27a0ecSDave Kleikamp #include <linux/seq_file.h> 383197ebdbSTheodore Ts'o #include <linux/ctype.h> 391330593eSVignesh Babu #include <linux/log2.h> 40717d50e4SAndreas Dilger #include <linux/crc16.h> 41ef510424SDan Williams #include <linux/dax.h> 427abc52c2SDan Magenheimer #include <linux/cleancache.h> 437c0f6ba6SLinus Torvalds #include <linux/uaccess.h> 44ee73f9a5SJeff Layton #include <linux/iversion.h> 45c83ad55eSGabriel Krisman Bertazi #include <linux/unicode.h> 46c6a564ffSChristoph Hellwig #include <linux/part_stat.h> 47bfff6873SLukas Czerner #include <linux/kthread.h> 48bfff6873SLukas Czerner #include <linux/freezer.h> 49bfff6873SLukas Czerner 503dcf5451SChristoph Hellwig #include "ext4.h" 514a092d73STheodore Ts'o #include "ext4_extents.h" /* Needed for trace points definition */ 523dcf5451SChristoph Hellwig #include "ext4_jbd2.h" 53ac27a0ecSDave Kleikamp #include "xattr.h" 54ac27a0ecSDave Kleikamp #include "acl.h" 553661d286STheodore Ts'o #include "mballoc.h" 560c9ec4beSDarrick J. Wong #include "fsmap.h" 57ac27a0ecSDave Kleikamp 589bffad1eSTheodore Ts'o #define CREATE_TRACE_POINTS 599bffad1eSTheodore Ts'o #include <trace/events/ext4.h> 609bffad1eSTheodore Ts'o 610b75a840SLukas Czerner static struct ext4_lazy_init *ext4_li_info; 620b75a840SLukas Czerner static struct mutex ext4_li_mtx; 63e294a537STheodore Ts'o static struct ratelimit_state ext4_mount_msg_ratelimit; 649f6200bbSTheodore Ts'o 65617ba13bSMingming Cao static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 66ac27a0ecSDave Kleikamp unsigned long journal_devnum); 672adf6da8STheodore Ts'o static int ext4_show_options(struct seq_file *seq, struct dentry *root); 68e2d67052STheodore Ts'o static int ext4_commit_super(struct super_block *sb, int sync); 6911215630SJan Kara static int ext4_mark_recovery_complete(struct super_block *sb, 70617ba13bSMingming Cao struct ext4_super_block *es); 7111215630SJan Kara static int ext4_clear_journal_err(struct super_block *sb, 72617ba13bSMingming Cao struct ext4_super_block *es); 73617ba13bSMingming Cao static int ext4_sync_fs(struct super_block *sb, int wait); 74617ba13bSMingming Cao static int ext4_remount(struct super_block *sb, int *flags, char *data); 75617ba13bSMingming Cao static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 76c4be0c1dSTakashi Sato static int ext4_unfreeze(struct super_block *sb); 77c4be0c1dSTakashi Sato static int ext4_freeze(struct super_block *sb); 78152a0836SAl Viro static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 79152a0836SAl Viro const char *dev_name, void *data); 802035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb); 812035e776STheodore Ts'o static inline int ext3_feature_set_ok(struct super_block *sb); 82d39195c3SAmir Goldstein static int ext4_feature_set_ok(struct super_block *sb, int readonly); 83bfff6873SLukas Czerner static void ext4_destroy_lazyinit_thread(void); 84bfff6873SLukas Czerner static void ext4_unregister_li_request(struct super_block *sb); 858f1f7453SEric Sandeen static void ext4_clear_request_list(void); 86c6cb7e77SEric Whitney static struct inode *ext4_get_journal_inode(struct super_block *sb, 87c6cb7e77SEric Whitney unsigned int journal_inum); 88ac27a0ecSDave Kleikamp 89e74031fdSJan Kara /* 90e74031fdSJan Kara * Lock ordering 91e74031fdSJan Kara * 92e74031fdSJan Kara * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and 93e74031fdSJan Kara * i_mmap_rwsem (inode->i_mmap_rwsem)! 94e74031fdSJan Kara * 95e74031fdSJan Kara * page fault path: 96c1e8d7c6SMichel Lespinasse * mmap_lock -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start -> 97e74031fdSJan Kara * page lock -> i_data_sem (rw) 98e74031fdSJan Kara * 99e74031fdSJan Kara * buffered write path: 100c1e8d7c6SMichel Lespinasse * sb_start_write -> i_mutex -> mmap_lock 101e74031fdSJan Kara * sb_start_write -> i_mutex -> transaction start -> page lock -> 102e74031fdSJan Kara * i_data_sem (rw) 103e74031fdSJan Kara * 104e74031fdSJan Kara * truncate: 1051d39834fSNikolay Borisov * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock 1061d39834fSNikolay Borisov * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start -> 1071d39834fSNikolay Borisov * i_data_sem (rw) 108e74031fdSJan Kara * 109e74031fdSJan Kara * direct IO: 110c1e8d7c6SMichel Lespinasse * sb_start_write -> i_mutex -> mmap_lock 1111d39834fSNikolay Borisov * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw) 112e74031fdSJan Kara * 113e74031fdSJan Kara * writepages: 114e74031fdSJan Kara * transaction start -> page lock(s) -> i_data_sem (rw) 115e74031fdSJan Kara */ 116e74031fdSJan Kara 117c290ea01SJan Kara #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 1182035e776STheodore Ts'o static struct file_system_type ext2_fs_type = { 1192035e776STheodore Ts'o .owner = THIS_MODULE, 1202035e776STheodore Ts'o .name = "ext2", 1212035e776STheodore Ts'o .mount = ext4_mount, 1222035e776STheodore Ts'o .kill_sb = kill_block_super, 1232035e776STheodore Ts'o .fs_flags = FS_REQUIRES_DEV, 1242035e776STheodore Ts'o }; 1257f78e035SEric W. Biederman MODULE_ALIAS_FS("ext2"); 126fa7614ddSEric W. Biederman MODULE_ALIAS("ext2"); 1272035e776STheodore Ts'o #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 1282035e776STheodore Ts'o #else 1292035e776STheodore Ts'o #define IS_EXT2_SB(sb) (0) 1302035e776STheodore Ts'o #endif 1312035e776STheodore Ts'o 1322035e776STheodore Ts'o 133ba69f9abSJan Kara static struct file_system_type ext3_fs_type = { 134ba69f9abSJan Kara .owner = THIS_MODULE, 135ba69f9abSJan Kara .name = "ext3", 136152a0836SAl Viro .mount = ext4_mount, 137ba69f9abSJan Kara .kill_sb = kill_block_super, 138ba69f9abSJan Kara .fs_flags = FS_REQUIRES_DEV, 139ba69f9abSJan Kara }; 1407f78e035SEric W. Biederman MODULE_ALIAS_FS("ext3"); 141fa7614ddSEric W. Biederman MODULE_ALIAS("ext3"); 142ba69f9abSJan Kara #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 143bd81d8eeSLaurent Vivier 144fb265c9cSTheodore Ts'o /* 145fb265c9cSTheodore Ts'o * This works like sb_bread() except it uses ERR_PTR for error 146fb265c9cSTheodore Ts'o * returns. Currently with sb_bread it's impossible to distinguish 147fb265c9cSTheodore Ts'o * between ENOMEM and EIO situations (since both result in a NULL 148fb265c9cSTheodore Ts'o * return. 149fb265c9cSTheodore Ts'o */ 150fb265c9cSTheodore Ts'o struct buffer_head * 151fb265c9cSTheodore Ts'o ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags) 152fb265c9cSTheodore Ts'o { 153fb265c9cSTheodore Ts'o struct buffer_head *bh = sb_getblk(sb, block); 154fb265c9cSTheodore Ts'o 155fb265c9cSTheodore Ts'o if (bh == NULL) 156fb265c9cSTheodore Ts'o return ERR_PTR(-ENOMEM); 157cf2834a5STheodore Ts'o if (ext4_buffer_uptodate(bh)) 158fb265c9cSTheodore Ts'o return bh; 159fb265c9cSTheodore Ts'o ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh); 160fb265c9cSTheodore Ts'o wait_on_buffer(bh); 161fb265c9cSTheodore Ts'o if (buffer_uptodate(bh)) 162fb265c9cSTheodore Ts'o return bh; 163fb265c9cSTheodore Ts'o put_bh(bh); 164fb265c9cSTheodore Ts'o return ERR_PTR(-EIO); 165fb265c9cSTheodore Ts'o } 166fb265c9cSTheodore Ts'o 167d25425f8SDarrick J. Wong static int ext4_verify_csum_type(struct super_block *sb, 168d25425f8SDarrick J. Wong struct ext4_super_block *es) 169d25425f8SDarrick J. Wong { 170e2b911c5SDarrick J. Wong if (!ext4_has_feature_metadata_csum(sb)) 171d25425f8SDarrick J. Wong return 1; 172d25425f8SDarrick J. Wong 173d25425f8SDarrick J. Wong return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 174d25425f8SDarrick J. Wong } 175d25425f8SDarrick J. Wong 176a9c47317SDarrick J. Wong static __le32 ext4_superblock_csum(struct super_block *sb, 177a9c47317SDarrick J. Wong struct ext4_super_block *es) 178a9c47317SDarrick J. Wong { 179a9c47317SDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 180a9c47317SDarrick J. Wong int offset = offsetof(struct ext4_super_block, s_checksum); 181a9c47317SDarrick J. Wong __u32 csum; 182a9c47317SDarrick J. Wong 183a9c47317SDarrick J. Wong csum = ext4_chksum(sbi, ~0, (char *)es, offset); 184a9c47317SDarrick J. Wong 185a9c47317SDarrick J. Wong return cpu_to_le32(csum); 186a9c47317SDarrick J. Wong } 187a9c47317SDarrick J. Wong 188c197855eSStephen Hemminger static int ext4_superblock_csum_verify(struct super_block *sb, 189a9c47317SDarrick J. Wong struct ext4_super_block *es) 190a9c47317SDarrick J. Wong { 1919aa5d32bSDmitry Monakhov if (!ext4_has_metadata_csum(sb)) 192a9c47317SDarrick J. Wong return 1; 193a9c47317SDarrick J. Wong 194a9c47317SDarrick J. Wong return es->s_checksum == ext4_superblock_csum(sb, es); 195a9c47317SDarrick J. Wong } 196a9c47317SDarrick J. Wong 19706db49e6STheodore Ts'o void ext4_superblock_csum_set(struct super_block *sb) 198a9c47317SDarrick J. Wong { 19906db49e6STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 20006db49e6STheodore Ts'o 2019aa5d32bSDmitry Monakhov if (!ext4_has_metadata_csum(sb)) 202a9c47317SDarrick J. Wong return; 203a9c47317SDarrick J. Wong 204acaa5326SConstantine Sapuntzakis /* 205acaa5326SConstantine Sapuntzakis * Locking the superblock prevents the scenario 206acaa5326SConstantine Sapuntzakis * where: 207acaa5326SConstantine Sapuntzakis * 1) a first thread pauses during checksum calculation. 208acaa5326SConstantine Sapuntzakis * 2) a second thread updates the superblock, recalculates 209acaa5326SConstantine Sapuntzakis * the checksum, and updates s_checksum 210acaa5326SConstantine Sapuntzakis * 3) the first thread resumes and finishes its checksum calculation 211acaa5326SConstantine Sapuntzakis * and updates s_checksum with a potentially stale or torn value. 212acaa5326SConstantine Sapuntzakis */ 213acaa5326SConstantine Sapuntzakis lock_buffer(EXT4_SB(sb)->s_sbh); 214a9c47317SDarrick J. Wong es->s_checksum = ext4_superblock_csum(sb, es); 215acaa5326SConstantine Sapuntzakis unlock_buffer(EXT4_SB(sb)->s_sbh); 216a9c47317SDarrick J. Wong } 217a9c47317SDarrick J. Wong 2188fadc143SAlexandre Ratchov ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 2198fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 220bd81d8eeSLaurent Vivier { 2213a14589cSAneesh Kumar K.V return le32_to_cpu(bg->bg_block_bitmap_lo) | 2228fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 2238fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 224bd81d8eeSLaurent Vivier } 225bd81d8eeSLaurent Vivier 2268fadc143SAlexandre Ratchov ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 2278fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 228bd81d8eeSLaurent Vivier { 2295272f837SAneesh Kumar K.V return le32_to_cpu(bg->bg_inode_bitmap_lo) | 2308fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 2318fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 232bd81d8eeSLaurent Vivier } 233bd81d8eeSLaurent Vivier 2348fadc143SAlexandre Ratchov ext4_fsblk_t ext4_inode_table(struct super_block *sb, 2358fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 236bd81d8eeSLaurent Vivier { 2375272f837SAneesh Kumar K.V return le32_to_cpu(bg->bg_inode_table_lo) | 2388fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 2398fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 240bd81d8eeSLaurent Vivier } 241bd81d8eeSLaurent Vivier 242021b65bbSTheodore Ts'o __u32 ext4_free_group_clusters(struct super_block *sb, 243560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 244560671a0SAneesh Kumar K.V { 245560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_free_blocks_count_lo) | 246560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 247560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 248560671a0SAneesh Kumar K.V } 249560671a0SAneesh Kumar K.V 250560671a0SAneesh Kumar K.V __u32 ext4_free_inodes_count(struct super_block *sb, 251560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 252560671a0SAneesh Kumar K.V { 253560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_free_inodes_count_lo) | 254560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 255560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 256560671a0SAneesh Kumar K.V } 257560671a0SAneesh Kumar K.V 258560671a0SAneesh Kumar K.V __u32 ext4_used_dirs_count(struct super_block *sb, 259560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 260560671a0SAneesh Kumar K.V { 261560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_used_dirs_count_lo) | 262560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 263560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 264560671a0SAneesh Kumar K.V } 265560671a0SAneesh Kumar K.V 266560671a0SAneesh Kumar K.V __u32 ext4_itable_unused_count(struct super_block *sb, 267560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 268560671a0SAneesh Kumar K.V { 269560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_itable_unused_lo) | 270560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 271560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 272560671a0SAneesh Kumar K.V } 273560671a0SAneesh Kumar K.V 2748fadc143SAlexandre Ratchov void ext4_block_bitmap_set(struct super_block *sb, 2758fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 276bd81d8eeSLaurent Vivier { 2773a14589cSAneesh Kumar K.V bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 2788fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 2798fadc143SAlexandre Ratchov bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 280bd81d8eeSLaurent Vivier } 281bd81d8eeSLaurent Vivier 2828fadc143SAlexandre Ratchov void ext4_inode_bitmap_set(struct super_block *sb, 2838fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 284bd81d8eeSLaurent Vivier { 2855272f837SAneesh Kumar K.V bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 2868fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 2878fadc143SAlexandre Ratchov bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 288bd81d8eeSLaurent Vivier } 289bd81d8eeSLaurent Vivier 2908fadc143SAlexandre Ratchov void ext4_inode_table_set(struct super_block *sb, 2918fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 292bd81d8eeSLaurent Vivier { 2935272f837SAneesh Kumar K.V bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 2948fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 2958fadc143SAlexandre Ratchov bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 296bd81d8eeSLaurent Vivier } 297bd81d8eeSLaurent Vivier 298021b65bbSTheodore Ts'o void ext4_free_group_clusters_set(struct super_block *sb, 299560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 300560671a0SAneesh Kumar K.V { 301560671a0SAneesh Kumar K.V bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 302560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 303560671a0SAneesh Kumar K.V bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 304560671a0SAneesh Kumar K.V } 305560671a0SAneesh Kumar K.V 306560671a0SAneesh Kumar K.V void ext4_free_inodes_set(struct super_block *sb, 307560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 308560671a0SAneesh Kumar K.V { 309560671a0SAneesh Kumar K.V bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 310560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 311560671a0SAneesh Kumar K.V bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 312560671a0SAneesh Kumar K.V } 313560671a0SAneesh Kumar K.V 314560671a0SAneesh Kumar K.V void ext4_used_dirs_set(struct super_block *sb, 315560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 316560671a0SAneesh Kumar K.V { 317560671a0SAneesh Kumar K.V bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 318560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 319560671a0SAneesh Kumar K.V bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 320560671a0SAneesh Kumar K.V } 321560671a0SAneesh Kumar K.V 322560671a0SAneesh Kumar K.V void ext4_itable_unused_set(struct super_block *sb, 323560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 324560671a0SAneesh Kumar K.V { 325560671a0SAneesh Kumar K.V bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 326560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 327560671a0SAneesh Kumar K.V bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 328560671a0SAneesh Kumar K.V } 329560671a0SAneesh Kumar K.V 3306a0678a7SArnd Bergmann static void __ext4_update_tstamp(__le32 *lo, __u8 *hi) 3316a0678a7SArnd Bergmann { 3326a0678a7SArnd Bergmann time64_t now = ktime_get_real_seconds(); 3336a0678a7SArnd Bergmann 3346a0678a7SArnd Bergmann now = clamp_val(now, 0, (1ull << 40) - 1); 3356a0678a7SArnd Bergmann 3366a0678a7SArnd Bergmann *lo = cpu_to_le32(lower_32_bits(now)); 3376a0678a7SArnd Bergmann *hi = upper_32_bits(now); 3386a0678a7SArnd Bergmann } 3396a0678a7SArnd Bergmann 3406a0678a7SArnd Bergmann static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi) 3416a0678a7SArnd Bergmann { 3426a0678a7SArnd Bergmann return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo); 3436a0678a7SArnd Bergmann } 3446a0678a7SArnd Bergmann #define ext4_update_tstamp(es, tstamp) \ 3456a0678a7SArnd Bergmann __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 3466a0678a7SArnd Bergmann #define ext4_get_tstamp(es, tstamp) \ 3476a0678a7SArnd Bergmann __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 348d3d1faf6SCurt Wohlgemuth 34954d3adbcSTheodore Ts'o static void __save_error_info(struct super_block *sb, int error, 35054d3adbcSTheodore Ts'o __u32 ino, __u64 block, 35154d3adbcSTheodore Ts'o const char *func, unsigned int line) 3521c13d5c0STheodore Ts'o { 3531c13d5c0STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 35454d3adbcSTheodore Ts'o int err; 3551c13d5c0STheodore Ts'o 3561c13d5c0STheodore Ts'o EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 3571b46617bSTheodore Ts'o if (bdev_read_only(sb->s_bdev)) 3581b46617bSTheodore Ts'o return; 3591c13d5c0STheodore Ts'o es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 3606a0678a7SArnd Bergmann ext4_update_tstamp(es, s_last_error_time); 3611c13d5c0STheodore Ts'o strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); 3621c13d5c0STheodore Ts'o es->s_last_error_line = cpu_to_le32(line); 36354d3adbcSTheodore Ts'o es->s_last_error_ino = cpu_to_le32(ino); 36454d3adbcSTheodore Ts'o es->s_last_error_block = cpu_to_le64(block); 36554d3adbcSTheodore Ts'o switch (error) { 36654d3adbcSTheodore Ts'o case EIO: 36754d3adbcSTheodore Ts'o err = EXT4_ERR_EIO; 36854d3adbcSTheodore Ts'o break; 36954d3adbcSTheodore Ts'o case ENOMEM: 37054d3adbcSTheodore Ts'o err = EXT4_ERR_ENOMEM; 37154d3adbcSTheodore Ts'o break; 37254d3adbcSTheodore Ts'o case EFSBADCRC: 37354d3adbcSTheodore Ts'o err = EXT4_ERR_EFSBADCRC; 37454d3adbcSTheodore Ts'o break; 37554d3adbcSTheodore Ts'o case 0: 37654d3adbcSTheodore Ts'o case EFSCORRUPTED: 37754d3adbcSTheodore Ts'o err = EXT4_ERR_EFSCORRUPTED; 37854d3adbcSTheodore Ts'o break; 37954d3adbcSTheodore Ts'o case ENOSPC: 38054d3adbcSTheodore Ts'o err = EXT4_ERR_ENOSPC; 38154d3adbcSTheodore Ts'o break; 38254d3adbcSTheodore Ts'o case ENOKEY: 38354d3adbcSTheodore Ts'o err = EXT4_ERR_ENOKEY; 38454d3adbcSTheodore Ts'o break; 38554d3adbcSTheodore Ts'o case EROFS: 38654d3adbcSTheodore Ts'o err = EXT4_ERR_EROFS; 38754d3adbcSTheodore Ts'o break; 38854d3adbcSTheodore Ts'o case EFBIG: 38954d3adbcSTheodore Ts'o err = EXT4_ERR_EFBIG; 39054d3adbcSTheodore Ts'o break; 39154d3adbcSTheodore Ts'o case EEXIST: 39254d3adbcSTheodore Ts'o err = EXT4_ERR_EEXIST; 39354d3adbcSTheodore Ts'o break; 39454d3adbcSTheodore Ts'o case ERANGE: 39554d3adbcSTheodore Ts'o err = EXT4_ERR_ERANGE; 39654d3adbcSTheodore Ts'o break; 39754d3adbcSTheodore Ts'o case EOVERFLOW: 39854d3adbcSTheodore Ts'o err = EXT4_ERR_EOVERFLOW; 39954d3adbcSTheodore Ts'o break; 40054d3adbcSTheodore Ts'o case EBUSY: 40154d3adbcSTheodore Ts'o err = EXT4_ERR_EBUSY; 40254d3adbcSTheodore Ts'o break; 40354d3adbcSTheodore Ts'o case ENOTDIR: 40454d3adbcSTheodore Ts'o err = EXT4_ERR_ENOTDIR; 40554d3adbcSTheodore Ts'o break; 40654d3adbcSTheodore Ts'o case ENOTEMPTY: 40754d3adbcSTheodore Ts'o err = EXT4_ERR_ENOTEMPTY; 40854d3adbcSTheodore Ts'o break; 40954d3adbcSTheodore Ts'o case ESHUTDOWN: 41054d3adbcSTheodore Ts'o err = EXT4_ERR_ESHUTDOWN; 41154d3adbcSTheodore Ts'o break; 41254d3adbcSTheodore Ts'o case EFAULT: 41354d3adbcSTheodore Ts'o err = EXT4_ERR_EFAULT; 41454d3adbcSTheodore Ts'o break; 41554d3adbcSTheodore Ts'o default: 41654d3adbcSTheodore Ts'o err = EXT4_ERR_UNKNOWN; 41754d3adbcSTheodore Ts'o } 41854d3adbcSTheodore Ts'o es->s_last_error_errcode = err; 4191c13d5c0STheodore Ts'o if (!es->s_first_error_time) { 4201c13d5c0STheodore Ts'o es->s_first_error_time = es->s_last_error_time; 4216a0678a7SArnd Bergmann es->s_first_error_time_hi = es->s_last_error_time_hi; 4221c13d5c0STheodore Ts'o strncpy(es->s_first_error_func, func, 4231c13d5c0STheodore Ts'o sizeof(es->s_first_error_func)); 4241c13d5c0STheodore Ts'o es->s_first_error_line = cpu_to_le32(line); 4251c13d5c0STheodore Ts'o es->s_first_error_ino = es->s_last_error_ino; 4261c13d5c0STheodore Ts'o es->s_first_error_block = es->s_last_error_block; 427878520acSTheodore Ts'o es->s_first_error_errcode = es->s_last_error_errcode; 4281c13d5c0STheodore Ts'o } 42966e61a9eSTheodore Ts'o /* 43066e61a9eSTheodore Ts'o * Start the daily error reporting function if it hasn't been 43166e61a9eSTheodore Ts'o * started already 43266e61a9eSTheodore Ts'o */ 43366e61a9eSTheodore Ts'o if (!es->s_error_count) 43466e61a9eSTheodore Ts'o mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); 435ba39ebb6SWei Yongjun le32_add_cpu(&es->s_error_count, 1); 4361c13d5c0STheodore Ts'o } 4371c13d5c0STheodore Ts'o 43854d3adbcSTheodore Ts'o static void save_error_info(struct super_block *sb, int error, 43954d3adbcSTheodore Ts'o __u32 ino, __u64 block, 44054d3adbcSTheodore Ts'o const char *func, unsigned int line) 4411c13d5c0STheodore Ts'o { 44254d3adbcSTheodore Ts'o __save_error_info(sb, error, ino, block, func, line); 443c96e2b85SEric Sandeen if (!bdev_read_only(sb->s_bdev)) 4441c13d5c0STheodore Ts'o ext4_commit_super(sb, 1); 4451c13d5c0STheodore Ts'o } 4461c13d5c0STheodore Ts'o 447bdfe0cbdSTheodore Ts'o /* 448bdfe0cbdSTheodore Ts'o * The del_gendisk() function uninitializes the disk-specific data 449bdfe0cbdSTheodore Ts'o * structures, including the bdi structure, without telling anyone 450bdfe0cbdSTheodore Ts'o * else. Once this happens, any attempt to call mark_buffer_dirty() 451bdfe0cbdSTheodore Ts'o * (for example, by ext4_commit_super), will cause a kernel OOPS. 452bdfe0cbdSTheodore Ts'o * This is a kludge to prevent these oops until we can put in a proper 453bdfe0cbdSTheodore Ts'o * hook in del_gendisk() to inform the VFS and file system layers. 454bdfe0cbdSTheodore Ts'o */ 455bdfe0cbdSTheodore Ts'o static int block_device_ejected(struct super_block *sb) 456bdfe0cbdSTheodore Ts'o { 457bdfe0cbdSTheodore Ts'o struct inode *bd_inode = sb->s_bdev->bd_inode; 458bdfe0cbdSTheodore Ts'o struct backing_dev_info *bdi = inode_to_bdi(bd_inode); 459bdfe0cbdSTheodore Ts'o 460bdfe0cbdSTheodore Ts'o return bdi->dev == NULL; 461bdfe0cbdSTheodore Ts'o } 462bdfe0cbdSTheodore Ts'o 46318aadd47SBobi Jam static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 46418aadd47SBobi Jam { 46518aadd47SBobi Jam struct super_block *sb = journal->j_private; 46618aadd47SBobi Jam struct ext4_sb_info *sbi = EXT4_SB(sb); 46718aadd47SBobi Jam int error = is_journal_aborted(journal); 4685d3ee208SDmitry Monakhov struct ext4_journal_cb_entry *jce; 46918aadd47SBobi Jam 4705d3ee208SDmitry Monakhov BUG_ON(txn->t_state == T_FINISHED); 471a0154344SDaeho Jeong 472a0154344SDaeho Jeong ext4_process_freed_data(sb, txn->t_tid); 473a0154344SDaeho Jeong 47418aadd47SBobi Jam spin_lock(&sbi->s_md_lock); 4755d3ee208SDmitry Monakhov while (!list_empty(&txn->t_private_list)) { 4765d3ee208SDmitry Monakhov jce = list_entry(txn->t_private_list.next, 4775d3ee208SDmitry Monakhov struct ext4_journal_cb_entry, jce_list); 47818aadd47SBobi Jam list_del_init(&jce->jce_list); 47918aadd47SBobi Jam spin_unlock(&sbi->s_md_lock); 48018aadd47SBobi Jam jce->jce_func(sb, jce, error); 48118aadd47SBobi Jam spin_lock(&sbi->s_md_lock); 48218aadd47SBobi Jam } 48318aadd47SBobi Jam spin_unlock(&sbi->s_md_lock); 48418aadd47SBobi Jam } 4851c13d5c0STheodore Ts'o 4861dc1097fSJan Kara static bool system_going_down(void) 4871dc1097fSJan Kara { 4881dc1097fSJan Kara return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 4891dc1097fSJan Kara || system_state == SYSTEM_RESTART; 4901dc1097fSJan Kara } 4911dc1097fSJan Kara 492ac27a0ecSDave Kleikamp /* Deal with the reporting of failure conditions on a filesystem such as 493ac27a0ecSDave Kleikamp * inconsistencies detected or read IO failures. 494ac27a0ecSDave Kleikamp * 495ac27a0ecSDave Kleikamp * On ext2, we can store the error state of the filesystem in the 496617ba13bSMingming Cao * superblock. That is not possible on ext4, because we may have other 497ac27a0ecSDave Kleikamp * write ordering constraints on the superblock which prevent us from 498ac27a0ecSDave Kleikamp * writing it out straight away; and given that the journal is about to 499ac27a0ecSDave Kleikamp * be aborted, we can't rely on the current, or future, transactions to 500ac27a0ecSDave Kleikamp * write out the superblock safely. 501ac27a0ecSDave Kleikamp * 502dab291afSMingming Cao * We'll just use the jbd2_journal_abort() error code to record an error in 503d6b198bcSThadeu Lima de Souza Cascardo * the journal instead. On recovery, the journal will complain about 504ac27a0ecSDave Kleikamp * that error until we've noted it down and cleared it. 505ac27a0ecSDave Kleikamp */ 506ac27a0ecSDave Kleikamp 507617ba13bSMingming Cao static void ext4_handle_error(struct super_block *sb) 508ac27a0ecSDave Kleikamp { 509327eaf73STheodore Ts'o if (test_opt(sb, WARN_ON_ERROR)) 510327eaf73STheodore Ts'o WARN_ON_ONCE(1); 511327eaf73STheodore Ts'o 512bc98a42cSDavid Howells if (sb_rdonly(sb)) 513ac27a0ecSDave Kleikamp return; 514ac27a0ecSDave Kleikamp 515ac27a0ecSDave Kleikamp if (!test_opt(sb, ERRORS_CONT)) { 516617ba13bSMingming Cao journal_t *journal = EXT4_SB(sb)->s_journal; 517ac27a0ecSDave Kleikamp 5184ab2f15bSTheodore Ts'o EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 519ac27a0ecSDave Kleikamp if (journal) 520dab291afSMingming Cao jbd2_journal_abort(journal, -EIO); 521ac27a0ecSDave Kleikamp } 5221dc1097fSJan Kara /* 5231dc1097fSJan Kara * We force ERRORS_RO behavior when system is rebooting. Otherwise we 5241dc1097fSJan Kara * could panic during 'reboot -f' as the underlying device got already 5251dc1097fSJan Kara * disabled. 5261dc1097fSJan Kara */ 5271dc1097fSJan Kara if (test_opt(sb, ERRORS_RO) || system_going_down()) { 528b31e1552SEric Sandeen ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 5294418e141SDmitry Monakhov /* 5304418e141SDmitry Monakhov * Make sure updated value of ->s_mount_flags will be visible 5314418e141SDmitry Monakhov * before ->s_flags update 5324418e141SDmitry Monakhov */ 5334418e141SDmitry Monakhov smp_wmb(); 5341751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 5351dc1097fSJan Kara } else if (test_opt(sb, ERRORS_PANIC)) { 536617ba13bSMingming Cao panic("EXT4-fs (device %s): panic forced after error\n", 537ac27a0ecSDave Kleikamp sb->s_id); 538ac27a0ecSDave Kleikamp } 5394327ba52SDaeho Jeong } 540ac27a0ecSDave Kleikamp 541efbed4dcSTheodore Ts'o #define ext4_error_ratelimit(sb) \ 542efbed4dcSTheodore Ts'o ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 543efbed4dcSTheodore Ts'o "EXT4-fs error") 544efbed4dcSTheodore Ts'o 54512062dddSEric Sandeen void __ext4_error(struct super_block *sb, const char *function, 54654d3adbcSTheodore Ts'o unsigned int line, int error, __u64 block, 54754d3adbcSTheodore Ts'o const char *fmt, ...) 548ac27a0ecSDave Kleikamp { 5490ff2ea7dSJoe Perches struct va_format vaf; 550ac27a0ecSDave Kleikamp va_list args; 551ac27a0ecSDave Kleikamp 5520db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 5530db1ff22STheodore Ts'o return; 5540db1ff22STheodore Ts'o 555ccf0f32aSTheodore Ts'o trace_ext4_error(sb, function, line); 556efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 557ac27a0ecSDave Kleikamp va_start(args, fmt); 5580ff2ea7dSJoe Perches vaf.fmt = fmt; 5590ff2ea7dSJoe Perches vaf.va = &args; 560efbed4dcSTheodore Ts'o printk(KERN_CRIT 561efbed4dcSTheodore Ts'o "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 5620ff2ea7dSJoe Perches sb->s_id, function, line, current->comm, &vaf); 563ac27a0ecSDave Kleikamp va_end(args); 564efbed4dcSTheodore Ts'o } 56554d3adbcSTheodore Ts'o save_error_info(sb, error, 0, block, function, line); 566617ba13bSMingming Cao ext4_handle_error(sb); 567ac27a0ecSDave Kleikamp } 568ac27a0ecSDave Kleikamp 569e7c96e8eSJoe Perches void __ext4_error_inode(struct inode *inode, const char *function, 57054d3adbcSTheodore Ts'o unsigned int line, ext4_fsblk_t block, int error, 571273df556SFrank Mayhar const char *fmt, ...) 572273df556SFrank Mayhar { 573273df556SFrank Mayhar va_list args; 574f7c21177STheodore Ts'o struct va_format vaf; 575273df556SFrank Mayhar 5760db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 5770db1ff22STheodore Ts'o return; 5780db1ff22STheodore Ts'o 579ccf0f32aSTheodore Ts'o trace_ext4_error(inode->i_sb, function, line); 580efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(inode->i_sb)) { 581273df556SFrank Mayhar va_start(args, fmt); 582f7c21177STheodore Ts'o vaf.fmt = fmt; 583f7c21177STheodore Ts'o vaf.va = &args; 584c398eda0STheodore Ts'o if (block) 585d9ee81daSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 586d9ee81daSJoe Perches "inode #%lu: block %llu: comm %s: %pV\n", 587d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 588d9ee81daSJoe Perches block, current->comm, &vaf); 589d9ee81daSJoe Perches else 590d9ee81daSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 591d9ee81daSJoe Perches "inode #%lu: comm %s: %pV\n", 592d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 593d9ee81daSJoe Perches current->comm, &vaf); 594273df556SFrank Mayhar va_end(args); 595efbed4dcSTheodore Ts'o } 59654d3adbcSTheodore Ts'o save_error_info(inode->i_sb, error, inode->i_ino, block, 59754d3adbcSTheodore Ts'o function, line); 598273df556SFrank Mayhar ext4_handle_error(inode->i_sb); 599273df556SFrank Mayhar } 600273df556SFrank Mayhar 601e7c96e8eSJoe Perches void __ext4_error_file(struct file *file, const char *function, 602f7c21177STheodore Ts'o unsigned int line, ext4_fsblk_t block, 603f7c21177STheodore Ts'o const char *fmt, ...) 604273df556SFrank Mayhar { 605273df556SFrank Mayhar va_list args; 606f7c21177STheodore Ts'o struct va_format vaf; 607496ad9aaSAl Viro struct inode *inode = file_inode(file); 608273df556SFrank Mayhar char pathname[80], *path; 609273df556SFrank Mayhar 6100db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 6110db1ff22STheodore Ts'o return; 6120db1ff22STheodore Ts'o 613ccf0f32aSTheodore Ts'o trace_ext4_error(inode->i_sb, function, line); 614efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(inode->i_sb)) { 6159bf39ab2SMiklos Szeredi path = file_path(file, pathname, sizeof(pathname)); 616f9a62d09SDan Carpenter if (IS_ERR(path)) 617273df556SFrank Mayhar path = "(unknown)"; 618f7c21177STheodore Ts'o va_start(args, fmt); 619f7c21177STheodore Ts'o vaf.fmt = fmt; 620f7c21177STheodore Ts'o vaf.va = &args; 621d9ee81daSJoe Perches if (block) 622d9ee81daSJoe Perches printk(KERN_CRIT 623d9ee81daSJoe Perches "EXT4-fs error (device %s): %s:%d: inode #%lu: " 624d9ee81daSJoe Perches "block %llu: comm %s: path %s: %pV\n", 625d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 626d9ee81daSJoe Perches block, current->comm, path, &vaf); 627d9ee81daSJoe Perches else 628d9ee81daSJoe Perches printk(KERN_CRIT 629d9ee81daSJoe Perches "EXT4-fs error (device %s): %s:%d: inode #%lu: " 630d9ee81daSJoe Perches "comm %s: path %s: %pV\n", 631d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 632d9ee81daSJoe Perches current->comm, path, &vaf); 633273df556SFrank Mayhar va_end(args); 634efbed4dcSTheodore Ts'o } 63554d3adbcSTheodore Ts'o save_error_info(inode->i_sb, EFSCORRUPTED, inode->i_ino, block, 63654d3adbcSTheodore Ts'o function, line); 637273df556SFrank Mayhar ext4_handle_error(inode->i_sb); 638273df556SFrank Mayhar } 639273df556SFrank Mayhar 640722887ddSTheodore Ts'o const char *ext4_decode_error(struct super_block *sb, int errno, 641ac27a0ecSDave Kleikamp char nbuf[16]) 642ac27a0ecSDave Kleikamp { 643ac27a0ecSDave Kleikamp char *errstr = NULL; 644ac27a0ecSDave Kleikamp 645ac27a0ecSDave Kleikamp switch (errno) { 6466a797d27SDarrick J. Wong case -EFSCORRUPTED: 6476a797d27SDarrick J. Wong errstr = "Corrupt filesystem"; 6486a797d27SDarrick J. Wong break; 6496a797d27SDarrick J. Wong case -EFSBADCRC: 6506a797d27SDarrick J. Wong errstr = "Filesystem failed CRC"; 6516a797d27SDarrick J. Wong break; 652ac27a0ecSDave Kleikamp case -EIO: 653ac27a0ecSDave Kleikamp errstr = "IO failure"; 654ac27a0ecSDave Kleikamp break; 655ac27a0ecSDave Kleikamp case -ENOMEM: 656ac27a0ecSDave Kleikamp errstr = "Out of memory"; 657ac27a0ecSDave Kleikamp break; 658ac27a0ecSDave Kleikamp case -EROFS: 65978f1ddbbSTheodore Ts'o if (!sb || (EXT4_SB(sb)->s_journal && 66078f1ddbbSTheodore Ts'o EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 661ac27a0ecSDave Kleikamp errstr = "Journal has aborted"; 662ac27a0ecSDave Kleikamp else 663ac27a0ecSDave Kleikamp errstr = "Readonly filesystem"; 664ac27a0ecSDave Kleikamp break; 665ac27a0ecSDave Kleikamp default: 666ac27a0ecSDave Kleikamp /* If the caller passed in an extra buffer for unknown 667ac27a0ecSDave Kleikamp * errors, textualise them now. Else we just return 668ac27a0ecSDave Kleikamp * NULL. */ 669ac27a0ecSDave Kleikamp if (nbuf) { 670ac27a0ecSDave Kleikamp /* Check for truncated error codes... */ 671ac27a0ecSDave Kleikamp if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 672ac27a0ecSDave Kleikamp errstr = nbuf; 673ac27a0ecSDave Kleikamp } 674ac27a0ecSDave Kleikamp break; 675ac27a0ecSDave Kleikamp } 676ac27a0ecSDave Kleikamp 677ac27a0ecSDave Kleikamp return errstr; 678ac27a0ecSDave Kleikamp } 679ac27a0ecSDave Kleikamp 680617ba13bSMingming Cao /* __ext4_std_error decodes expected errors from journaling functions 681ac27a0ecSDave Kleikamp * automatically and invokes the appropriate error response. */ 682ac27a0ecSDave Kleikamp 683c398eda0STheodore Ts'o void __ext4_std_error(struct super_block *sb, const char *function, 684c398eda0STheodore Ts'o unsigned int line, int errno) 685ac27a0ecSDave Kleikamp { 686ac27a0ecSDave Kleikamp char nbuf[16]; 687ac27a0ecSDave Kleikamp const char *errstr; 688ac27a0ecSDave Kleikamp 6890db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 6900db1ff22STheodore Ts'o return; 6910db1ff22STheodore Ts'o 692ac27a0ecSDave Kleikamp /* Special case: if the error is EROFS, and we're not already 693ac27a0ecSDave Kleikamp * inside a transaction, then there's really no point in logging 694ac27a0ecSDave Kleikamp * an error. */ 695bc98a42cSDavid Howells if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb)) 696ac27a0ecSDave Kleikamp return; 697ac27a0ecSDave Kleikamp 698efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 699617ba13bSMingming Cao errstr = ext4_decode_error(sb, errno, nbuf); 700c398eda0STheodore Ts'o printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 701c398eda0STheodore Ts'o sb->s_id, function, line, errstr); 702efbed4dcSTheodore Ts'o } 703ac27a0ecSDave Kleikamp 70454d3adbcSTheodore Ts'o save_error_info(sb, -errno, 0, 0, function, line); 705617ba13bSMingming Cao ext4_handle_error(sb); 706ac27a0ecSDave Kleikamp } 707ac27a0ecSDave Kleikamp 708ac27a0ecSDave Kleikamp /* 709617ba13bSMingming Cao * ext4_abort is a much stronger failure handler than ext4_error. The 710ac27a0ecSDave Kleikamp * abort function may be used to deal with unrecoverable failures such 711ac27a0ecSDave Kleikamp * as journal IO errors or ENOMEM at a critical moment in log management. 712ac27a0ecSDave Kleikamp * 713ac27a0ecSDave Kleikamp * We unconditionally force the filesystem into an ABORT|READONLY state, 714ac27a0ecSDave Kleikamp * unless the error response on the fs has been set to panic in which 715ac27a0ecSDave Kleikamp * case we take the easy way out and panic immediately. 716ac27a0ecSDave Kleikamp */ 717ac27a0ecSDave Kleikamp 718c67d859eSTheodore Ts'o void __ext4_abort(struct super_block *sb, const char *function, 71954d3adbcSTheodore Ts'o unsigned int line, int error, const char *fmt, ...) 720ac27a0ecSDave Kleikamp { 721651e1c3bSJoe Perches struct va_format vaf; 722ac27a0ecSDave Kleikamp va_list args; 723ac27a0ecSDave Kleikamp 7240db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 7250db1ff22STheodore Ts'o return; 7260db1ff22STheodore Ts'o 72754d3adbcSTheodore Ts'o save_error_info(sb, error, 0, 0, function, line); 728ac27a0ecSDave Kleikamp va_start(args, fmt); 729651e1c3bSJoe Perches vaf.fmt = fmt; 730651e1c3bSJoe Perches vaf.va = &args; 731651e1c3bSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n", 732651e1c3bSJoe Perches sb->s_id, function, line, &vaf); 733ac27a0ecSDave Kleikamp va_end(args); 734ac27a0ecSDave Kleikamp 735bc98a42cSDavid Howells if (sb_rdonly(sb) == 0) { 7364ab2f15bSTheodore Ts'o EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 7377b97d868Szhangyi (F) if (EXT4_SB(sb)->s_journal) 7387b97d868Szhangyi (F) jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); 7397b97d868Szhangyi (F) 7407b97d868Szhangyi (F) ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 7414418e141SDmitry Monakhov /* 7424418e141SDmitry Monakhov * Make sure updated value of ->s_mount_flags will be visible 7434418e141SDmitry Monakhov * before ->s_flags update 7444418e141SDmitry Monakhov */ 7454418e141SDmitry Monakhov smp_wmb(); 7461751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 7471c13d5c0STheodore Ts'o } 7487b97d868Szhangyi (F) if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) 7491c13d5c0STheodore Ts'o panic("EXT4-fs panic from previous error\n"); 750ac27a0ecSDave Kleikamp } 751ac27a0ecSDave Kleikamp 752e7c96e8eSJoe Perches void __ext4_msg(struct super_block *sb, 753e7c96e8eSJoe Perches const char *prefix, const char *fmt, ...) 754b31e1552SEric Sandeen { 7550ff2ea7dSJoe Perches struct va_format vaf; 756b31e1552SEric Sandeen va_list args; 757b31e1552SEric Sandeen 7581cf006edSDmitry Monakhov atomic_inc(&EXT4_SB(sb)->s_msg_count); 759efbed4dcSTheodore Ts'o if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs")) 760efbed4dcSTheodore Ts'o return; 761efbed4dcSTheodore Ts'o 762b31e1552SEric Sandeen va_start(args, fmt); 7630ff2ea7dSJoe Perches vaf.fmt = fmt; 7640ff2ea7dSJoe Perches vaf.va = &args; 7650ff2ea7dSJoe Perches printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 766b31e1552SEric Sandeen va_end(args); 767b31e1552SEric Sandeen } 768b31e1552SEric Sandeen 7691cf006edSDmitry Monakhov static int ext4_warning_ratelimit(struct super_block *sb) 7701cf006edSDmitry Monakhov { 7711cf006edSDmitry Monakhov atomic_inc(&EXT4_SB(sb)->s_warning_count); 7721cf006edSDmitry Monakhov return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), 7731cf006edSDmitry Monakhov "EXT4-fs warning"); 7741cf006edSDmitry Monakhov } 775b03a2f7eSAndreas Dilger 77612062dddSEric Sandeen void __ext4_warning(struct super_block *sb, const char *function, 777c398eda0STheodore Ts'o unsigned int line, const char *fmt, ...) 778ac27a0ecSDave Kleikamp { 7790ff2ea7dSJoe Perches struct va_format vaf; 780ac27a0ecSDave Kleikamp va_list args; 781ac27a0ecSDave Kleikamp 782b03a2f7eSAndreas Dilger if (!ext4_warning_ratelimit(sb)) 783efbed4dcSTheodore Ts'o return; 784efbed4dcSTheodore Ts'o 785ac27a0ecSDave Kleikamp va_start(args, fmt); 7860ff2ea7dSJoe Perches vaf.fmt = fmt; 7870ff2ea7dSJoe Perches vaf.va = &args; 7880ff2ea7dSJoe Perches printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 7890ff2ea7dSJoe Perches sb->s_id, function, line, &vaf); 790ac27a0ecSDave Kleikamp va_end(args); 791ac27a0ecSDave Kleikamp } 792ac27a0ecSDave Kleikamp 793b03a2f7eSAndreas Dilger void __ext4_warning_inode(const struct inode *inode, const char *function, 794b03a2f7eSAndreas Dilger unsigned int line, const char *fmt, ...) 795b03a2f7eSAndreas Dilger { 796b03a2f7eSAndreas Dilger struct va_format vaf; 797b03a2f7eSAndreas Dilger va_list args; 798b03a2f7eSAndreas Dilger 799b03a2f7eSAndreas Dilger if (!ext4_warning_ratelimit(inode->i_sb)) 800b03a2f7eSAndreas Dilger return; 801b03a2f7eSAndreas Dilger 802b03a2f7eSAndreas Dilger va_start(args, fmt); 803b03a2f7eSAndreas Dilger vaf.fmt = fmt; 804b03a2f7eSAndreas Dilger vaf.va = &args; 805b03a2f7eSAndreas Dilger printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " 806b03a2f7eSAndreas Dilger "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, 807b03a2f7eSAndreas Dilger function, line, inode->i_ino, current->comm, &vaf); 808b03a2f7eSAndreas Dilger va_end(args); 809b03a2f7eSAndreas Dilger } 810b03a2f7eSAndreas Dilger 811e29136f8STheodore Ts'o void __ext4_grp_locked_error(const char *function, unsigned int line, 812e29136f8STheodore Ts'o struct super_block *sb, ext4_group_t grp, 813e29136f8STheodore Ts'o unsigned long ino, ext4_fsblk_t block, 814e29136f8STheodore Ts'o const char *fmt, ...) 8155d1b1b3fSAneesh Kumar K.V __releases(bitlock) 8165d1b1b3fSAneesh Kumar K.V __acquires(bitlock) 8175d1b1b3fSAneesh Kumar K.V { 8180ff2ea7dSJoe Perches struct va_format vaf; 8195d1b1b3fSAneesh Kumar K.V va_list args; 8205d1b1b3fSAneesh Kumar K.V 8210db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 8220db1ff22STheodore Ts'o return; 8230db1ff22STheodore Ts'o 824ccf0f32aSTheodore Ts'o trace_ext4_error(sb, function, line); 82554d3adbcSTheodore Ts'o __save_error_info(sb, EFSCORRUPTED, ino, block, function, line); 8260ff2ea7dSJoe Perches 827efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 8285d1b1b3fSAneesh Kumar K.V va_start(args, fmt); 8290ff2ea7dSJoe Perches vaf.fmt = fmt; 8300ff2ea7dSJoe Perches vaf.va = &args; 83121149d61SRobin Dong printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 832e29136f8STheodore Ts'o sb->s_id, function, line, grp); 833e29136f8STheodore Ts'o if (ino) 8340ff2ea7dSJoe Perches printk(KERN_CONT "inode %lu: ", ino); 835e29136f8STheodore Ts'o if (block) 836efbed4dcSTheodore Ts'o printk(KERN_CONT "block %llu:", 837efbed4dcSTheodore Ts'o (unsigned long long) block); 8380ff2ea7dSJoe Perches printk(KERN_CONT "%pV\n", &vaf); 8395d1b1b3fSAneesh Kumar K.V va_end(args); 840efbed4dcSTheodore Ts'o } 8415d1b1b3fSAneesh Kumar K.V 842327eaf73STheodore Ts'o if (test_opt(sb, WARN_ON_ERROR)) 843327eaf73STheodore Ts'o WARN_ON_ONCE(1); 844327eaf73STheodore Ts'o 8455d1b1b3fSAneesh Kumar K.V if (test_opt(sb, ERRORS_CONT)) { 846e2d67052STheodore Ts'o ext4_commit_super(sb, 0); 8475d1b1b3fSAneesh Kumar K.V return; 8485d1b1b3fSAneesh Kumar K.V } 8491c13d5c0STheodore Ts'o 8505d1b1b3fSAneesh Kumar K.V ext4_unlock_group(sb, grp); 85106f29cc8SZhouyi Zhou ext4_commit_super(sb, 1); 8525d1b1b3fSAneesh Kumar K.V ext4_handle_error(sb); 8535d1b1b3fSAneesh Kumar K.V /* 8545d1b1b3fSAneesh Kumar K.V * We only get here in the ERRORS_RO case; relocking the group 8555d1b1b3fSAneesh Kumar K.V * may be dangerous, but nothing bad will happen since the 8565d1b1b3fSAneesh Kumar K.V * filesystem will have already been marked read/only and the 8575d1b1b3fSAneesh Kumar K.V * journal has been aborted. We return 1 as a hint to callers 8585d1b1b3fSAneesh Kumar K.V * who might what to use the return value from 85925985edcSLucas De Marchi * ext4_grp_locked_error() to distinguish between the 8605d1b1b3fSAneesh Kumar K.V * ERRORS_CONT and ERRORS_RO case, and perhaps return more 8615d1b1b3fSAneesh Kumar K.V * aggressively from the ext4 function in question, with a 8625d1b1b3fSAneesh Kumar K.V * more appropriate error code. 8635d1b1b3fSAneesh Kumar K.V */ 8645d1b1b3fSAneesh Kumar K.V ext4_lock_group(sb, grp); 8655d1b1b3fSAneesh Kumar K.V return; 8665d1b1b3fSAneesh Kumar K.V } 8675d1b1b3fSAneesh Kumar K.V 868db79e6d1SWang Shilong void ext4_mark_group_bitmap_corrupted(struct super_block *sb, 869db79e6d1SWang Shilong ext4_group_t group, 870db79e6d1SWang Shilong unsigned int flags) 871db79e6d1SWang Shilong { 872db79e6d1SWang Shilong struct ext4_sb_info *sbi = EXT4_SB(sb); 873db79e6d1SWang Shilong struct ext4_group_info *grp = ext4_get_group_info(sb, group); 874db79e6d1SWang Shilong struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); 8759af0b3d1SWang Shilong int ret; 876db79e6d1SWang Shilong 8779af0b3d1SWang Shilong if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) { 8789af0b3d1SWang Shilong ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, 8799af0b3d1SWang Shilong &grp->bb_state); 8809af0b3d1SWang Shilong if (!ret) 881db79e6d1SWang Shilong percpu_counter_sub(&sbi->s_freeclusters_counter, 882db79e6d1SWang Shilong grp->bb_free); 883db79e6d1SWang Shilong } 884db79e6d1SWang Shilong 8859af0b3d1SWang Shilong if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) { 8869af0b3d1SWang Shilong ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, 8879af0b3d1SWang Shilong &grp->bb_state); 8889af0b3d1SWang Shilong if (!ret && gdp) { 889db79e6d1SWang Shilong int count; 890db79e6d1SWang Shilong 891db79e6d1SWang Shilong count = ext4_free_inodes_count(sb, gdp); 892db79e6d1SWang Shilong percpu_counter_sub(&sbi->s_freeinodes_counter, 893db79e6d1SWang Shilong count); 894db79e6d1SWang Shilong } 895db79e6d1SWang Shilong } 896db79e6d1SWang Shilong } 897db79e6d1SWang Shilong 898617ba13bSMingming Cao void ext4_update_dynamic_rev(struct super_block *sb) 899ac27a0ecSDave Kleikamp { 900617ba13bSMingming Cao struct ext4_super_block *es = EXT4_SB(sb)->s_es; 901ac27a0ecSDave Kleikamp 902617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 903ac27a0ecSDave Kleikamp return; 904ac27a0ecSDave Kleikamp 90512062dddSEric Sandeen ext4_warning(sb, 906ac27a0ecSDave Kleikamp "updating to rev %d because of new feature flag, " 907ac27a0ecSDave Kleikamp "running e2fsck is recommended", 908617ba13bSMingming Cao EXT4_DYNAMIC_REV); 909ac27a0ecSDave Kleikamp 910617ba13bSMingming Cao es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 911617ba13bSMingming Cao es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 912617ba13bSMingming Cao es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 913ac27a0ecSDave Kleikamp /* leave es->s_feature_*compat flags alone */ 914ac27a0ecSDave Kleikamp /* es->s_uuid will be set by e2fsck if empty */ 915ac27a0ecSDave Kleikamp 916ac27a0ecSDave Kleikamp /* 917ac27a0ecSDave Kleikamp * The rest of the superblock fields should be zero, and if not it 918ac27a0ecSDave Kleikamp * means they are likely already in use, so leave them alone. We 919ac27a0ecSDave Kleikamp * can leave it up to e2fsck to clean up any inconsistencies there. 920ac27a0ecSDave Kleikamp */ 921ac27a0ecSDave Kleikamp } 922ac27a0ecSDave Kleikamp 923ac27a0ecSDave Kleikamp /* 924ac27a0ecSDave Kleikamp * Open the external journal device 925ac27a0ecSDave Kleikamp */ 926b31e1552SEric Sandeen static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 927ac27a0ecSDave Kleikamp { 928ac27a0ecSDave Kleikamp struct block_device *bdev; 929ac27a0ecSDave Kleikamp 930d4d77629STejun Heo bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 931ac27a0ecSDave Kleikamp if (IS_ERR(bdev)) 932ac27a0ecSDave Kleikamp goto fail; 933ac27a0ecSDave Kleikamp return bdev; 934ac27a0ecSDave Kleikamp 935ac27a0ecSDave Kleikamp fail: 936ea3edd4dSChristoph Hellwig ext4_msg(sb, KERN_ERR, 937ea3edd4dSChristoph Hellwig "failed to open journal device unknown-block(%u,%u) %ld", 938ea3edd4dSChristoph Hellwig MAJOR(dev), MINOR(dev), PTR_ERR(bdev)); 939ac27a0ecSDave Kleikamp return NULL; 940ac27a0ecSDave Kleikamp } 941ac27a0ecSDave Kleikamp 942ac27a0ecSDave Kleikamp /* 943ac27a0ecSDave Kleikamp * Release the journal device 944ac27a0ecSDave Kleikamp */ 9454385bab1SAl Viro static void ext4_blkdev_put(struct block_device *bdev) 946ac27a0ecSDave Kleikamp { 9474385bab1SAl Viro blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 948ac27a0ecSDave Kleikamp } 949ac27a0ecSDave Kleikamp 9504385bab1SAl Viro static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 951ac27a0ecSDave Kleikamp { 952ac27a0ecSDave Kleikamp struct block_device *bdev; 953ee7ed3aaSChunguang Xu bdev = sbi->s_journal_bdev; 954ac27a0ecSDave Kleikamp if (bdev) { 9554385bab1SAl Viro ext4_blkdev_put(bdev); 956ee7ed3aaSChunguang Xu sbi->s_journal_bdev = NULL; 957ac27a0ecSDave Kleikamp } 958ac27a0ecSDave Kleikamp } 959ac27a0ecSDave Kleikamp 960ac27a0ecSDave Kleikamp static inline struct inode *orphan_list_entry(struct list_head *l) 961ac27a0ecSDave Kleikamp { 962617ba13bSMingming Cao return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 963ac27a0ecSDave Kleikamp } 964ac27a0ecSDave Kleikamp 965617ba13bSMingming Cao static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 966ac27a0ecSDave Kleikamp { 967ac27a0ecSDave Kleikamp struct list_head *l; 968ac27a0ecSDave Kleikamp 969b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 970ac27a0ecSDave Kleikamp le32_to_cpu(sbi->s_es->s_last_orphan)); 971ac27a0ecSDave Kleikamp 972ac27a0ecSDave Kleikamp printk(KERN_ERR "sb_info orphan list:\n"); 973ac27a0ecSDave Kleikamp list_for_each(l, &sbi->s_orphan) { 974ac27a0ecSDave Kleikamp struct inode *inode = orphan_list_entry(l); 975ac27a0ecSDave Kleikamp printk(KERN_ERR " " 976ac27a0ecSDave Kleikamp "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 977ac27a0ecSDave Kleikamp inode->i_sb->s_id, inode->i_ino, inode, 978ac27a0ecSDave Kleikamp inode->i_mode, inode->i_nlink, 979ac27a0ecSDave Kleikamp NEXT_ORPHAN(inode)); 980ac27a0ecSDave Kleikamp } 981ac27a0ecSDave Kleikamp } 982ac27a0ecSDave Kleikamp 983957153fcSJan Kara #ifdef CONFIG_QUOTA 984957153fcSJan Kara static int ext4_quota_off(struct super_block *sb, int type); 985957153fcSJan Kara 986957153fcSJan Kara static inline void ext4_quota_off_umount(struct super_block *sb) 987957153fcSJan Kara { 988957153fcSJan Kara int type; 989957153fcSJan Kara 990957153fcSJan Kara /* Use our quota_off function to clear inode flags etc. */ 991957153fcSJan Kara for (type = 0; type < EXT4_MAXQUOTAS; type++) 992957153fcSJan Kara ext4_quota_off(sb, type); 993957153fcSJan Kara } 99433458eabSTheodore Ts'o 99533458eabSTheodore Ts'o /* 99633458eabSTheodore Ts'o * This is a helper function which is used in the mount/remount 99733458eabSTheodore Ts'o * codepaths (which holds s_umount) to fetch the quota file name. 99833458eabSTheodore Ts'o */ 99933458eabSTheodore Ts'o static inline char *get_qf_name(struct super_block *sb, 100033458eabSTheodore Ts'o struct ext4_sb_info *sbi, 100133458eabSTheodore Ts'o int type) 100233458eabSTheodore Ts'o { 100333458eabSTheodore Ts'o return rcu_dereference_protected(sbi->s_qf_names[type], 100433458eabSTheodore Ts'o lockdep_is_held(&sb->s_umount)); 100533458eabSTheodore Ts'o } 1006957153fcSJan Kara #else 1007957153fcSJan Kara static inline void ext4_quota_off_umount(struct super_block *sb) 1008957153fcSJan Kara { 1009957153fcSJan Kara } 1010957153fcSJan Kara #endif 1011957153fcSJan Kara 1012617ba13bSMingming Cao static void ext4_put_super(struct super_block *sb) 1013ac27a0ecSDave Kleikamp { 1014617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 1015617ba13bSMingming Cao struct ext4_super_block *es = sbi->s_es; 10161d0c3924STheodore Ts'o struct buffer_head **group_desc; 10177c990728SSuraj Jitindar Singh struct flex_groups **flex_groups; 101897abd7d4STheodore Ts'o int aborted = 0; 1019ef2cabf7SHidehiro Kawai int i, err; 1020ac27a0ecSDave Kleikamp 1021857ac889SLukas Czerner ext4_unregister_li_request(sb); 1022957153fcSJan Kara ext4_quota_off_umount(sb); 1023e0ccfd95SChristoph Hellwig 10242e8fa54eSJan Kara destroy_workqueue(sbi->rsv_conversion_wq); 10254c0425ffSMingming Cao 10265e47868fSRitesh Harjani /* 10275e47868fSRitesh Harjani * Unregister sysfs before destroying jbd2 journal. 10285e47868fSRitesh Harjani * Since we could still access attr_journal_task attribute via sysfs 10295e47868fSRitesh Harjani * path which could have sbi->s_journal->j_task as NULL 10305e47868fSRitesh Harjani */ 10315e47868fSRitesh Harjani ext4_unregister_sysfs(sb); 10325e47868fSRitesh Harjani 10330390131bSFrank Mayhar if (sbi->s_journal) { 103497abd7d4STheodore Ts'o aborted = is_journal_aborted(sbi->s_journal); 1035ef2cabf7SHidehiro Kawai err = jbd2_journal_destroy(sbi->s_journal); 103647b4a50bSJan Kara sbi->s_journal = NULL; 1037878520acSTheodore Ts'o if ((err < 0) && !aborted) { 103854d3adbcSTheodore Ts'o ext4_abort(sb, -err, "Couldn't clean up the journal"); 10390390131bSFrank Mayhar } 1040878520acSTheodore Ts'o } 1041d4edac31SJosef Bacik 1042d3922a77SZheng Liu ext4_es_unregister_shrinker(sbi); 10439105bb14SAl Viro del_timer_sync(&sbi->s_err_report); 1044d4edac31SJosef Bacik ext4_release_system_zone(sb); 1045d4edac31SJosef Bacik ext4_mb_release(sb); 1046d4edac31SJosef Bacik ext4_ext_release(sb); 1047d4edac31SJosef Bacik 1048bc98a42cSDavid Howells if (!sb_rdonly(sb) && !aborted) { 1049e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 1050ac27a0ecSDave Kleikamp es->s_state = cpu_to_le16(sbi->s_mount_state); 1051ac27a0ecSDave Kleikamp } 1052bc98a42cSDavid Howells if (!sb_rdonly(sb)) 1053a8e25a83SArtem Bityutskiy ext4_commit_super(sb, 1); 1054a8e25a83SArtem Bityutskiy 10551d0c3924STheodore Ts'o rcu_read_lock(); 10561d0c3924STheodore Ts'o group_desc = rcu_dereference(sbi->s_group_desc); 1057ac27a0ecSDave Kleikamp for (i = 0; i < sbi->s_gdb_count; i++) 10581d0c3924STheodore Ts'o brelse(group_desc[i]); 10591d0c3924STheodore Ts'o kvfree(group_desc); 10607c990728SSuraj Jitindar Singh flex_groups = rcu_dereference(sbi->s_flex_groups); 10617c990728SSuraj Jitindar Singh if (flex_groups) { 10627c990728SSuraj Jitindar Singh for (i = 0; i < sbi->s_flex_groups_allocated; i++) 10637c990728SSuraj Jitindar Singh kvfree(flex_groups[i]); 10647c990728SSuraj Jitindar Singh kvfree(flex_groups); 10657c990728SSuraj Jitindar Singh } 10661d0c3924STheodore Ts'o rcu_read_unlock(); 106757042651STheodore Ts'o percpu_counter_destroy(&sbi->s_freeclusters_counter); 1068ac27a0ecSDave Kleikamp percpu_counter_destroy(&sbi->s_freeinodes_counter); 1069ac27a0ecSDave Kleikamp percpu_counter_destroy(&sbi->s_dirs_counter); 107057042651STheodore Ts'o percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 1071bbd55937SEric Biggers percpu_free_rwsem(&sbi->s_writepages_rwsem); 1072ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1073a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 107433458eabSTheodore Ts'o kfree(get_qf_name(sb, sbi, i)); 1075ac27a0ecSDave Kleikamp #endif 1076ac27a0ecSDave Kleikamp 1077ac27a0ecSDave Kleikamp /* Debugging code just in case the in-memory inode orphan list 1078ac27a0ecSDave Kleikamp * isn't empty. The on-disk one can be non-empty if we've 1079ac27a0ecSDave Kleikamp * detected an error and taken the fs readonly, but the 1080ac27a0ecSDave Kleikamp * in-memory list had better be clean by this point. */ 1081ac27a0ecSDave Kleikamp if (!list_empty(&sbi->s_orphan)) 1082ac27a0ecSDave Kleikamp dump_orphan_list(sb, sbi); 1083ac27a0ecSDave Kleikamp J_ASSERT(list_empty(&sbi->s_orphan)); 1084ac27a0ecSDave Kleikamp 108589d96a6fSTheodore Ts'o sync_blockdev(sb->s_bdev); 1086f98393a6SPeter Zijlstra invalidate_bdev(sb->s_bdev); 1087ee7ed3aaSChunguang Xu if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) { 1088ac27a0ecSDave Kleikamp /* 1089ac27a0ecSDave Kleikamp * Invalidate the journal device's buffers. We don't want them 1090ac27a0ecSDave Kleikamp * floating about in memory - the physical journal device may 1091ac27a0ecSDave Kleikamp * hotswapped, and it breaks the `ro-after' testing code. 1092ac27a0ecSDave Kleikamp */ 1093ee7ed3aaSChunguang Xu sync_blockdev(sbi->s_journal_bdev); 1094ee7ed3aaSChunguang Xu invalidate_bdev(sbi->s_journal_bdev); 1095617ba13bSMingming Cao ext4_blkdev_remove(sbi); 1096ac27a0ecSDave Kleikamp } 109750c15df6SChengguang Xu 1098dec214d0STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 1099dec214d0STahsin Erdogan sbi->s_ea_inode_cache = NULL; 110050c15df6SChengguang Xu 110147387409STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 110247387409STahsin Erdogan sbi->s_ea_block_cache = NULL; 110350c15df6SChengguang Xu 1104c5e06d10SJohann Lombardi if (sbi->s_mmp_tsk) 1105c5e06d10SJohann Lombardi kthread_stop(sbi->s_mmp_tsk); 11069060dd2cSEric Sandeen brelse(sbi->s_sbh); 1107ac27a0ecSDave Kleikamp sb->s_fs_info = NULL; 11083197ebdbSTheodore Ts'o /* 11093197ebdbSTheodore Ts'o * Now that we are completely done shutting down the 11103197ebdbSTheodore Ts'o * superblock, we need to actually destroy the kobject. 11113197ebdbSTheodore Ts'o */ 11123197ebdbSTheodore Ts'o kobject_put(&sbi->s_kobj); 11133197ebdbSTheodore Ts'o wait_for_completion(&sbi->s_kobj_unregister); 11140441984aSDarrick J. Wong if (sbi->s_chksum_driver) 11150441984aSDarrick J. Wong crypto_free_shash(sbi->s_chksum_driver); 1116705895b6SPekka Enberg kfree(sbi->s_blockgroup_lock); 11175e405595SDan Williams fs_put_dax(sbi->s_daxdev); 1118ed318a6cSEric Biggers fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx); 1119c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 1120c83ad55eSGabriel Krisman Bertazi utf8_unload(sbi->s_encoding); 1121c83ad55eSGabriel Krisman Bertazi #endif 1122ac27a0ecSDave Kleikamp kfree(sbi); 1123ac27a0ecSDave Kleikamp } 1124ac27a0ecSDave Kleikamp 1125e18b890bSChristoph Lameter static struct kmem_cache *ext4_inode_cachep; 1126ac27a0ecSDave Kleikamp 1127ac27a0ecSDave Kleikamp /* 1128ac27a0ecSDave Kleikamp * Called inside transaction, so use GFP_NOFS 1129ac27a0ecSDave Kleikamp */ 1130617ba13bSMingming Cao static struct inode *ext4_alloc_inode(struct super_block *sb) 1131ac27a0ecSDave Kleikamp { 1132617ba13bSMingming Cao struct ext4_inode_info *ei; 1133ac27a0ecSDave Kleikamp 1134e6b4f8daSChristoph Lameter ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 1135ac27a0ecSDave Kleikamp if (!ei) 1136ac27a0ecSDave Kleikamp return NULL; 11370b8e58a1SAndreas Dilger 1138ee73f9a5SJeff Layton inode_set_iversion(&ei->vfs_inode, 1); 1139202ee5dfSTheodore Ts'o spin_lock_init(&ei->i_raw_lock); 1140c9de560dSAlex Tomas INIT_LIST_HEAD(&ei->i_prealloc_list); 114127bc446eSbrookxu atomic_set(&ei->i_prealloc_active, 0); 1142c9de560dSAlex Tomas spin_lock_init(&ei->i_prealloc_lock); 11439a26b661SZheng Liu ext4_es_init_tree(&ei->i_es_tree); 11449a26b661SZheng Liu rwlock_init(&ei->i_es_lock); 1145edaa53caSZheng Liu INIT_LIST_HEAD(&ei->i_es_list); 1146eb68d0e2SZheng Liu ei->i_es_all_nr = 0; 1147edaa53caSZheng Liu ei->i_es_shk_nr = 0; 1148dd475925SJan Kara ei->i_es_shrink_lblk = 0; 1149d2a17637SMingming Cao ei->i_reserved_data_blocks = 0; 1150d2a17637SMingming Cao spin_lock_init(&(ei->i_block_reservation_lock)); 11511dc0aa46SEric Whitney ext4_init_pending_tree(&ei->i_pending_tree); 1152a9e7f447SDmitry Monakhov #ifdef CONFIG_QUOTA 1153a9e7f447SDmitry Monakhov ei->i_reserved_quota = 0; 115496c7e0d9SJan Kara memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 1155a9e7f447SDmitry Monakhov #endif 11568aefcd55STheodore Ts'o ei->jinode = NULL; 11572e8fa54eSJan Kara INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 1158744692dcSJiaying Zhang spin_lock_init(&ei->i_completed_io_lock); 1159b436b9beSJan Kara ei->i_sync_tid = 0; 1160b436b9beSJan Kara ei->i_datasync_tid = 0; 1161e27f41e1SDmitry Monakhov atomic_set(&ei->i_unwritten, 0); 11622e8fa54eSJan Kara INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 1163ac27a0ecSDave Kleikamp return &ei->vfs_inode; 1164ac27a0ecSDave Kleikamp } 1165ac27a0ecSDave Kleikamp 11667ff9c073STheodore Ts'o static int ext4_drop_inode(struct inode *inode) 11677ff9c073STheodore Ts'o { 11687ff9c073STheodore Ts'o int drop = generic_drop_inode(inode); 11697ff9c073STheodore Ts'o 117029b3692eSEric Biggers if (!drop) 117129b3692eSEric Biggers drop = fscrypt_drop_inode(inode); 117229b3692eSEric Biggers 11737ff9c073STheodore Ts'o trace_ext4_drop_inode(inode, drop); 11747ff9c073STheodore Ts'o return drop; 11757ff9c073STheodore Ts'o } 11767ff9c073STheodore Ts'o 117794053139SAl Viro static void ext4_free_in_core_inode(struct inode *inode) 1178fa0d7e3dSNick Piggin { 11792c58d548SEric Biggers fscrypt_free_inode(inode); 1180fa0d7e3dSNick Piggin kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 1181fa0d7e3dSNick Piggin } 1182fa0d7e3dSNick Piggin 1183617ba13bSMingming Cao static void ext4_destroy_inode(struct inode *inode) 1184ac27a0ecSDave Kleikamp { 11859f7dd93dSVasily Averin if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 1186b31e1552SEric Sandeen ext4_msg(inode->i_sb, KERN_ERR, 1187b31e1552SEric Sandeen "Inode %lu (%p): orphan list check failed!", 1188b31e1552SEric Sandeen inode->i_ino, EXT4_I(inode)); 11899f7dd93dSVasily Averin print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 11909f7dd93dSVasily Averin EXT4_I(inode), sizeof(struct ext4_inode_info), 11919f7dd93dSVasily Averin true); 11929f7dd93dSVasily Averin dump_stack(); 11939f7dd93dSVasily Averin } 1194ac27a0ecSDave Kleikamp } 1195ac27a0ecSDave Kleikamp 119651cc5068SAlexey Dobriyan static void init_once(void *foo) 1197ac27a0ecSDave Kleikamp { 1198617ba13bSMingming Cao struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 1199ac27a0ecSDave Kleikamp 1200ac27a0ecSDave Kleikamp INIT_LIST_HEAD(&ei->i_orphan); 1201ac27a0ecSDave Kleikamp init_rwsem(&ei->xattr_sem); 12020e855ac8SAneesh Kumar K.V init_rwsem(&ei->i_data_sem); 1203ea3d7209SJan Kara init_rwsem(&ei->i_mmap_sem); 1204ac27a0ecSDave Kleikamp inode_init_once(&ei->vfs_inode); 1205ac27a0ecSDave Kleikamp } 1206ac27a0ecSDave Kleikamp 1207e67bc2b3SFabian Frederick static int __init init_inodecache(void) 1208ac27a0ecSDave Kleikamp { 1209f8dd7c70SDavid Windsor ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache", 1210f8dd7c70SDavid Windsor sizeof(struct ext4_inode_info), 0, 1211f8dd7c70SDavid Windsor (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD| 1212f8dd7c70SDavid Windsor SLAB_ACCOUNT), 1213f8dd7c70SDavid Windsor offsetof(struct ext4_inode_info, i_data), 1214f8dd7c70SDavid Windsor sizeof_field(struct ext4_inode_info, i_data), 121520c2df83SPaul Mundt init_once); 1216617ba13bSMingming Cao if (ext4_inode_cachep == NULL) 1217ac27a0ecSDave Kleikamp return -ENOMEM; 1218ac27a0ecSDave Kleikamp return 0; 1219ac27a0ecSDave Kleikamp } 1220ac27a0ecSDave Kleikamp 1221ac27a0ecSDave Kleikamp static void destroy_inodecache(void) 1222ac27a0ecSDave Kleikamp { 12238c0a8537SKirill A. Shutemov /* 12248c0a8537SKirill A. Shutemov * Make sure all delayed rcu free inodes are flushed before we 12258c0a8537SKirill A. Shutemov * destroy cache. 12268c0a8537SKirill A. Shutemov */ 12278c0a8537SKirill A. Shutemov rcu_barrier(); 1228617ba13bSMingming Cao kmem_cache_destroy(ext4_inode_cachep); 1229ac27a0ecSDave Kleikamp } 1230ac27a0ecSDave Kleikamp 12310930fcc1SAl Viro void ext4_clear_inode(struct inode *inode) 1232ac27a0ecSDave Kleikamp { 12330930fcc1SAl Viro invalidate_inode_buffers(inode); 1234dbd5768fSJan Kara clear_inode(inode); 123527bc446eSbrookxu ext4_discard_preallocations(inode, 0); 123651865fdaSZheng Liu ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1237f4c2d372SJan Kara dquot_drop(inode); 12388aefcd55STheodore Ts'o if (EXT4_I(inode)->jinode) { 12398aefcd55STheodore Ts'o jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 12408aefcd55STheodore Ts'o EXT4_I(inode)->jinode); 12418aefcd55STheodore Ts'o jbd2_free_inode(EXT4_I(inode)->jinode); 12428aefcd55STheodore Ts'o EXT4_I(inode)->jinode = NULL; 12438aefcd55STheodore Ts'o } 12443d204e24SEric Biggers fscrypt_put_encryption_info(inode); 1245c93d8f88SEric Biggers fsverity_cleanup_inode(inode); 1246ac27a0ecSDave Kleikamp } 1247ac27a0ecSDave Kleikamp 12481b961ac0SChristoph Hellwig static struct inode *ext4_nfs_get_inode(struct super_block *sb, 12491b961ac0SChristoph Hellwig u64 ino, u32 generation) 1250ac27a0ecSDave Kleikamp { 1251ac27a0ecSDave Kleikamp struct inode *inode; 1252ac27a0ecSDave Kleikamp 12538a363970STheodore Ts'o /* 1254ac27a0ecSDave Kleikamp * Currently we don't know the generation for parent directory, so 1255ac27a0ecSDave Kleikamp * a generation of 0 means "accept any" 1256ac27a0ecSDave Kleikamp */ 12578a363970STheodore Ts'o inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); 12581d1fe1eeSDavid Howells if (IS_ERR(inode)) 12591d1fe1eeSDavid Howells return ERR_CAST(inode); 12601d1fe1eeSDavid Howells if (generation && inode->i_generation != generation) { 1261ac27a0ecSDave Kleikamp iput(inode); 1262ac27a0ecSDave Kleikamp return ERR_PTR(-ESTALE); 1263ac27a0ecSDave Kleikamp } 12641b961ac0SChristoph Hellwig 12651b961ac0SChristoph Hellwig return inode; 1266ac27a0ecSDave Kleikamp } 12671b961ac0SChristoph Hellwig 12681b961ac0SChristoph Hellwig static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 12691b961ac0SChristoph Hellwig int fh_len, int fh_type) 12701b961ac0SChristoph Hellwig { 12711b961ac0SChristoph Hellwig return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 12721b961ac0SChristoph Hellwig ext4_nfs_get_inode); 12731b961ac0SChristoph Hellwig } 12741b961ac0SChristoph Hellwig 12751b961ac0SChristoph Hellwig static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 12761b961ac0SChristoph Hellwig int fh_len, int fh_type) 12771b961ac0SChristoph Hellwig { 12781b961ac0SChristoph Hellwig return generic_fh_to_parent(sb, fid, fh_len, fh_type, 12791b961ac0SChristoph Hellwig ext4_nfs_get_inode); 1280ac27a0ecSDave Kleikamp } 1281ac27a0ecSDave Kleikamp 1282fde87268STheodore Ts'o static int ext4_nfs_commit_metadata(struct inode *inode) 1283fde87268STheodore Ts'o { 1284fde87268STheodore Ts'o struct writeback_control wbc = { 1285fde87268STheodore Ts'o .sync_mode = WB_SYNC_ALL 1286fde87268STheodore Ts'o }; 1287fde87268STheodore Ts'o 1288fde87268STheodore Ts'o trace_ext4_nfs_commit_metadata(inode); 1289fde87268STheodore Ts'o return ext4_write_inode(inode, &wbc); 1290fde87268STheodore Ts'o } 1291fde87268STheodore Ts'o 1292c39a7f84SToshiyuki Okajima /* 1293c39a7f84SToshiyuki Okajima * Try to release metadata pages (indirect blocks, directories) which are 1294c39a7f84SToshiyuki Okajima * mapped via the block device. Since these pages could have journal heads 1295c39a7f84SToshiyuki Okajima * which would prevent try_to_free_buffers() from freeing them, we must use 1296c39a7f84SToshiyuki Okajima * jbd2 layer's try_to_free_buffers() function to release them. 1297c39a7f84SToshiyuki Okajima */ 12980b8e58a1SAndreas Dilger static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 12990b8e58a1SAndreas Dilger gfp_t wait) 1300c39a7f84SToshiyuki Okajima { 1301c39a7f84SToshiyuki Okajima journal_t *journal = EXT4_SB(sb)->s_journal; 1302c39a7f84SToshiyuki Okajima 1303c39a7f84SToshiyuki Okajima WARN_ON(PageChecked(page)); 1304c39a7f84SToshiyuki Okajima if (!page_has_buffers(page)) 1305c39a7f84SToshiyuki Okajima return 0; 1306c39a7f84SToshiyuki Okajima if (journal) 1307529a781eSzhangyi (F) return jbd2_journal_try_to_free_buffers(journal, page); 1308529a781eSzhangyi (F) 1309c39a7f84SToshiyuki Okajima return try_to_free_buffers(page); 1310c39a7f84SToshiyuki Okajima } 1311c39a7f84SToshiyuki Okajima 1312643fa961SChandan Rajendra #ifdef CONFIG_FS_ENCRYPTION 1313a7550b30SJaegeuk Kim static int ext4_get_context(struct inode *inode, void *ctx, size_t len) 1314a7550b30SJaegeuk Kim { 1315a7550b30SJaegeuk Kim return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION, 1316a7550b30SJaegeuk Kim EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len); 1317a7550b30SJaegeuk Kim } 1318a7550b30SJaegeuk Kim 1319a7550b30SJaegeuk Kim static int ext4_set_context(struct inode *inode, const void *ctx, size_t len, 1320a7550b30SJaegeuk Kim void *fs_data) 1321a7550b30SJaegeuk Kim { 13222f8f5e76SEric Biggers handle_t *handle = fs_data; 1323c1a5d5f6STahsin Erdogan int res, res2, credits, retries = 0; 1324a7550b30SJaegeuk Kim 13259ce0151aSEric Biggers /* 13269ce0151aSEric Biggers * Encrypting the root directory is not allowed because e2fsck expects 13279ce0151aSEric Biggers * lost+found to exist and be unencrypted, and encrypting the root 13289ce0151aSEric Biggers * directory would imply encrypting the lost+found directory as well as 13299ce0151aSEric Biggers * the filename "lost+found" itself. 13309ce0151aSEric Biggers */ 13319ce0151aSEric Biggers if (inode->i_ino == EXT4_ROOT_INO) 13329ce0151aSEric Biggers return -EPERM; 1333a7550b30SJaegeuk Kim 13347d3e06a8SRoss Zwisler if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode))) 13357d3e06a8SRoss Zwisler return -EINVAL; 13367d3e06a8SRoss Zwisler 1337b383a73fSIra Weiny if (ext4_test_inode_flag(inode, EXT4_INODE_DAX)) 1338b383a73fSIra Weiny return -EOPNOTSUPP; 1339b383a73fSIra Weiny 134094840e3cSEric Biggers res = ext4_convert_inline_data(inode); 134194840e3cSEric Biggers if (res) 134294840e3cSEric Biggers return res; 134394840e3cSEric Biggers 13442f8f5e76SEric Biggers /* 13452f8f5e76SEric Biggers * If a journal handle was specified, then the encryption context is 13462f8f5e76SEric Biggers * being set on a new inode via inheritance and is part of a larger 13472f8f5e76SEric Biggers * transaction to create the inode. Otherwise the encryption context is 13482f8f5e76SEric Biggers * being set on an existing inode in its own transaction. Only in the 13492f8f5e76SEric Biggers * latter case should the "retry on ENOSPC" logic be used. 13502f8f5e76SEric Biggers */ 13512f8f5e76SEric Biggers 13522f8f5e76SEric Biggers if (handle) { 13532f8f5e76SEric Biggers res = ext4_xattr_set_handle(handle, inode, 13542f8f5e76SEric Biggers EXT4_XATTR_INDEX_ENCRYPTION, 13552f8f5e76SEric Biggers EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 13562f8f5e76SEric Biggers ctx, len, 0); 1357a7550b30SJaegeuk Kim if (!res) { 1358a7550b30SJaegeuk Kim ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1359a7550b30SJaegeuk Kim ext4_clear_inode_state(inode, 1360a7550b30SJaegeuk Kim EXT4_STATE_MAY_INLINE_DATA); 1361a3caa24bSJan Kara /* 13622ee6a576SEric Biggers * Update inode->i_flags - S_ENCRYPTED will be enabled, 13632ee6a576SEric Biggers * S_DAX may be disabled 1364a3caa24bSJan Kara */ 1365043546e4SIra Weiny ext4_set_inode_flags(inode, false); 1366a7550b30SJaegeuk Kim } 1367a7550b30SJaegeuk Kim return res; 1368a7550b30SJaegeuk Kim } 1369a7550b30SJaegeuk Kim 1370b8cb5a54STahsin Erdogan res = dquot_initialize(inode); 1371b8cb5a54STahsin Erdogan if (res) 1372b8cb5a54STahsin Erdogan return res; 13732f8f5e76SEric Biggers retry: 1374af65207cSTahsin Erdogan res = ext4_xattr_set_credits(inode, len, false /* is_create */, 1375af65207cSTahsin Erdogan &credits); 1376dec214d0STahsin Erdogan if (res) 1377dec214d0STahsin Erdogan return res; 1378dec214d0STahsin Erdogan 1379c1a5d5f6STahsin Erdogan handle = ext4_journal_start(inode, EXT4_HT_MISC, credits); 1380a7550b30SJaegeuk Kim if (IS_ERR(handle)) 1381a7550b30SJaegeuk Kim return PTR_ERR(handle); 1382a7550b30SJaegeuk Kim 13832f8f5e76SEric Biggers res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION, 13842f8f5e76SEric Biggers EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 13852f8f5e76SEric Biggers ctx, len, 0); 1386a7550b30SJaegeuk Kim if (!res) { 1387a7550b30SJaegeuk Kim ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 13882ee6a576SEric Biggers /* 13892ee6a576SEric Biggers * Update inode->i_flags - S_ENCRYPTED will be enabled, 13902ee6a576SEric Biggers * S_DAX may be disabled 13912ee6a576SEric Biggers */ 1392043546e4SIra Weiny ext4_set_inode_flags(inode, false); 1393a7550b30SJaegeuk Kim res = ext4_mark_inode_dirty(handle, inode); 1394a7550b30SJaegeuk Kim if (res) 1395a7550b30SJaegeuk Kim EXT4_ERROR_INODE(inode, "Failed to mark inode dirty"); 1396a7550b30SJaegeuk Kim } 1397a7550b30SJaegeuk Kim res2 = ext4_journal_stop(handle); 13982f8f5e76SEric Biggers 13992f8f5e76SEric Biggers if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) 14002f8f5e76SEric Biggers goto retry; 1401a7550b30SJaegeuk Kim if (!res) 1402a7550b30SJaegeuk Kim res = res2; 1403a7550b30SJaegeuk Kim return res; 1404a7550b30SJaegeuk Kim } 1405a7550b30SJaegeuk Kim 1406ed318a6cSEric Biggers static const union fscrypt_context * 1407ed318a6cSEric Biggers ext4_get_dummy_context(struct super_block *sb) 1408a7550b30SJaegeuk Kim { 1409ed318a6cSEric Biggers return EXT4_SB(sb)->s_dummy_enc_ctx.ctx; 1410a7550b30SJaegeuk Kim } 1411a7550b30SJaegeuk Kim 1412b925acb8SEric Biggers static bool ext4_has_stable_inodes(struct super_block *sb) 1413b925acb8SEric Biggers { 1414b925acb8SEric Biggers return ext4_has_feature_stable_inodes(sb); 1415b925acb8SEric Biggers } 1416b925acb8SEric Biggers 1417b925acb8SEric Biggers static void ext4_get_ino_and_lblk_bits(struct super_block *sb, 1418b925acb8SEric Biggers int *ino_bits_ret, int *lblk_bits_ret) 1419b925acb8SEric Biggers { 1420b925acb8SEric Biggers *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count); 1421b925acb8SEric Biggers *lblk_bits_ret = 8 * sizeof(ext4_lblk_t); 1422b925acb8SEric Biggers } 1423b925acb8SEric Biggers 14246f69f0edSEric Biggers static const struct fscrypt_operations ext4_cryptops = { 1425a5d431efSEric Biggers .key_prefix = "ext4:", 1426a7550b30SJaegeuk Kim .get_context = ext4_get_context, 1427a7550b30SJaegeuk Kim .set_context = ext4_set_context, 1428ed318a6cSEric Biggers .get_dummy_context = ext4_get_dummy_context, 1429a7550b30SJaegeuk Kim .empty_dir = ext4_empty_dir, 1430e12ee683SEric Biggers .max_namelen = EXT4_NAME_LEN, 1431b925acb8SEric Biggers .has_stable_inodes = ext4_has_stable_inodes, 1432b925acb8SEric Biggers .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits, 1433a7550b30SJaegeuk Kim }; 1434a7550b30SJaegeuk Kim #endif 1435a7550b30SJaegeuk Kim 1436ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1437d6006186SEric Biggers static const char * const quotatypes[] = INITQFNAMES; 1438689c958cSLi Xi #define QTYPE2NAME(t) (quotatypes[t]) 1439ac27a0ecSDave Kleikamp 1440617ba13bSMingming Cao static int ext4_write_dquot(struct dquot *dquot); 1441617ba13bSMingming Cao static int ext4_acquire_dquot(struct dquot *dquot); 1442617ba13bSMingming Cao static int ext4_release_dquot(struct dquot *dquot); 1443617ba13bSMingming Cao static int ext4_mark_dquot_dirty(struct dquot *dquot); 1444617ba13bSMingming Cao static int ext4_write_info(struct super_block *sb, int type); 14456f28e087SJan Kara static int ext4_quota_on(struct super_block *sb, int type, int format_id, 14468c54ca9cSAl Viro const struct path *path); 1447617ba13bSMingming Cao static int ext4_quota_on_mount(struct super_block *sb, int type); 1448617ba13bSMingming Cao static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1449ac27a0ecSDave Kleikamp size_t len, loff_t off); 1450617ba13bSMingming Cao static ssize_t ext4_quota_write(struct super_block *sb, int type, 1451ac27a0ecSDave Kleikamp const char *data, size_t len, loff_t off); 14527c319d32SAditya Kali static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 14537c319d32SAditya Kali unsigned int flags); 14547c319d32SAditya Kali static int ext4_enable_quotas(struct super_block *sb); 1455ac27a0ecSDave Kleikamp 145696c7e0d9SJan Kara static struct dquot **ext4_get_dquots(struct inode *inode) 145796c7e0d9SJan Kara { 145896c7e0d9SJan Kara return EXT4_I(inode)->i_dquot; 145996c7e0d9SJan Kara } 146096c7e0d9SJan Kara 146161e225dcSAlexey Dobriyan static const struct dquot_operations ext4_quota_operations = { 146260e58e0fSMingming Cao .get_reserved_space = ext4_get_reserved_space, 1463617ba13bSMingming Cao .write_dquot = ext4_write_dquot, 1464617ba13bSMingming Cao .acquire_dquot = ext4_acquire_dquot, 1465617ba13bSMingming Cao .release_dquot = ext4_release_dquot, 1466617ba13bSMingming Cao .mark_dirty = ext4_mark_dquot_dirty, 1467a5b5ee32SJan Kara .write_info = ext4_write_info, 1468a5b5ee32SJan Kara .alloc_dquot = dquot_alloc, 1469a5b5ee32SJan Kara .destroy_dquot = dquot_destroy, 1470040cb378SLi Xi .get_projid = ext4_get_projid, 14717a9ca53aSTahsin Erdogan .get_inode_usage = ext4_get_inode_usage, 1472ebc11f7bSChengguang Xu .get_next_id = dquot_get_next_id, 1473ac27a0ecSDave Kleikamp }; 1474ac27a0ecSDave Kleikamp 14750d54b217SAlexey Dobriyan static const struct quotactl_ops ext4_qctl_operations = { 1476617ba13bSMingming Cao .quota_on = ext4_quota_on, 1477ca0e05e4SDmitry Monakhov .quota_off = ext4_quota_off, 1478287a8095SChristoph Hellwig .quota_sync = dquot_quota_sync, 14790a240339SJan Kara .get_state = dquot_get_state, 1480287a8095SChristoph Hellwig .set_info = dquot_set_dqinfo, 1481287a8095SChristoph Hellwig .get_dqblk = dquot_get_dqblk, 14826332b9b5SEric Sandeen .set_dqblk = dquot_set_dqblk, 14836332b9b5SEric Sandeen .get_nextdqblk = dquot_get_next_dqblk, 1484ac27a0ecSDave Kleikamp }; 1485ac27a0ecSDave Kleikamp #endif 1486ac27a0ecSDave Kleikamp 1487ee9b6d61SJosef 'Jeff' Sipek static const struct super_operations ext4_sops = { 1488617ba13bSMingming Cao .alloc_inode = ext4_alloc_inode, 148994053139SAl Viro .free_inode = ext4_free_in_core_inode, 1490617ba13bSMingming Cao .destroy_inode = ext4_destroy_inode, 1491617ba13bSMingming Cao .write_inode = ext4_write_inode, 1492617ba13bSMingming Cao .dirty_inode = ext4_dirty_inode, 14937ff9c073STheodore Ts'o .drop_inode = ext4_drop_inode, 14940930fcc1SAl Viro .evict_inode = ext4_evict_inode, 1495617ba13bSMingming Cao .put_super = ext4_put_super, 1496617ba13bSMingming Cao .sync_fs = ext4_sync_fs, 1497c4be0c1dSTakashi Sato .freeze_fs = ext4_freeze, 1498c4be0c1dSTakashi Sato .unfreeze_fs = ext4_unfreeze, 1499617ba13bSMingming Cao .statfs = ext4_statfs, 1500617ba13bSMingming Cao .remount_fs = ext4_remount, 1501617ba13bSMingming Cao .show_options = ext4_show_options, 1502ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1503617ba13bSMingming Cao .quota_read = ext4_quota_read, 1504617ba13bSMingming Cao .quota_write = ext4_quota_write, 150596c7e0d9SJan Kara .get_dquots = ext4_get_dquots, 1506ac27a0ecSDave Kleikamp #endif 1507c39a7f84SToshiyuki Okajima .bdev_try_to_free_page = bdev_try_to_free_page, 1508ac27a0ecSDave Kleikamp }; 1509ac27a0ecSDave Kleikamp 151039655164SChristoph Hellwig static const struct export_operations ext4_export_ops = { 15111b961ac0SChristoph Hellwig .fh_to_dentry = ext4_fh_to_dentry, 15121b961ac0SChristoph Hellwig .fh_to_parent = ext4_fh_to_parent, 1513617ba13bSMingming Cao .get_parent = ext4_get_parent, 1514fde87268STheodore Ts'o .commit_metadata = ext4_nfs_commit_metadata, 1515ac27a0ecSDave Kleikamp }; 1516ac27a0ecSDave Kleikamp 1517ac27a0ecSDave Kleikamp enum { 1518ac27a0ecSDave Kleikamp Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1519ac27a0ecSDave Kleikamp Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 152072578c33STheodore Ts'o Opt_nouid32, Opt_debug, Opt_removed, 1521ac27a0ecSDave Kleikamp Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 152272578c33STheodore Ts'o Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1523ad4eec61SEric Sandeen Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1524ad4eec61SEric Sandeen Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1525ac27a0ecSDave Kleikamp Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 15266ddb2447STheodore Ts'o Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 15274f74d15fSEric Biggers Opt_inlinecrypt, 1528ac27a0ecSDave Kleikamp Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 15295a20bdfcSJan Kara Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1530ee4a3fcdSTheodore Ts'o Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 15319cb20f94SIra Weiny Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, 15329cb20f94SIra Weiny Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never, 1533327eaf73STheodore Ts'o Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1534327eaf73STheodore Ts'o Opt_nowarn_on_error, Opt_mblk_io_submit, 1535670e9875STheodore Ts'o Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize, 15361449032bSTheodore Ts'o Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 15375328e635SEric Sandeen Opt_inode_readahead_blks, Opt_journal_ioprio, 1538744692dcSJiaying Zhang Opt_dioread_nolock, Opt_dioread_lock, 1539fc6cb1cdSTheodore Ts'o Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1540cdb7ee4cSTahsin Erdogan Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 15413d392b26STheodore Ts'o Opt_prefetch_block_bitmaps, 1542ac27a0ecSDave Kleikamp }; 1543ac27a0ecSDave Kleikamp 1544a447c093SSteven Whitehouse static const match_table_t tokens = { 1545ac27a0ecSDave Kleikamp {Opt_bsd_df, "bsddf"}, 1546ac27a0ecSDave Kleikamp {Opt_minix_df, "minixdf"}, 1547ac27a0ecSDave Kleikamp {Opt_grpid, "grpid"}, 1548ac27a0ecSDave Kleikamp {Opt_grpid, "bsdgroups"}, 1549ac27a0ecSDave Kleikamp {Opt_nogrpid, "nogrpid"}, 1550ac27a0ecSDave Kleikamp {Opt_nogrpid, "sysvgroups"}, 1551ac27a0ecSDave Kleikamp {Opt_resgid, "resgid=%u"}, 1552ac27a0ecSDave Kleikamp {Opt_resuid, "resuid=%u"}, 1553ac27a0ecSDave Kleikamp {Opt_sb, "sb=%u"}, 1554ac27a0ecSDave Kleikamp {Opt_err_cont, "errors=continue"}, 1555ac27a0ecSDave Kleikamp {Opt_err_panic, "errors=panic"}, 1556ac27a0ecSDave Kleikamp {Opt_err_ro, "errors=remount-ro"}, 1557ac27a0ecSDave Kleikamp {Opt_nouid32, "nouid32"}, 1558ac27a0ecSDave Kleikamp {Opt_debug, "debug"}, 155972578c33STheodore Ts'o {Opt_removed, "oldalloc"}, 156072578c33STheodore Ts'o {Opt_removed, "orlov"}, 1561ac27a0ecSDave Kleikamp {Opt_user_xattr, "user_xattr"}, 1562ac27a0ecSDave Kleikamp {Opt_nouser_xattr, "nouser_xattr"}, 1563ac27a0ecSDave Kleikamp {Opt_acl, "acl"}, 1564ac27a0ecSDave Kleikamp {Opt_noacl, "noacl"}, 1565e3bb52aeSEric Sandeen {Opt_noload, "norecovery"}, 15665a916be1STheodore Ts'o {Opt_noload, "noload"}, 156772578c33STheodore Ts'o {Opt_removed, "nobh"}, 156872578c33STheodore Ts'o {Opt_removed, "bh"}, 1569ac27a0ecSDave Kleikamp {Opt_commit, "commit=%u"}, 157030773840STheodore Ts'o {Opt_min_batch_time, "min_batch_time=%u"}, 157130773840STheodore Ts'o {Opt_max_batch_time, "max_batch_time=%u"}, 1572ac27a0ecSDave Kleikamp {Opt_journal_dev, "journal_dev=%u"}, 1573ad4eec61SEric Sandeen {Opt_journal_path, "journal_path=%s"}, 1574818d276cSGirish Shilamkar {Opt_journal_checksum, "journal_checksum"}, 1575c6d3d56dSDarrick J. Wong {Opt_nojournal_checksum, "nojournal_checksum"}, 1576818d276cSGirish Shilamkar {Opt_journal_async_commit, "journal_async_commit"}, 1577ac27a0ecSDave Kleikamp {Opt_abort, "abort"}, 1578ac27a0ecSDave Kleikamp {Opt_data_journal, "data=journal"}, 1579ac27a0ecSDave Kleikamp {Opt_data_ordered, "data=ordered"}, 1580ac27a0ecSDave Kleikamp {Opt_data_writeback, "data=writeback"}, 15815bf5683aSHidehiro Kawai {Opt_data_err_abort, "data_err=abort"}, 15825bf5683aSHidehiro Kawai {Opt_data_err_ignore, "data_err=ignore"}, 1583ac27a0ecSDave Kleikamp {Opt_offusrjquota, "usrjquota="}, 1584ac27a0ecSDave Kleikamp {Opt_usrjquota, "usrjquota=%s"}, 1585ac27a0ecSDave Kleikamp {Opt_offgrpjquota, "grpjquota="}, 1586ac27a0ecSDave Kleikamp {Opt_grpjquota, "grpjquota=%s"}, 1587ac27a0ecSDave Kleikamp {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1588ac27a0ecSDave Kleikamp {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 15895a20bdfcSJan Kara {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1590ac27a0ecSDave Kleikamp {Opt_grpquota, "grpquota"}, 1591ac27a0ecSDave Kleikamp {Opt_noquota, "noquota"}, 1592ac27a0ecSDave Kleikamp {Opt_quota, "quota"}, 1593ac27a0ecSDave Kleikamp {Opt_usrquota, "usrquota"}, 159449da9392SJan Kara {Opt_prjquota, "prjquota"}, 1595ac27a0ecSDave Kleikamp {Opt_barrier, "barrier=%u"}, 159606705bffSTheodore Ts'o {Opt_barrier, "barrier"}, 159706705bffSTheodore Ts'o {Opt_nobarrier, "nobarrier"}, 159825ec56b5SJean Noel Cordenner {Opt_i_version, "i_version"}, 1599923ae0ffSRoss Zwisler {Opt_dax, "dax"}, 16009cb20f94SIra Weiny {Opt_dax_always, "dax=always"}, 16019cb20f94SIra Weiny {Opt_dax_inode, "dax=inode"}, 16029cb20f94SIra Weiny {Opt_dax_never, "dax=never"}, 1603c9de560dSAlex Tomas {Opt_stripe, "stripe=%u"}, 160464769240SAlex Tomas {Opt_delalloc, "delalloc"}, 1605327eaf73STheodore Ts'o {Opt_warn_on_error, "warn_on_error"}, 1606327eaf73STheodore Ts'o {Opt_nowarn_on_error, "nowarn_on_error"}, 1607a26f4992STheodore Ts'o {Opt_lazytime, "lazytime"}, 1608a26f4992STheodore Ts'o {Opt_nolazytime, "nolazytime"}, 1609670e9875STheodore Ts'o {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"}, 1610dd919b98SAneesh Kumar K.V {Opt_nodelalloc, "nodelalloc"}, 161136ade451SJan Kara {Opt_removed, "mblk_io_submit"}, 161236ade451SJan Kara {Opt_removed, "nomblk_io_submit"}, 16136fd058f7STheodore Ts'o {Opt_block_validity, "block_validity"}, 16146fd058f7STheodore Ts'o {Opt_noblock_validity, "noblock_validity"}, 1615240799cdSTheodore Ts'o {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1616b3881f74STheodore Ts'o {Opt_journal_ioprio, "journal_ioprio=%u"}, 1617afd4672dSTheodore Ts'o {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 161806705bffSTheodore Ts'o {Opt_auto_da_alloc, "auto_da_alloc"}, 161906705bffSTheodore Ts'o {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1620744692dcSJiaying Zhang {Opt_dioread_nolock, "dioread_nolock"}, 1621244adf64STheodore Ts'o {Opt_dioread_lock, "nodioread_nolock"}, 1622744692dcSJiaying Zhang {Opt_dioread_lock, "dioread_lock"}, 16235328e635SEric Sandeen {Opt_discard, "discard"}, 16245328e635SEric Sandeen {Opt_nodiscard, "nodiscard"}, 1625fc6cb1cdSTheodore Ts'o {Opt_init_itable, "init_itable=%u"}, 1626fc6cb1cdSTheodore Ts'o {Opt_init_itable, "init_itable"}, 1627fc6cb1cdSTheodore Ts'o {Opt_noinit_itable, "noinit_itable"}, 1628df981d03STheodore Ts'o {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 1629ed318a6cSEric Biggers {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 16306ddb2447STheodore Ts'o {Opt_test_dummy_encryption, "test_dummy_encryption"}, 16314f74d15fSEric Biggers {Opt_inlinecrypt, "inlinecrypt"}, 1632cdb7ee4cSTahsin Erdogan {Opt_nombcache, "nombcache"}, 1633cdb7ee4cSTahsin Erdogan {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */ 16343d392b26STheodore Ts'o {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"}, 1635c7198b9cSTheodore Ts'o {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1636c7198b9cSTheodore Ts'o {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1637c7198b9cSTheodore Ts'o {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1638c7198b9cSTheodore Ts'o {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1639c7198b9cSTheodore Ts'o {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1640f3f12faaSJosef Bacik {Opt_err, NULL}, 1641ac27a0ecSDave Kleikamp }; 1642ac27a0ecSDave Kleikamp 1643617ba13bSMingming Cao static ext4_fsblk_t get_sb_block(void **data) 1644ac27a0ecSDave Kleikamp { 1645617ba13bSMingming Cao ext4_fsblk_t sb_block; 1646ac27a0ecSDave Kleikamp char *options = (char *) *data; 1647ac27a0ecSDave Kleikamp 1648ac27a0ecSDave Kleikamp if (!options || strncmp(options, "sb=", 3) != 0) 1649ac27a0ecSDave Kleikamp return 1; /* Default location */ 16500b8e58a1SAndreas Dilger 1651ac27a0ecSDave Kleikamp options += 3; 16520b8e58a1SAndreas Dilger /* TODO: use simple_strtoll with >32bit ext4 */ 1653ac27a0ecSDave Kleikamp sb_block = simple_strtoul(options, &options, 0); 1654ac27a0ecSDave Kleikamp if (*options && *options != ',') { 16554776004fSTheodore Ts'o printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1656ac27a0ecSDave Kleikamp (char *) *data); 1657ac27a0ecSDave Kleikamp return 1; 1658ac27a0ecSDave Kleikamp } 1659ac27a0ecSDave Kleikamp if (*options == ',') 1660ac27a0ecSDave Kleikamp options++; 1661ac27a0ecSDave Kleikamp *data = (void *) options; 16620b8e58a1SAndreas Dilger 1663ac27a0ecSDave Kleikamp return sb_block; 1664ac27a0ecSDave Kleikamp } 1665ac27a0ecSDave Kleikamp 1666b3881f74STheodore Ts'o #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1667d6006186SEric Biggers static const char deprecated_msg[] = 1668d6006186SEric Biggers "Mount option \"%s\" will be removed by %s\n" 1669437ca0fdSDmitry Monakhov "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1670b3881f74STheodore Ts'o 167156c50f11SDmitry Monakhov #ifdef CONFIG_QUOTA 167256c50f11SDmitry Monakhov static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 167356c50f11SDmitry Monakhov { 167456c50f11SDmitry Monakhov struct ext4_sb_info *sbi = EXT4_SB(sb); 167533458eabSTheodore Ts'o char *qname, *old_qname = get_qf_name(sb, sbi, qtype); 167603dafb5fSChen Gang int ret = -1; 167756c50f11SDmitry Monakhov 167833458eabSTheodore Ts'o if (sb_any_quota_loaded(sb) && !old_qname) { 167956c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 168056c50f11SDmitry Monakhov "Cannot change journaled " 168156c50f11SDmitry Monakhov "quota options when quota turned on"); 168257f73c2cSTheodore Ts'o return -1; 168356c50f11SDmitry Monakhov } 1684e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) { 1685c325a67cSTheodore Ts'o ext4_msg(sb, KERN_INFO, "Journaled quota options " 1686c325a67cSTheodore Ts'o "ignored when QUOTA feature is enabled"); 1687c325a67cSTheodore Ts'o return 1; 1688262b4662SJan Kara } 168956c50f11SDmitry Monakhov qname = match_strdup(args); 169056c50f11SDmitry Monakhov if (!qname) { 169156c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 169256c50f11SDmitry Monakhov "Not enough memory for storing quotafile name"); 169357f73c2cSTheodore Ts'o return -1; 169456c50f11SDmitry Monakhov } 169533458eabSTheodore Ts'o if (old_qname) { 169633458eabSTheodore Ts'o if (strcmp(old_qname, qname) == 0) 169703dafb5fSChen Gang ret = 1; 169803dafb5fSChen Gang else 169956c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 170003dafb5fSChen Gang "%s quota file already specified", 170103dafb5fSChen Gang QTYPE2NAME(qtype)); 170203dafb5fSChen Gang goto errout; 170356c50f11SDmitry Monakhov } 170403dafb5fSChen Gang if (strchr(qname, '/')) { 170556c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 170656c50f11SDmitry Monakhov "quotafile must be on filesystem root"); 170703dafb5fSChen Gang goto errout; 170856c50f11SDmitry Monakhov } 170933458eabSTheodore Ts'o rcu_assign_pointer(sbi->s_qf_names[qtype], qname); 1710fd8c37ecSTheodore Ts'o set_opt(sb, QUOTA); 171156c50f11SDmitry Monakhov return 1; 171203dafb5fSChen Gang errout: 171303dafb5fSChen Gang kfree(qname); 171403dafb5fSChen Gang return ret; 171556c50f11SDmitry Monakhov } 171656c50f11SDmitry Monakhov 171756c50f11SDmitry Monakhov static int clear_qf_name(struct super_block *sb, int qtype) 171856c50f11SDmitry Monakhov { 171956c50f11SDmitry Monakhov 172056c50f11SDmitry Monakhov struct ext4_sb_info *sbi = EXT4_SB(sb); 172133458eabSTheodore Ts'o char *old_qname = get_qf_name(sb, sbi, qtype); 172256c50f11SDmitry Monakhov 172333458eabSTheodore Ts'o if (sb_any_quota_loaded(sb) && old_qname) { 172456c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 172556c50f11SDmitry Monakhov " when quota turned on"); 172657f73c2cSTheodore Ts'o return -1; 172756c50f11SDmitry Monakhov } 172833458eabSTheodore Ts'o rcu_assign_pointer(sbi->s_qf_names[qtype], NULL); 172933458eabSTheodore Ts'o synchronize_rcu(); 173033458eabSTheodore Ts'o kfree(old_qname); 173156c50f11SDmitry Monakhov return 1; 173256c50f11SDmitry Monakhov } 173356c50f11SDmitry Monakhov #endif 173456c50f11SDmitry Monakhov 173526092bf5STheodore Ts'o #define MOPT_SET 0x0001 173626092bf5STheodore Ts'o #define MOPT_CLEAR 0x0002 173726092bf5STheodore Ts'o #define MOPT_NOSUPPORT 0x0004 173826092bf5STheodore Ts'o #define MOPT_EXPLICIT 0x0008 173926092bf5STheodore Ts'o #define MOPT_CLEAR_ERR 0x0010 174026092bf5STheodore Ts'o #define MOPT_GTE0 0x0020 174126092bf5STheodore Ts'o #ifdef CONFIG_QUOTA 174226092bf5STheodore Ts'o #define MOPT_Q 0 174326092bf5STheodore Ts'o #define MOPT_QFMT 0x0040 174426092bf5STheodore Ts'o #else 174526092bf5STheodore Ts'o #define MOPT_Q MOPT_NOSUPPORT 174626092bf5STheodore Ts'o #define MOPT_QFMT MOPT_NOSUPPORT 174726092bf5STheodore Ts'o #endif 174826092bf5STheodore Ts'o #define MOPT_DATAJ 0x0080 17498dc0aa8cSTheodore Ts'o #define MOPT_NO_EXT2 0x0100 17508dc0aa8cSTheodore Ts'o #define MOPT_NO_EXT3 0x0200 17518dc0aa8cSTheodore Ts'o #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1752ad4eec61SEric Sandeen #define MOPT_STRING 0x0400 17539cb20f94SIra Weiny #define MOPT_SKIP 0x0800 175426092bf5STheodore Ts'o 175526092bf5STheodore Ts'o static const struct mount_opts { 175626092bf5STheodore Ts'o int token; 175726092bf5STheodore Ts'o int mount_opt; 175826092bf5STheodore Ts'o int flags; 175926092bf5STheodore Ts'o } ext4_mount_opts[] = { 176026092bf5STheodore Ts'o {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 176126092bf5STheodore Ts'o {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 176226092bf5STheodore Ts'o {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 176326092bf5STheodore Ts'o {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 176426092bf5STheodore Ts'o {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 176526092bf5STheodore Ts'o {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 17668dc0aa8cSTheodore Ts'o {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 17678dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_SET}, 17688dc0aa8cSTheodore Ts'o {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 17698dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_CLEAR}, 177026092bf5STheodore Ts'o {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 177126092bf5STheodore Ts'o {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 17728dc0aa8cSTheodore Ts'o {Opt_delalloc, EXT4_MOUNT_DELALLOC, 17738dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 17748dc0aa8cSTheodore Ts'o {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 177559d9fa5cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_CLEAR}, 1776327eaf73STheodore Ts'o {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1777327eaf73STheodore Ts'o {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1778c6d3d56dSDarrick J. Wong {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1779c6d3d56dSDarrick J. Wong MOPT_EXT4_ONLY | MOPT_CLEAR}, 17808dc0aa8cSTheodore Ts'o {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 17811e381f60SDmitry Monakhov MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 178226092bf5STheodore Ts'o {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 17838dc0aa8cSTheodore Ts'o EXT4_MOUNT_JOURNAL_CHECKSUM), 17841e381f60SDmitry Monakhov MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 17858dc0aa8cSTheodore Ts'o {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 178626092bf5STheodore Ts'o {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 178726092bf5STheodore Ts'o {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 178826092bf5STheodore Ts'o {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 17898dc0aa8cSTheodore Ts'o {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 17907915a861SAles Novak MOPT_NO_EXT2}, 17918dc0aa8cSTheodore Ts'o {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 17927915a861SAles Novak MOPT_NO_EXT2}, 179326092bf5STheodore Ts'o {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 179426092bf5STheodore Ts'o {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 179526092bf5STheodore Ts'o {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 179626092bf5STheodore Ts'o {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 179726092bf5STheodore Ts'o {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 179826092bf5STheodore Ts'o {Opt_commit, 0, MOPT_GTE0}, 179926092bf5STheodore Ts'o {Opt_max_batch_time, 0, MOPT_GTE0}, 180026092bf5STheodore Ts'o {Opt_min_batch_time, 0, MOPT_GTE0}, 180126092bf5STheodore Ts'o {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 180226092bf5STheodore Ts'o {Opt_init_itable, 0, MOPT_GTE0}, 18039cb20f94SIra Weiny {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP}, 18049cb20f94SIra Weiny {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS, 18059cb20f94SIra Weiny MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 18069cb20f94SIra Weiny {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE, 18079cb20f94SIra Weiny MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 18089cb20f94SIra Weiny {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER, 18099cb20f94SIra Weiny MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 181026092bf5STheodore Ts'o {Opt_stripe, 0, MOPT_GTE0}, 18110efb3b23SJan Kara {Opt_resuid, 0, MOPT_GTE0}, 18120efb3b23SJan Kara {Opt_resgid, 0, MOPT_GTE0}, 18135ba92bcfSCarlos Maiolino {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 18145ba92bcfSCarlos Maiolino {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING}, 18155ba92bcfSCarlos Maiolino {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 18168dc0aa8cSTheodore Ts'o {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 18178dc0aa8cSTheodore Ts'o {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 18188dc0aa8cSTheodore Ts'o {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 18198dc0aa8cSTheodore Ts'o MOPT_NO_EXT2 | MOPT_DATAJ}, 182026092bf5STheodore Ts'o {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 182126092bf5STheodore Ts'o {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 182226092bf5STheodore Ts'o #ifdef CONFIG_EXT4_FS_POSIX_ACL 182326092bf5STheodore Ts'o {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 182426092bf5STheodore Ts'o {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 182526092bf5STheodore Ts'o #else 182626092bf5STheodore Ts'o {Opt_acl, 0, MOPT_NOSUPPORT}, 182726092bf5STheodore Ts'o {Opt_noacl, 0, MOPT_NOSUPPORT}, 182826092bf5STheodore Ts'o #endif 182926092bf5STheodore Ts'o {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 183026092bf5STheodore Ts'o {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1831670e9875STheodore Ts'o {Opt_debug_want_extra_isize, 0, MOPT_GTE0}, 183226092bf5STheodore Ts'o {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 183326092bf5STheodore Ts'o {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 183426092bf5STheodore Ts'o MOPT_SET | MOPT_Q}, 183526092bf5STheodore Ts'o {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 183626092bf5STheodore Ts'o MOPT_SET | MOPT_Q}, 183749da9392SJan Kara {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 183849da9392SJan Kara MOPT_SET | MOPT_Q}, 183926092bf5STheodore Ts'o {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 184049da9392SJan Kara EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 184149da9392SJan Kara MOPT_CLEAR | MOPT_Q}, 184226092bf5STheodore Ts'o {Opt_usrjquota, 0, MOPT_Q}, 184326092bf5STheodore Ts'o {Opt_grpjquota, 0, MOPT_Q}, 184426092bf5STheodore Ts'o {Opt_offusrjquota, 0, MOPT_Q}, 184526092bf5STheodore Ts'o {Opt_offgrpjquota, 0, MOPT_Q}, 184626092bf5STheodore Ts'o {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 184726092bf5STheodore Ts'o {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 184826092bf5STheodore Ts'o {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 1849df981d03STheodore Ts'o {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 1850ed318a6cSEric Biggers {Opt_test_dummy_encryption, 0, MOPT_STRING}, 1851cdb7ee4cSTahsin Erdogan {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 18523d392b26STheodore Ts'o {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS, 18533d392b26STheodore Ts'o MOPT_SET}, 185426092bf5STheodore Ts'o {Opt_err, 0, 0} 185526092bf5STheodore Ts'o }; 185626092bf5STheodore Ts'o 1857c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 1858c83ad55eSGabriel Krisman Bertazi static const struct ext4_sb_encodings { 1859c83ad55eSGabriel Krisman Bertazi __u16 magic; 1860c83ad55eSGabriel Krisman Bertazi char *name; 1861c83ad55eSGabriel Krisman Bertazi char *version; 1862c83ad55eSGabriel Krisman Bertazi } ext4_sb_encoding_map[] = { 1863c83ad55eSGabriel Krisman Bertazi {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"}, 1864c83ad55eSGabriel Krisman Bertazi }; 1865c83ad55eSGabriel Krisman Bertazi 1866c83ad55eSGabriel Krisman Bertazi static int ext4_sb_read_encoding(const struct ext4_super_block *es, 1867c83ad55eSGabriel Krisman Bertazi const struct ext4_sb_encodings **encoding, 1868c83ad55eSGabriel Krisman Bertazi __u16 *flags) 1869c83ad55eSGabriel Krisman Bertazi { 1870c83ad55eSGabriel Krisman Bertazi __u16 magic = le16_to_cpu(es->s_encoding); 1871c83ad55eSGabriel Krisman Bertazi int i; 1872c83ad55eSGabriel Krisman Bertazi 1873c83ad55eSGabriel Krisman Bertazi for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) 1874c83ad55eSGabriel Krisman Bertazi if (magic == ext4_sb_encoding_map[i].magic) 1875c83ad55eSGabriel Krisman Bertazi break; 1876c83ad55eSGabriel Krisman Bertazi 1877c83ad55eSGabriel Krisman Bertazi if (i >= ARRAY_SIZE(ext4_sb_encoding_map)) 1878c83ad55eSGabriel Krisman Bertazi return -EINVAL; 1879c83ad55eSGabriel Krisman Bertazi 1880c83ad55eSGabriel Krisman Bertazi *encoding = &ext4_sb_encoding_map[i]; 1881c83ad55eSGabriel Krisman Bertazi *flags = le16_to_cpu(es->s_encoding_flags); 1882c83ad55eSGabriel Krisman Bertazi 1883c83ad55eSGabriel Krisman Bertazi return 0; 1884c83ad55eSGabriel Krisman Bertazi } 1885c83ad55eSGabriel Krisman Bertazi #endif 1886c83ad55eSGabriel Krisman Bertazi 1887ed318a6cSEric Biggers static int ext4_set_test_dummy_encryption(struct super_block *sb, 1888ed318a6cSEric Biggers const char *opt, 1889ed318a6cSEric Biggers const substring_t *arg, 1890ed318a6cSEric Biggers bool is_remount) 1891ed318a6cSEric Biggers { 1892ed318a6cSEric Biggers #ifdef CONFIG_FS_ENCRYPTION 1893ed318a6cSEric Biggers struct ext4_sb_info *sbi = EXT4_SB(sb); 1894ed318a6cSEric Biggers int err; 1895ed318a6cSEric Biggers 1896ed318a6cSEric Biggers /* 1897ed318a6cSEric Biggers * This mount option is just for testing, and it's not worthwhile to 1898ed318a6cSEric Biggers * implement the extra complexity (e.g. RCU protection) that would be 1899ed318a6cSEric Biggers * needed to allow it to be set or changed during remount. We do allow 1900ed318a6cSEric Biggers * it to be specified during remount, but only if there is no change. 1901ed318a6cSEric Biggers */ 1902ed318a6cSEric Biggers if (is_remount && !sbi->s_dummy_enc_ctx.ctx) { 1903ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 1904ed318a6cSEric Biggers "Can't set test_dummy_encryption on remount"); 1905ed318a6cSEric Biggers return -1; 1906ed318a6cSEric Biggers } 1907ed318a6cSEric Biggers err = fscrypt_set_test_dummy_encryption(sb, arg, &sbi->s_dummy_enc_ctx); 1908ed318a6cSEric Biggers if (err) { 1909ed318a6cSEric Biggers if (err == -EEXIST) 1910ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 1911ed318a6cSEric Biggers "Can't change test_dummy_encryption on remount"); 1912ed318a6cSEric Biggers else if (err == -EINVAL) 1913ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 1914ed318a6cSEric Biggers "Value of option \"%s\" is unrecognized", opt); 1915ed318a6cSEric Biggers else 1916ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 1917ed318a6cSEric Biggers "Error processing option \"%s\" [%d]", 1918ed318a6cSEric Biggers opt, err); 1919ed318a6cSEric Biggers return -1; 1920ed318a6cSEric Biggers } 1921ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled"); 1922ed318a6cSEric Biggers #else 1923ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 1924ed318a6cSEric Biggers "Test dummy encryption mount option ignored"); 1925ed318a6cSEric Biggers #endif 1926ed318a6cSEric Biggers return 1; 1927ed318a6cSEric Biggers } 1928ed318a6cSEric Biggers 192926092bf5STheodore Ts'o static int handle_mount_opt(struct super_block *sb, char *opt, int token, 193026092bf5STheodore Ts'o substring_t *args, unsigned long *journal_devnum, 193126092bf5STheodore Ts'o unsigned int *journal_ioprio, int is_remount) 193226092bf5STheodore Ts'o { 193326092bf5STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 193426092bf5STheodore Ts'o const struct mount_opts *m; 193508cefc7aSEric W. Biederman kuid_t uid; 193608cefc7aSEric W. Biederman kgid_t gid; 193726092bf5STheodore Ts'o int arg = 0; 193826092bf5STheodore Ts'o 193957f73c2cSTheodore Ts'o #ifdef CONFIG_QUOTA 194057f73c2cSTheodore Ts'o if (token == Opt_usrjquota) 194157f73c2cSTheodore Ts'o return set_qf_name(sb, USRQUOTA, &args[0]); 194257f73c2cSTheodore Ts'o else if (token == Opt_grpjquota) 194357f73c2cSTheodore Ts'o return set_qf_name(sb, GRPQUOTA, &args[0]); 194457f73c2cSTheodore Ts'o else if (token == Opt_offusrjquota) 194557f73c2cSTheodore Ts'o return clear_qf_name(sb, USRQUOTA); 194657f73c2cSTheodore Ts'o else if (token == Opt_offgrpjquota) 194757f73c2cSTheodore Ts'o return clear_qf_name(sb, GRPQUOTA); 194857f73c2cSTheodore Ts'o #endif 194926092bf5STheodore Ts'o switch (token) { 1950f7048605STheodore Ts'o case Opt_noacl: 1951f7048605STheodore Ts'o case Opt_nouser_xattr: 1952f7048605STheodore Ts'o ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 1953f7048605STheodore Ts'o break; 195426092bf5STheodore Ts'o case Opt_sb: 195526092bf5STheodore Ts'o return 1; /* handled by get_sb_block() */ 195626092bf5STheodore Ts'o case Opt_removed: 19575f3633e3SJan Kara ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 195826092bf5STheodore Ts'o return 1; 195926092bf5STheodore Ts'o case Opt_abort: 196026092bf5STheodore Ts'o sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 196126092bf5STheodore Ts'o return 1; 196226092bf5STheodore Ts'o case Opt_i_version: 1963357fdad0SMatthew Garrett sb->s_flags |= SB_I_VERSION; 196426092bf5STheodore Ts'o return 1; 1965a26f4992STheodore Ts'o case Opt_lazytime: 19661751e8a6SLinus Torvalds sb->s_flags |= SB_LAZYTIME; 1967a26f4992STheodore Ts'o return 1; 1968a26f4992STheodore Ts'o case Opt_nolazytime: 19691751e8a6SLinus Torvalds sb->s_flags &= ~SB_LAZYTIME; 1970a26f4992STheodore Ts'o return 1; 19714f74d15fSEric Biggers case Opt_inlinecrypt: 19724f74d15fSEric Biggers #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 19734f74d15fSEric Biggers sb->s_flags |= SB_INLINECRYPT; 19744f74d15fSEric Biggers #else 19754f74d15fSEric Biggers ext4_msg(sb, KERN_ERR, "inline encryption not supported"); 19764f74d15fSEric Biggers #endif 19774f74d15fSEric Biggers return 1; 197826092bf5STheodore Ts'o } 197926092bf5STheodore Ts'o 19805f3633e3SJan Kara for (m = ext4_mount_opts; m->token != Opt_err; m++) 19815f3633e3SJan Kara if (token == m->token) 19825f3633e3SJan Kara break; 19835f3633e3SJan Kara 19845f3633e3SJan Kara if (m->token == Opt_err) { 19855f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 19865f3633e3SJan Kara "or missing value", opt); 19875f3633e3SJan Kara return -1; 19885f3633e3SJan Kara } 19895f3633e3SJan Kara 19908dc0aa8cSTheodore Ts'o if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 19918dc0aa8cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 19928dc0aa8cSTheodore Ts'o "Mount option \"%s\" incompatible with ext2", opt); 19938dc0aa8cSTheodore Ts'o return -1; 19948dc0aa8cSTheodore Ts'o } 19958dc0aa8cSTheodore Ts'o if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 19968dc0aa8cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 19978dc0aa8cSTheodore Ts'o "Mount option \"%s\" incompatible with ext3", opt); 19988dc0aa8cSTheodore Ts'o return -1; 19998dc0aa8cSTheodore Ts'o } 20008dc0aa8cSTheodore Ts'o 2001ad4eec61SEric Sandeen if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) 20020efb3b23SJan Kara return -1; 200326092bf5STheodore Ts'o if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 200426092bf5STheodore Ts'o return -1; 2005c93cf2d7SDmitry Monakhov if (m->flags & MOPT_EXPLICIT) { 2006c93cf2d7SDmitry Monakhov if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 200726092bf5STheodore Ts'o set_opt2(sb, EXPLICIT_DELALLOC); 20081e381f60SDmitry Monakhov } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 20091e381f60SDmitry Monakhov set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM); 2010c93cf2d7SDmitry Monakhov } else 2011c93cf2d7SDmitry Monakhov return -1; 2012c93cf2d7SDmitry Monakhov } 201326092bf5STheodore Ts'o if (m->flags & MOPT_CLEAR_ERR) 201426092bf5STheodore Ts'o clear_opt(sb, ERRORS_MASK); 201526092bf5STheodore Ts'o if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 201626092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "Cannot change quota " 201726092bf5STheodore Ts'o "options when quota turned on"); 201826092bf5STheodore Ts'o return -1; 201926092bf5STheodore Ts'o } 202026092bf5STheodore Ts'o 202126092bf5STheodore Ts'o if (m->flags & MOPT_NOSUPPORT) { 202226092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 202326092bf5STheodore Ts'o } else if (token == Opt_commit) { 202426092bf5STheodore Ts'o if (arg == 0) 202526092bf5STheodore Ts'o arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 20269ba55543Szhangyi (F) else if (arg > INT_MAX / HZ) { 20279ba55543Szhangyi (F) ext4_msg(sb, KERN_ERR, 20289ba55543Szhangyi (F) "Invalid commit interval %d, " 20299ba55543Szhangyi (F) "must be smaller than %d", 20309ba55543Szhangyi (F) arg, INT_MAX / HZ); 20319ba55543Szhangyi (F) return -1; 20329ba55543Szhangyi (F) } 203326092bf5STheodore Ts'o sbi->s_commit_interval = HZ * arg; 2034670e9875STheodore Ts'o } else if (token == Opt_debug_want_extra_isize) { 20359803387cSTheodore Ts'o if ((arg & 1) || 20369803387cSTheodore Ts'o (arg < 4) || 20379803387cSTheodore Ts'o (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) { 20389803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 20399803387cSTheodore Ts'o "Invalid want_extra_isize %d", arg); 20409803387cSTheodore Ts'o return -1; 20419803387cSTheodore Ts'o } 2042670e9875STheodore Ts'o sbi->s_want_extra_isize = arg; 204326092bf5STheodore Ts'o } else if (token == Opt_max_batch_time) { 204426092bf5STheodore Ts'o sbi->s_max_batch_time = arg; 204526092bf5STheodore Ts'o } else if (token == Opt_min_batch_time) { 204626092bf5STheodore Ts'o sbi->s_min_batch_time = arg; 204726092bf5STheodore Ts'o } else if (token == Opt_inode_readahead_blks) { 2048e33e60eaSJan Kara if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 2049e33e60eaSJan Kara ext4_msg(sb, KERN_ERR, 2050e33e60eaSJan Kara "EXT4-fs: inode_readahead_blks must be " 2051e33e60eaSJan Kara "0 or a power of 2 smaller than 2^31"); 205226092bf5STheodore Ts'o return -1; 205326092bf5STheodore Ts'o } 205426092bf5STheodore Ts'o sbi->s_inode_readahead_blks = arg; 205526092bf5STheodore Ts'o } else if (token == Opt_init_itable) { 205626092bf5STheodore Ts'o set_opt(sb, INIT_INODE_TABLE); 205726092bf5STheodore Ts'o if (!args->from) 205826092bf5STheodore Ts'o arg = EXT4_DEF_LI_WAIT_MULT; 205926092bf5STheodore Ts'o sbi->s_li_wait_mult = arg; 2060df981d03STheodore Ts'o } else if (token == Opt_max_dir_size_kb) { 2061df981d03STheodore Ts'o sbi->s_max_dir_size_kb = arg; 206226092bf5STheodore Ts'o } else if (token == Opt_stripe) { 206326092bf5STheodore Ts'o sbi->s_stripe = arg; 20640efb3b23SJan Kara } else if (token == Opt_resuid) { 20650efb3b23SJan Kara uid = make_kuid(current_user_ns(), arg); 20660efb3b23SJan Kara if (!uid_valid(uid)) { 20675f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 20680efb3b23SJan Kara return -1; 20690efb3b23SJan Kara } 20700efb3b23SJan Kara sbi->s_resuid = uid; 20710efb3b23SJan Kara } else if (token == Opt_resgid) { 20720efb3b23SJan Kara gid = make_kgid(current_user_ns(), arg); 20730efb3b23SJan Kara if (!gid_valid(gid)) { 20745f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 20750efb3b23SJan Kara return -1; 20760efb3b23SJan Kara } 20770efb3b23SJan Kara sbi->s_resgid = gid; 20780efb3b23SJan Kara } else if (token == Opt_journal_dev) { 20790efb3b23SJan Kara if (is_remount) { 20800efb3b23SJan Kara ext4_msg(sb, KERN_ERR, 20810efb3b23SJan Kara "Cannot specify journal on remount"); 20820efb3b23SJan Kara return -1; 20830efb3b23SJan Kara } 20840efb3b23SJan Kara *journal_devnum = arg; 2085ad4eec61SEric Sandeen } else if (token == Opt_journal_path) { 2086ad4eec61SEric Sandeen char *journal_path; 2087ad4eec61SEric Sandeen struct inode *journal_inode; 2088ad4eec61SEric Sandeen struct path path; 2089ad4eec61SEric Sandeen int error; 2090ad4eec61SEric Sandeen 2091ad4eec61SEric Sandeen if (is_remount) { 2092ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, 2093ad4eec61SEric Sandeen "Cannot specify journal on remount"); 2094ad4eec61SEric Sandeen return -1; 2095ad4eec61SEric Sandeen } 2096ad4eec61SEric Sandeen journal_path = match_strdup(&args[0]); 2097ad4eec61SEric Sandeen if (!journal_path) { 2098ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: could not dup " 2099ad4eec61SEric Sandeen "journal device string"); 2100ad4eec61SEric Sandeen return -1; 2101ad4eec61SEric Sandeen } 2102ad4eec61SEric Sandeen 2103ad4eec61SEric Sandeen error = kern_path(journal_path, LOOKUP_FOLLOW, &path); 2104ad4eec61SEric Sandeen if (error) { 2105ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: could not find " 2106ad4eec61SEric Sandeen "journal device path: error %d", error); 2107ad4eec61SEric Sandeen kfree(journal_path); 2108ad4eec61SEric Sandeen return -1; 2109ad4eec61SEric Sandeen } 2110ad4eec61SEric Sandeen 21112b0143b5SDavid Howells journal_inode = d_inode(path.dentry); 2112ad4eec61SEric Sandeen if (!S_ISBLK(journal_inode->i_mode)) { 2113ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: journal path %s " 2114ad4eec61SEric Sandeen "is not a block device", journal_path); 2115ad4eec61SEric Sandeen path_put(&path); 2116ad4eec61SEric Sandeen kfree(journal_path); 2117ad4eec61SEric Sandeen return -1; 2118ad4eec61SEric Sandeen } 2119ad4eec61SEric Sandeen 2120ad4eec61SEric Sandeen *journal_devnum = new_encode_dev(journal_inode->i_rdev); 2121ad4eec61SEric Sandeen path_put(&path); 2122ad4eec61SEric Sandeen kfree(journal_path); 21230efb3b23SJan Kara } else if (token == Opt_journal_ioprio) { 21240efb3b23SJan Kara if (arg > 7) { 21255f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 21260efb3b23SJan Kara " (must be 0-7)"); 21270efb3b23SJan Kara return -1; 21280efb3b23SJan Kara } 21290efb3b23SJan Kara *journal_ioprio = 21300efb3b23SJan Kara IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 21316ddb2447STheodore Ts'o } else if (token == Opt_test_dummy_encryption) { 2132ed318a6cSEric Biggers return ext4_set_test_dummy_encryption(sb, opt, &args[0], 2133ed318a6cSEric Biggers is_remount); 213426092bf5STheodore Ts'o } else if (m->flags & MOPT_DATAJ) { 213526092bf5STheodore Ts'o if (is_remount) { 213626092bf5STheodore Ts'o if (!sbi->s_journal) 213726092bf5STheodore Ts'o ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 21385f3633e3SJan Kara else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 213926092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, 214026092bf5STheodore Ts'o "Cannot change data mode on remount"); 214126092bf5STheodore Ts'o return -1; 214226092bf5STheodore Ts'o } 214326092bf5STheodore Ts'o } else { 214426092bf5STheodore Ts'o clear_opt(sb, DATA_FLAGS); 214526092bf5STheodore Ts'o sbi->s_mount_opt |= m->mount_opt; 214626092bf5STheodore Ts'o } 214726092bf5STheodore Ts'o #ifdef CONFIG_QUOTA 214826092bf5STheodore Ts'o } else if (m->flags & MOPT_QFMT) { 214926092bf5STheodore Ts'o if (sb_any_quota_loaded(sb) && 215026092bf5STheodore Ts'o sbi->s_jquota_fmt != m->mount_opt) { 21515f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Cannot change journaled " 21525f3633e3SJan Kara "quota options when quota turned on"); 215326092bf5STheodore Ts'o return -1; 215426092bf5STheodore Ts'o } 2155e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) { 2156c325a67cSTheodore Ts'o ext4_msg(sb, KERN_INFO, 2157c325a67cSTheodore Ts'o "Quota format mount options ignored " 2158262b4662SJan Kara "when QUOTA feature is enabled"); 2159c325a67cSTheodore Ts'o return 1; 2160262b4662SJan Kara } 216126092bf5STheodore Ts'o sbi->s_jquota_fmt = m->mount_opt; 216226092bf5STheodore Ts'o #endif 21639cb20f94SIra Weiny } else if (token == Opt_dax || token == Opt_dax_always || 21649cb20f94SIra Weiny token == Opt_dax_inode || token == Opt_dax_never) { 2165ef83b6e8SDan Williams #ifdef CONFIG_FS_DAX 21669cb20f94SIra Weiny switch (token) { 21679cb20f94SIra Weiny case Opt_dax: 21689cb20f94SIra Weiny case Opt_dax_always: 2169829b37b8STheodore Ts'o if (is_remount && 2170829b37b8STheodore Ts'o (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2171829b37b8STheodore Ts'o (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) { 2172829b37b8STheodore Ts'o fail_dax_change_remount: 2173829b37b8STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't change " 2174829b37b8STheodore Ts'o "dax mount option while remounting"); 2175829b37b8STheodore Ts'o return -1; 2176829b37b8STheodore Ts'o } 2177829b37b8STheodore Ts'o if (is_remount && 2178829b37b8STheodore Ts'o (test_opt(sb, DATA_FLAGS) == 2179829b37b8STheodore Ts'o EXT4_MOUNT_JOURNAL_DATA)) { 2180829b37b8STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't mount with " 2181829b37b8STheodore Ts'o "both data=journal and dax"); 2182829b37b8STheodore Ts'o return -1; 2183829b37b8STheodore Ts'o } 2184ef83b6e8SDan Williams ext4_msg(sb, KERN_WARNING, 2185ef83b6e8SDan Williams "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); 21869cb20f94SIra Weiny sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS; 21879cb20f94SIra Weiny sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER; 21889cb20f94SIra Weiny break; 21899cb20f94SIra Weiny case Opt_dax_never: 2190829b37b8STheodore Ts'o if (is_remount && 2191829b37b8STheodore Ts'o (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2192829b37b8STheodore Ts'o (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) 2193829b37b8STheodore Ts'o goto fail_dax_change_remount; 21949cb20f94SIra Weiny sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER; 21959cb20f94SIra Weiny sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 21969cb20f94SIra Weiny break; 21979cb20f94SIra Weiny case Opt_dax_inode: 2198829b37b8STheodore Ts'o if (is_remount && 2199829b37b8STheodore Ts'o ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2200829b37b8STheodore Ts'o (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2201829b37b8STheodore Ts'o !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) 2202829b37b8STheodore Ts'o goto fail_dax_change_remount; 22039cb20f94SIra Weiny sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 22049cb20f94SIra Weiny sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER; 22059cb20f94SIra Weiny /* Strictly for printing options */ 22069cb20f94SIra Weiny sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE; 22079cb20f94SIra Weiny break; 22089cb20f94SIra Weiny } 2209ef83b6e8SDan Williams #else 2210923ae0ffSRoss Zwisler ext4_msg(sb, KERN_INFO, "dax option not supported"); 22119cb20f94SIra Weiny sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER; 22129cb20f94SIra Weiny sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 2213923ae0ffSRoss Zwisler return -1; 2214923ae0ffSRoss Zwisler #endif 22157915a861SAles Novak } else if (token == Opt_data_err_abort) { 22167915a861SAles Novak sbi->s_mount_opt |= m->mount_opt; 22177915a861SAles Novak } else if (token == Opt_data_err_ignore) { 22187915a861SAles Novak sbi->s_mount_opt &= ~m->mount_opt; 221926092bf5STheodore Ts'o } else { 222026092bf5STheodore Ts'o if (!args->from) 222126092bf5STheodore Ts'o arg = 1; 222226092bf5STheodore Ts'o if (m->flags & MOPT_CLEAR) 222326092bf5STheodore Ts'o arg = !arg; 222426092bf5STheodore Ts'o else if (unlikely(!(m->flags & MOPT_SET))) { 222526092bf5STheodore Ts'o ext4_msg(sb, KERN_WARNING, 222626092bf5STheodore Ts'o "buggy handling of option %s", opt); 222726092bf5STheodore Ts'o WARN_ON(1); 222826092bf5STheodore Ts'o return -1; 222926092bf5STheodore Ts'o } 223026092bf5STheodore Ts'o if (arg != 0) 223126092bf5STheodore Ts'o sbi->s_mount_opt |= m->mount_opt; 223226092bf5STheodore Ts'o else 223326092bf5STheodore Ts'o sbi->s_mount_opt &= ~m->mount_opt; 223426092bf5STheodore Ts'o } 223526092bf5STheodore Ts'o return 1; 223626092bf5STheodore Ts'o } 223726092bf5STheodore Ts'o 2238ac27a0ecSDave Kleikamp static int parse_options(char *options, struct super_block *sb, 2239c3191067STheodore Ts'o unsigned long *journal_devnum, 2240b3881f74STheodore Ts'o unsigned int *journal_ioprio, 2241661aa520SEric Sandeen int is_remount) 2242ac27a0ecSDave Kleikamp { 22431e1a76edSOlof Johansson struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb); 224433458eabSTheodore Ts'o char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name; 2245ac27a0ecSDave Kleikamp substring_t args[MAX_OPT_ARGS]; 224626092bf5STheodore Ts'o int token; 2247ac27a0ecSDave Kleikamp 2248ac27a0ecSDave Kleikamp if (!options) 2249ac27a0ecSDave Kleikamp return 1; 2250ac27a0ecSDave Kleikamp 2251ac27a0ecSDave Kleikamp while ((p = strsep(&options, ",")) != NULL) { 2252ac27a0ecSDave Kleikamp if (!*p) 2253ac27a0ecSDave Kleikamp continue; 225415121c18SEric Sandeen /* 225515121c18SEric Sandeen * Initialize args struct so we know whether arg was 225615121c18SEric Sandeen * found; some options take optional arguments. 225715121c18SEric Sandeen */ 2258caecd0afSSachin Kamat args[0].to = args[0].from = NULL; 2259ac27a0ecSDave Kleikamp token = match_token(p, tokens, args); 226026092bf5STheodore Ts'o if (handle_mount_opt(sb, p, token, args, journal_devnum, 226126092bf5STheodore Ts'o journal_ioprio, is_remount) < 0) 2262ac27a0ecSDave Kleikamp return 0; 2263ac27a0ecSDave Kleikamp } 2264ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 226549da9392SJan Kara /* 226649da9392SJan Kara * We do the test below only for project quotas. 'usrquota' and 226749da9392SJan Kara * 'grpquota' mount options are allowed even without quota feature 226849da9392SJan Kara * to support legacy quotas in quota files. 226949da9392SJan Kara */ 227049da9392SJan Kara if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) { 227149da9392SJan Kara ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. " 227249da9392SJan Kara "Cannot enable project quota enforcement."); 227349da9392SJan Kara return 0; 227449da9392SJan Kara } 227533458eabSTheodore Ts'o usr_qf_name = get_qf_name(sb, sbi, USRQUOTA); 227633458eabSTheodore Ts'o grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA); 227733458eabSTheodore Ts'o if (usr_qf_name || grp_qf_name) { 227833458eabSTheodore Ts'o if (test_opt(sb, USRQUOTA) && usr_qf_name) 2279fd8c37ecSTheodore Ts'o clear_opt(sb, USRQUOTA); 2280ac27a0ecSDave Kleikamp 228133458eabSTheodore Ts'o if (test_opt(sb, GRPQUOTA) && grp_qf_name) 2282fd8c37ecSTheodore Ts'o clear_opt(sb, GRPQUOTA); 2283ac27a0ecSDave Kleikamp 228456c50f11SDmitry Monakhov if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 2285b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "old and new quota " 2286b31e1552SEric Sandeen "format mixing"); 2287ac27a0ecSDave Kleikamp return 0; 2288ac27a0ecSDave Kleikamp } 2289ac27a0ecSDave Kleikamp 2290ac27a0ecSDave Kleikamp if (!sbi->s_jquota_fmt) { 2291b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journaled quota format " 2292b31e1552SEric Sandeen "not specified"); 2293ac27a0ecSDave Kleikamp return 0; 2294ac27a0ecSDave Kleikamp } 2295ac27a0ecSDave Kleikamp } 2296ac27a0ecSDave Kleikamp #endif 2297626b035bSRitesh Harjani if (test_opt(sb, DIOREAD_NOLOCK)) { 2298626b035bSRitesh Harjani int blocksize = 2299626b035bSRitesh Harjani BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 2300626b035bSRitesh Harjani if (blocksize < PAGE_SIZE) 2301626b035bSRitesh Harjani ext4_msg(sb, KERN_WARNING, "Warning: mounting with an " 2302626b035bSRitesh Harjani "experimental mount option 'dioread_nolock' " 2303626b035bSRitesh Harjani "for blocksize < PAGE_SIZE"); 2304626b035bSRitesh Harjani } 2305ac27a0ecSDave Kleikamp return 1; 2306ac27a0ecSDave Kleikamp } 2307ac27a0ecSDave Kleikamp 23082adf6da8STheodore Ts'o static inline void ext4_show_quota_options(struct seq_file *seq, 23092adf6da8STheodore Ts'o struct super_block *sb) 23102adf6da8STheodore Ts'o { 23112adf6da8STheodore Ts'o #if defined(CONFIG_QUOTA) 23122adf6da8STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 231333458eabSTheodore Ts'o char *usr_qf_name, *grp_qf_name; 23142adf6da8STheodore Ts'o 23152adf6da8STheodore Ts'o if (sbi->s_jquota_fmt) { 23162adf6da8STheodore Ts'o char *fmtname = ""; 23172adf6da8STheodore Ts'o 23182adf6da8STheodore Ts'o switch (sbi->s_jquota_fmt) { 23192adf6da8STheodore Ts'o case QFMT_VFS_OLD: 23202adf6da8STheodore Ts'o fmtname = "vfsold"; 23212adf6da8STheodore Ts'o break; 23222adf6da8STheodore Ts'o case QFMT_VFS_V0: 23232adf6da8STheodore Ts'o fmtname = "vfsv0"; 23242adf6da8STheodore Ts'o break; 23252adf6da8STheodore Ts'o case QFMT_VFS_V1: 23262adf6da8STheodore Ts'o fmtname = "vfsv1"; 23272adf6da8STheodore Ts'o break; 23282adf6da8STheodore Ts'o } 23292adf6da8STheodore Ts'o seq_printf(seq, ",jqfmt=%s", fmtname); 23302adf6da8STheodore Ts'o } 23312adf6da8STheodore Ts'o 233233458eabSTheodore Ts'o rcu_read_lock(); 233333458eabSTheodore Ts'o usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 233433458eabSTheodore Ts'o grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 233533458eabSTheodore Ts'o if (usr_qf_name) 233633458eabSTheodore Ts'o seq_show_option(seq, "usrjquota", usr_qf_name); 233733458eabSTheodore Ts'o if (grp_qf_name) 233833458eabSTheodore Ts'o seq_show_option(seq, "grpjquota", grp_qf_name); 233933458eabSTheodore Ts'o rcu_read_unlock(); 23402adf6da8STheodore Ts'o #endif 23412adf6da8STheodore Ts'o } 23422adf6da8STheodore Ts'o 23435a916be1STheodore Ts'o static const char *token2str(int token) 23445a916be1STheodore Ts'o { 234550df9fd5SHerton Ronaldo Krzesinski const struct match_token *t; 23465a916be1STheodore Ts'o 23475a916be1STheodore Ts'o for (t = tokens; t->token != Opt_err; t++) 23485a916be1STheodore Ts'o if (t->token == token && !strchr(t->pattern, '=')) 23495a916be1STheodore Ts'o break; 23505a916be1STheodore Ts'o return t->pattern; 23515a916be1STheodore Ts'o } 23525a916be1STheodore Ts'o 23532adf6da8STheodore Ts'o /* 23542adf6da8STheodore Ts'o * Show an option if 23552adf6da8STheodore Ts'o * - it's set to a non-default value OR 23562adf6da8STheodore Ts'o * - if the per-sb default is different from the global default 23572adf6da8STheodore Ts'o */ 235866acdcf4STheodore Ts'o static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 235966acdcf4STheodore Ts'o int nodefs) 23602adf6da8STheodore Ts'o { 23612adf6da8STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 23622adf6da8STheodore Ts'o struct ext4_super_block *es = sbi->s_es; 236368afa7e0STyson Nottingham int def_errors, def_mount_opt = sbi->s_def_mount_opt; 23645a916be1STheodore Ts'o const struct mount_opts *m; 236566acdcf4STheodore Ts'o char sep = nodefs ? '\n' : ','; 23662adf6da8STheodore Ts'o 236766acdcf4STheodore Ts'o #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 236866acdcf4STheodore Ts'o #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 23692adf6da8STheodore Ts'o 23702adf6da8STheodore Ts'o if (sbi->s_sb_block != 1) 23715a916be1STheodore Ts'o SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 23725a916be1STheodore Ts'o 23735a916be1STheodore Ts'o for (m = ext4_mount_opts; m->token != Opt_err; m++) { 23745a916be1STheodore Ts'o int want_set = m->flags & MOPT_SET; 23755a916be1STheodore Ts'o if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 23769cb20f94SIra Weiny (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP) 23775a916be1STheodore Ts'o continue; 237868afa7e0STyson Nottingham if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 23795a916be1STheodore Ts'o continue; /* skip if same as the default */ 23805a916be1STheodore Ts'o if ((want_set && 23815a916be1STheodore Ts'o (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 23825a916be1STheodore Ts'o (!want_set && (sbi->s_mount_opt & m->mount_opt))) 23835a916be1STheodore Ts'o continue; /* select Opt_noFoo vs Opt_Foo */ 23845a916be1STheodore Ts'o SEQ_OPTS_PRINT("%s", token2str(m->token)); 23855a916be1STheodore Ts'o } 23865a916be1STheodore Ts'o 238708cefc7aSEric W. Biederman if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 23885a916be1STheodore Ts'o le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 238908cefc7aSEric W. Biederman SEQ_OPTS_PRINT("resuid=%u", 239008cefc7aSEric W. Biederman from_kuid_munged(&init_user_ns, sbi->s_resuid)); 239108cefc7aSEric W. Biederman if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 23925a916be1STheodore Ts'o le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 239308cefc7aSEric W. Biederman SEQ_OPTS_PRINT("resgid=%u", 239408cefc7aSEric W. Biederman from_kgid_munged(&init_user_ns, sbi->s_resgid)); 239566acdcf4STheodore Ts'o def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 23965a916be1STheodore Ts'o if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 23975a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=remount-ro"); 23982adf6da8STheodore Ts'o if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 23995a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=continue"); 24002adf6da8STheodore Ts'o if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 24015a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=panic"); 240266acdcf4STheodore Ts'o if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 24035a916be1STheodore Ts'o SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 240466acdcf4STheodore Ts'o if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 24055a916be1STheodore Ts'o SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 240666acdcf4STheodore Ts'o if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 24075a916be1STheodore Ts'o SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 2408357fdad0SMatthew Garrett if (sb->s_flags & SB_I_VERSION) 24095a916be1STheodore Ts'o SEQ_OPTS_PUTS("i_version"); 241066acdcf4STheodore Ts'o if (nodefs || sbi->s_stripe) 24115a916be1STheodore Ts'o SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 241268afa7e0STyson Nottingham if (nodefs || EXT4_MOUNT_DATA_FLAGS & 241368afa7e0STyson Nottingham (sbi->s_mount_opt ^ def_mount_opt)) { 24142adf6da8STheodore Ts'o if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 24155a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=journal"); 24162adf6da8STheodore Ts'o else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 24175a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=ordered"); 24182adf6da8STheodore Ts'o else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 24195a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=writeback"); 24205a916be1STheodore Ts'o } 242166acdcf4STheodore Ts'o if (nodefs || 242266acdcf4STheodore Ts'o sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 24235a916be1STheodore Ts'o SEQ_OPTS_PRINT("inode_readahead_blks=%u", 24242adf6da8STheodore Ts'o sbi->s_inode_readahead_blks); 24252adf6da8STheodore Ts'o 2426ceec0376STyson Nottingham if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 242766acdcf4STheodore Ts'o (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 24285a916be1STheodore Ts'o SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 2429df981d03STheodore Ts'o if (nodefs || sbi->s_max_dir_size_kb) 2430df981d03STheodore Ts'o SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 24317915a861SAles Novak if (test_opt(sb, DATA_ERR_ABORT)) 24327915a861SAles Novak SEQ_OPTS_PUTS("data_err=abort"); 2433ed318a6cSEric Biggers 2434ed318a6cSEric Biggers fscrypt_show_test_dummy_encryption(seq, sep, sb); 24352adf6da8STheodore Ts'o 24364f74d15fSEric Biggers if (sb->s_flags & SB_INLINECRYPT) 24374f74d15fSEric Biggers SEQ_OPTS_PUTS("inlinecrypt"); 24384f74d15fSEric Biggers 24399cb20f94SIra Weiny if (test_opt(sb, DAX_ALWAYS)) { 24409cb20f94SIra Weiny if (IS_EXT2_SB(sb)) 24419cb20f94SIra Weiny SEQ_OPTS_PUTS("dax"); 24429cb20f94SIra Weiny else 24439cb20f94SIra Weiny SEQ_OPTS_PUTS("dax=always"); 24449cb20f94SIra Weiny } else if (test_opt2(sb, DAX_NEVER)) { 24459cb20f94SIra Weiny SEQ_OPTS_PUTS("dax=never"); 24469cb20f94SIra Weiny } else if (test_opt2(sb, DAX_INODE)) { 24479cb20f94SIra Weiny SEQ_OPTS_PUTS("dax=inode"); 24489cb20f94SIra Weiny } 24499cb20f94SIra Weiny 24502adf6da8STheodore Ts'o ext4_show_quota_options(seq, sb); 24512adf6da8STheodore Ts'o return 0; 24522adf6da8STheodore Ts'o } 24532adf6da8STheodore Ts'o 245466acdcf4STheodore Ts'o static int ext4_show_options(struct seq_file *seq, struct dentry *root) 245566acdcf4STheodore Ts'o { 245666acdcf4STheodore Ts'o return _ext4_show_options(seq, root->d_sb, 0); 245766acdcf4STheodore Ts'o } 245866acdcf4STheodore Ts'o 2459ebd173beSTheodore Ts'o int ext4_seq_options_show(struct seq_file *seq, void *offset) 246066acdcf4STheodore Ts'o { 246166acdcf4STheodore Ts'o struct super_block *sb = seq->private; 246266acdcf4STheodore Ts'o int rc; 246366acdcf4STheodore Ts'o 2464bc98a42cSDavid Howells seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 246566acdcf4STheodore Ts'o rc = _ext4_show_options(seq, sb, 1); 246666acdcf4STheodore Ts'o seq_puts(seq, "\n"); 246766acdcf4STheodore Ts'o return rc; 246866acdcf4STheodore Ts'o } 246966acdcf4STheodore Ts'o 2470617ba13bSMingming Cao static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 2471ac27a0ecSDave Kleikamp int read_only) 2472ac27a0ecSDave Kleikamp { 2473617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 2474c89128a0SJaegeuk Kim int err = 0; 2475ac27a0ecSDave Kleikamp 2476617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 2477b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "revision level too high, " 2478b31e1552SEric Sandeen "forcing read-only mode"); 2479c89128a0SJaegeuk Kim err = -EROFS; 24805adaccacSyangerkun goto done; 2481ac27a0ecSDave Kleikamp } 2482ac27a0ecSDave Kleikamp if (read_only) 2483281b5995STheodore Ts'o goto done; 2484617ba13bSMingming Cao if (!(sbi->s_mount_state & EXT4_VALID_FS)) 2485b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 2486b31e1552SEric Sandeen "running e2fsck is recommended"); 2487c8b459f4SLukas Czerner else if (sbi->s_mount_state & EXT4_ERROR_FS) 2488b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 2489b31e1552SEric Sandeen "warning: mounting fs with errors, " 2490b31e1552SEric Sandeen "running e2fsck is recommended"); 2491ed3ce80aSTao Ma else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 2492ac27a0ecSDave Kleikamp le16_to_cpu(es->s_mnt_count) >= 2493ac27a0ecSDave Kleikamp (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 2494b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 2495b31e1552SEric Sandeen "warning: maximal mount count reached, " 2496b31e1552SEric Sandeen "running e2fsck is recommended"); 2497ac27a0ecSDave Kleikamp else if (le32_to_cpu(es->s_checkinterval) && 24986a0678a7SArnd Bergmann (ext4_get_tstamp(es, s_lastcheck) + 24996a0678a7SArnd Bergmann le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 2500b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 2501b31e1552SEric Sandeen "warning: checktime reached, " 2502b31e1552SEric Sandeen "running e2fsck is recommended"); 25030390131bSFrank Mayhar if (!sbi->s_journal) 2504216c34b2SMarcin Slusarz es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 2505ac27a0ecSDave Kleikamp if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 2506617ba13bSMingming Cao es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 2507e8546d06SMarcin Slusarz le16_add_cpu(&es->s_mnt_count, 1); 25086a0678a7SArnd Bergmann ext4_update_tstamp(es, s_mtime); 25090390131bSFrank Mayhar if (sbi->s_journal) 2510e2b911c5SDarrick J. Wong ext4_set_feature_journal_needs_recovery(sb); 2511ac27a0ecSDave Kleikamp 2512c89128a0SJaegeuk Kim err = ext4_commit_super(sb, 1); 2513281b5995STheodore Ts'o done: 2514ac27a0ecSDave Kleikamp if (test_opt(sb, DEBUG)) 2515a9df9a49STheodore Ts'o printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 2516a2595b8aSTheodore Ts'o "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 2517ac27a0ecSDave Kleikamp sb->s_blocksize, 2518ac27a0ecSDave Kleikamp sbi->s_groups_count, 2519617ba13bSMingming Cao EXT4_BLOCKS_PER_GROUP(sb), 2520617ba13bSMingming Cao EXT4_INODES_PER_GROUP(sb), 2521a2595b8aSTheodore Ts'o sbi->s_mount_opt, sbi->s_mount_opt2); 2522ac27a0ecSDave Kleikamp 25237abc52c2SDan Magenheimer cleancache_init_fs(sb); 2524c89128a0SJaegeuk Kim return err; 2525ac27a0ecSDave Kleikamp } 2526ac27a0ecSDave Kleikamp 2527117fff10STheodore Ts'o int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 2528117fff10STheodore Ts'o { 2529117fff10STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 25307c990728SSuraj Jitindar Singh struct flex_groups **old_groups, **new_groups; 253137b0b6b8SDan Carpenter int size, i, j; 2532117fff10STheodore Ts'o 2533117fff10STheodore Ts'o if (!sbi->s_log_groups_per_flex) 2534117fff10STheodore Ts'o return 0; 2535117fff10STheodore Ts'o 2536117fff10STheodore Ts'o size = ext4_flex_group(sbi, ngroup - 1) + 1; 2537117fff10STheodore Ts'o if (size <= sbi->s_flex_groups_allocated) 2538117fff10STheodore Ts'o return 0; 2539117fff10STheodore Ts'o 25407c990728SSuraj Jitindar Singh new_groups = kvzalloc(roundup_pow_of_two(size * 25417c990728SSuraj Jitindar Singh sizeof(*sbi->s_flex_groups)), GFP_KERNEL); 2542117fff10STheodore Ts'o if (!new_groups) { 25437c990728SSuraj Jitindar Singh ext4_msg(sb, KERN_ERR, 25447c990728SSuraj Jitindar Singh "not enough memory for %d flex group pointers", size); 2545117fff10STheodore Ts'o return -ENOMEM; 2546117fff10STheodore Ts'o } 25477c990728SSuraj Jitindar Singh for (i = sbi->s_flex_groups_allocated; i < size; i++) { 25487c990728SSuraj Jitindar Singh new_groups[i] = kvzalloc(roundup_pow_of_two( 25497c990728SSuraj Jitindar Singh sizeof(struct flex_groups)), 25507c990728SSuraj Jitindar Singh GFP_KERNEL); 25517c990728SSuraj Jitindar Singh if (!new_groups[i]) { 255237b0b6b8SDan Carpenter for (j = sbi->s_flex_groups_allocated; j < i; j++) 255337b0b6b8SDan Carpenter kvfree(new_groups[j]); 25547c990728SSuraj Jitindar Singh kvfree(new_groups); 25557c990728SSuraj Jitindar Singh ext4_msg(sb, KERN_ERR, 25567c990728SSuraj Jitindar Singh "not enough memory for %d flex groups", size); 25577c990728SSuraj Jitindar Singh return -ENOMEM; 2558117fff10STheodore Ts'o } 25597c990728SSuraj Jitindar Singh } 25607c990728SSuraj Jitindar Singh rcu_read_lock(); 25617c990728SSuraj Jitindar Singh old_groups = rcu_dereference(sbi->s_flex_groups); 25627c990728SSuraj Jitindar Singh if (old_groups) 25637c990728SSuraj Jitindar Singh memcpy(new_groups, old_groups, 25647c990728SSuraj Jitindar Singh (sbi->s_flex_groups_allocated * 25657c990728SSuraj Jitindar Singh sizeof(struct flex_groups *))); 25667c990728SSuraj Jitindar Singh rcu_read_unlock(); 25677c990728SSuraj Jitindar Singh rcu_assign_pointer(sbi->s_flex_groups, new_groups); 25687c990728SSuraj Jitindar Singh sbi->s_flex_groups_allocated = size; 25697c990728SSuraj Jitindar Singh if (old_groups) 25707c990728SSuraj Jitindar Singh ext4_kvfree_array_rcu(old_groups); 2571117fff10STheodore Ts'o return 0; 2572117fff10STheodore Ts'o } 2573117fff10STheodore Ts'o 2574772cb7c8SJose R. Santos static int ext4_fill_flex_info(struct super_block *sb) 2575772cb7c8SJose R. Santos { 2576772cb7c8SJose R. Santos struct ext4_sb_info *sbi = EXT4_SB(sb); 2577772cb7c8SJose R. Santos struct ext4_group_desc *gdp = NULL; 25787c990728SSuraj Jitindar Singh struct flex_groups *fg; 2579772cb7c8SJose R. Santos ext4_group_t flex_group; 2580117fff10STheodore Ts'o int i, err; 2581772cb7c8SJose R. Santos 2582503358aeSTheodore Ts'o sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 2583d50f2ab6SXi Wang if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 2584772cb7c8SJose R. Santos sbi->s_log_groups_per_flex = 0; 2585772cb7c8SJose R. Santos return 1; 2586772cb7c8SJose R. Santos } 2587772cb7c8SJose R. Santos 2588117fff10STheodore Ts'o err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 2589117fff10STheodore Ts'o if (err) 2590772cb7c8SJose R. Santos goto failed; 2591772cb7c8SJose R. Santos 2592772cb7c8SJose R. Santos for (i = 0; i < sbi->s_groups_count; i++) { 259388b6edd1STheodore Ts'o gdp = ext4_get_group_desc(sb, i, NULL); 2594772cb7c8SJose R. Santos 2595772cb7c8SJose R. Santos flex_group = ext4_flex_group(sbi, i); 25967c990728SSuraj Jitindar Singh fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group); 25977c990728SSuraj Jitindar Singh atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes); 259890ba983fSTheodore Ts'o atomic64_add(ext4_free_group_clusters(sb, gdp), 25997c990728SSuraj Jitindar Singh &fg->free_clusters); 26007c990728SSuraj Jitindar Singh atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs); 2601772cb7c8SJose R. Santos } 2602772cb7c8SJose R. Santos 2603772cb7c8SJose R. Santos return 1; 2604772cb7c8SJose R. Santos failed: 2605772cb7c8SJose R. Santos return 0; 2606772cb7c8SJose R. Santos } 2607772cb7c8SJose R. Santos 2608e2b911c5SDarrick J. Wong static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 2609717d50e4SAndreas Dilger struct ext4_group_desc *gdp) 2610717d50e4SAndreas Dilger { 2611b47820edSDaeho Jeong int offset = offsetof(struct ext4_group_desc, bg_checksum); 2612717d50e4SAndreas Dilger __u16 crc = 0; 2613717d50e4SAndreas Dilger __le32 le_group = cpu_to_le32(block_group); 2614e2b911c5SDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 2615717d50e4SAndreas Dilger 26169aa5d32bSDmitry Monakhov if (ext4_has_metadata_csum(sbi->s_sb)) { 2617feb0ab32SDarrick J. Wong /* Use new metadata_csum algorithm */ 2618feb0ab32SDarrick J. Wong __u32 csum32; 2619b47820edSDaeho Jeong __u16 dummy_csum = 0; 2620feb0ab32SDarrick J. Wong 2621feb0ab32SDarrick J. Wong csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 2622feb0ab32SDarrick J. Wong sizeof(le_group)); 2623b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 2624b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 2625b47820edSDaeho Jeong sizeof(dummy_csum)); 2626b47820edSDaeho Jeong offset += sizeof(dummy_csum); 2627b47820edSDaeho Jeong if (offset < sbi->s_desc_size) 2628b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 2629b47820edSDaeho Jeong sbi->s_desc_size - offset); 2630feb0ab32SDarrick J. Wong 2631feb0ab32SDarrick J. Wong crc = csum32 & 0xFFFF; 2632feb0ab32SDarrick J. Wong goto out; 2633feb0ab32SDarrick J. Wong } 2634feb0ab32SDarrick J. Wong 2635feb0ab32SDarrick J. Wong /* old crc16 code */ 2636e2b911c5SDarrick J. Wong if (!ext4_has_feature_gdt_csum(sb)) 2637813d32f9SDarrick J. Wong return 0; 2638813d32f9SDarrick J. Wong 2639717d50e4SAndreas Dilger crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2640717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2641717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)gdp, offset); 2642717d50e4SAndreas Dilger offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2643717d50e4SAndreas Dilger /* for checksum of struct ext4_group_desc do the rest...*/ 2644e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb) && 2645717d50e4SAndreas Dilger offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2646717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)gdp + offset, 2647717d50e4SAndreas Dilger le16_to_cpu(sbi->s_es->s_desc_size) - 2648717d50e4SAndreas Dilger offset); 2649717d50e4SAndreas Dilger 2650feb0ab32SDarrick J. Wong out: 2651717d50e4SAndreas Dilger return cpu_to_le16(crc); 2652717d50e4SAndreas Dilger } 2653717d50e4SAndreas Dilger 2654feb0ab32SDarrick J. Wong int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2655717d50e4SAndreas Dilger struct ext4_group_desc *gdp) 2656717d50e4SAndreas Dilger { 2657feb0ab32SDarrick J. Wong if (ext4_has_group_desc_csum(sb) && 2658e2b911c5SDarrick J. Wong (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 2659717d50e4SAndreas Dilger return 0; 2660717d50e4SAndreas Dilger 2661717d50e4SAndreas Dilger return 1; 2662717d50e4SAndreas Dilger } 2663717d50e4SAndreas Dilger 2664feb0ab32SDarrick J. Wong void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2665feb0ab32SDarrick J. Wong struct ext4_group_desc *gdp) 2666feb0ab32SDarrick J. Wong { 2667feb0ab32SDarrick J. Wong if (!ext4_has_group_desc_csum(sb)) 2668feb0ab32SDarrick J. Wong return; 2669e2b911c5SDarrick J. Wong gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 2670feb0ab32SDarrick J. Wong } 2671feb0ab32SDarrick J. Wong 2672ac27a0ecSDave Kleikamp /* Called at mount-time, super-block is locked */ 2673bfff6873SLukas Czerner static int ext4_check_descriptors(struct super_block *sb, 2674829fa70dSTheodore Ts'o ext4_fsblk_t sb_block, 2675bfff6873SLukas Czerner ext4_group_t *first_not_zeroed) 2676ac27a0ecSDave Kleikamp { 2677617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 2678617ba13bSMingming Cao ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2679617ba13bSMingming Cao ext4_fsblk_t last_block; 268044de022cSTheodore Ts'o ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 2681bd81d8eeSLaurent Vivier ext4_fsblk_t block_bitmap; 2682bd81d8eeSLaurent Vivier ext4_fsblk_t inode_bitmap; 2683bd81d8eeSLaurent Vivier ext4_fsblk_t inode_table; 2684ce421581SJose R. Santos int flexbg_flag = 0; 2685bfff6873SLukas Czerner ext4_group_t i, grp = sbi->s_groups_count; 2686ac27a0ecSDave Kleikamp 2687e2b911c5SDarrick J. Wong if (ext4_has_feature_flex_bg(sb)) 2688ce421581SJose R. Santos flexbg_flag = 1; 2689ce421581SJose R. Santos 2690617ba13bSMingming Cao ext4_debug("Checking group descriptors"); 2691ac27a0ecSDave Kleikamp 2692197cd65aSAkinobu Mita for (i = 0; i < sbi->s_groups_count; i++) { 2693197cd65aSAkinobu Mita struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2694197cd65aSAkinobu Mita 2695ce421581SJose R. Santos if (i == sbi->s_groups_count - 1 || flexbg_flag) 2696bd81d8eeSLaurent Vivier last_block = ext4_blocks_count(sbi->s_es) - 1; 2697ac27a0ecSDave Kleikamp else 2698ac27a0ecSDave Kleikamp last_block = first_block + 2699617ba13bSMingming Cao (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2700ac27a0ecSDave Kleikamp 2701bfff6873SLukas Czerner if ((grp == sbi->s_groups_count) && 2702bfff6873SLukas Czerner !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2703bfff6873SLukas Czerner grp = i; 2704bfff6873SLukas Czerner 27058fadc143SAlexandre Ratchov block_bitmap = ext4_block_bitmap(sb, gdp); 2706829fa70dSTheodore Ts'o if (block_bitmap == sb_block) { 2707829fa70dSTheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2708829fa70dSTheodore Ts'o "Block bitmap for group %u overlaps " 2709829fa70dSTheodore Ts'o "superblock", i); 271018db4b4eSTheodore Ts'o if (!sb_rdonly(sb)) 271118db4b4eSTheodore Ts'o return 0; 2712829fa70dSTheodore Ts'o } 271377260807STheodore Ts'o if (block_bitmap >= sb_block + 1 && 271477260807STheodore Ts'o block_bitmap <= last_bg_block) { 271577260807STheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 271677260807STheodore Ts'o "Block bitmap for group %u overlaps " 271777260807STheodore Ts'o "block group descriptors", i); 271877260807STheodore Ts'o if (!sb_rdonly(sb)) 271977260807STheodore Ts'o return 0; 272077260807STheodore Ts'o } 27212b2d6d01STheodore Ts'o if (block_bitmap < first_block || block_bitmap > last_block) { 2722b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2723a9df9a49STheodore Ts'o "Block bitmap for group %u not in group " 2724b31e1552SEric Sandeen "(block %llu)!", i, block_bitmap); 2725ac27a0ecSDave Kleikamp return 0; 2726ac27a0ecSDave Kleikamp } 27278fadc143SAlexandre Ratchov inode_bitmap = ext4_inode_bitmap(sb, gdp); 2728829fa70dSTheodore Ts'o if (inode_bitmap == sb_block) { 2729829fa70dSTheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2730829fa70dSTheodore Ts'o "Inode bitmap for group %u overlaps " 2731829fa70dSTheodore Ts'o "superblock", i); 273218db4b4eSTheodore Ts'o if (!sb_rdonly(sb)) 273318db4b4eSTheodore Ts'o return 0; 2734829fa70dSTheodore Ts'o } 273577260807STheodore Ts'o if (inode_bitmap >= sb_block + 1 && 273677260807STheodore Ts'o inode_bitmap <= last_bg_block) { 273777260807STheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 273877260807STheodore Ts'o "Inode bitmap for group %u overlaps " 273977260807STheodore Ts'o "block group descriptors", i); 274077260807STheodore Ts'o if (!sb_rdonly(sb)) 274177260807STheodore Ts'o return 0; 274277260807STheodore Ts'o } 27432b2d6d01STheodore Ts'o if (inode_bitmap < first_block || inode_bitmap > last_block) { 2744b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2745a9df9a49STheodore Ts'o "Inode bitmap for group %u not in group " 2746b31e1552SEric Sandeen "(block %llu)!", i, inode_bitmap); 2747ac27a0ecSDave Kleikamp return 0; 2748ac27a0ecSDave Kleikamp } 27498fadc143SAlexandre Ratchov inode_table = ext4_inode_table(sb, gdp); 2750829fa70dSTheodore Ts'o if (inode_table == sb_block) { 2751829fa70dSTheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2752829fa70dSTheodore Ts'o "Inode table for group %u overlaps " 2753829fa70dSTheodore Ts'o "superblock", i); 275418db4b4eSTheodore Ts'o if (!sb_rdonly(sb)) 275518db4b4eSTheodore Ts'o return 0; 2756829fa70dSTheodore Ts'o } 275777260807STheodore Ts'o if (inode_table >= sb_block + 1 && 275877260807STheodore Ts'o inode_table <= last_bg_block) { 275977260807STheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 276077260807STheodore Ts'o "Inode table for group %u overlaps " 276177260807STheodore Ts'o "block group descriptors", i); 276277260807STheodore Ts'o if (!sb_rdonly(sb)) 276377260807STheodore Ts'o return 0; 276477260807STheodore Ts'o } 2765bd81d8eeSLaurent Vivier if (inode_table < first_block || 27662b2d6d01STheodore Ts'o inode_table + sbi->s_itb_per_group - 1 > last_block) { 2767b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2768a9df9a49STheodore Ts'o "Inode table for group %u not in group " 2769b31e1552SEric Sandeen "(block %llu)!", i, inode_table); 2770ac27a0ecSDave Kleikamp return 0; 2771ac27a0ecSDave Kleikamp } 2772955ce5f5SAneesh Kumar K.V ext4_lock_group(sb, i); 2773feb0ab32SDarrick J. Wong if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2774b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2775b31e1552SEric Sandeen "Checksum for group %u failed (%u!=%u)", 2776e2b911c5SDarrick J. Wong i, le16_to_cpu(ext4_group_desc_csum(sb, i, 2777fd2d4291SAvantika Mathur gdp)), le16_to_cpu(gdp->bg_checksum)); 2778bc98a42cSDavid Howells if (!sb_rdonly(sb)) { 2779955ce5f5SAneesh Kumar K.V ext4_unlock_group(sb, i); 2780717d50e4SAndreas Dilger return 0; 2781717d50e4SAndreas Dilger } 27827ee1ec4cSLi Zefan } 2783955ce5f5SAneesh Kumar K.V ext4_unlock_group(sb, i); 2784ce421581SJose R. Santos if (!flexbg_flag) 2785617ba13bSMingming Cao first_block += EXT4_BLOCKS_PER_GROUP(sb); 2786ac27a0ecSDave Kleikamp } 2787bfff6873SLukas Czerner if (NULL != first_not_zeroed) 2788bfff6873SLukas Czerner *first_not_zeroed = grp; 2789ac27a0ecSDave Kleikamp return 1; 2790ac27a0ecSDave Kleikamp } 2791ac27a0ecSDave Kleikamp 2792617ba13bSMingming Cao /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2793ac27a0ecSDave Kleikamp * the superblock) which were deleted from all directories, but held open by 2794ac27a0ecSDave Kleikamp * a process at the time of a crash. We walk the list and try to delete these 2795ac27a0ecSDave Kleikamp * inodes at recovery time (only with a read-write filesystem). 2796ac27a0ecSDave Kleikamp * 2797ac27a0ecSDave Kleikamp * In order to keep the orphan inode chain consistent during traversal (in 2798ac27a0ecSDave Kleikamp * case of crash during recovery), we link each inode into the superblock 2799ac27a0ecSDave Kleikamp * orphan list_head and handle it the same way as an inode deletion during 2800ac27a0ecSDave Kleikamp * normal operation (which journals the operations for us). 2801ac27a0ecSDave Kleikamp * 2802ac27a0ecSDave Kleikamp * We only do an iget() and an iput() on each inode, which is very safe if we 2803ac27a0ecSDave Kleikamp * accidentally point at an in-use or already deleted inode. The worst that 2804ac27a0ecSDave Kleikamp * can happen in this case is that we get a "bit already cleared" message from 2805617ba13bSMingming Cao * ext4_free_inode(). The only reason we would point at a wrong inode is if 2806ac27a0ecSDave Kleikamp * e2fsck was run on this filesystem, and it must have already done the orphan 2807ac27a0ecSDave Kleikamp * inode cleanup for us, so we can safely abort without any further action. 2808ac27a0ecSDave Kleikamp */ 2809617ba13bSMingming Cao static void ext4_orphan_cleanup(struct super_block *sb, 2810617ba13bSMingming Cao struct ext4_super_block *es) 2811ac27a0ecSDave Kleikamp { 2812ac27a0ecSDave Kleikamp unsigned int s_flags = sb->s_flags; 28132c98eb5eSTheodore Ts'o int ret, nr_orphans = 0, nr_truncates = 0; 2814ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 281595f1fda4Szhangyi (F) int quota_update = 0; 2816ac27a0ecSDave Kleikamp int i; 2817ac27a0ecSDave Kleikamp #endif 2818ac27a0ecSDave Kleikamp if (!es->s_last_orphan) { 2819ac27a0ecSDave Kleikamp jbd_debug(4, "no orphan inodes to clean up\n"); 2820ac27a0ecSDave Kleikamp return; 2821ac27a0ecSDave Kleikamp } 2822ac27a0ecSDave Kleikamp 2823a8f48a95SEric Sandeen if (bdev_read_only(sb->s_bdev)) { 2824b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "write access " 2825b31e1552SEric Sandeen "unavailable, skipping orphan cleanup"); 2826a8f48a95SEric Sandeen return; 2827a8f48a95SEric Sandeen } 2828a8f48a95SEric Sandeen 2829d39195c3SAmir Goldstein /* Check if feature set would not allow a r/w mount */ 2830d39195c3SAmir Goldstein if (!ext4_feature_set_ok(sb, 0)) { 2831d39195c3SAmir Goldstein ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2832d39195c3SAmir Goldstein "unknown ROCOMPAT features"); 2833d39195c3SAmir Goldstein return; 2834d39195c3SAmir Goldstein } 2835d39195c3SAmir Goldstein 2836617ba13bSMingming Cao if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2837c25f9bc6SEric Sandeen /* don't clear list on RO mount w/ errors */ 28381751e8a6SLinus Torvalds if (es->s_last_orphan && !(s_flags & SB_RDONLY)) { 283984474976SDmitry Monakhov ext4_msg(sb, KERN_INFO, "Errors on filesystem, " 2840ac27a0ecSDave Kleikamp "clearing orphan list.\n"); 2841ac27a0ecSDave Kleikamp es->s_last_orphan = 0; 2842c25f9bc6SEric Sandeen } 2843ac27a0ecSDave Kleikamp jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2844ac27a0ecSDave Kleikamp return; 2845ac27a0ecSDave Kleikamp } 2846ac27a0ecSDave Kleikamp 28471751e8a6SLinus Torvalds if (s_flags & SB_RDONLY) { 2848b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 28491751e8a6SLinus Torvalds sb->s_flags &= ~SB_RDONLY; 2850ac27a0ecSDave Kleikamp } 2851ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 2852ac27a0ecSDave Kleikamp /* Needed for iput() to work correctly and not trash data */ 28531751e8a6SLinus Torvalds sb->s_flags |= SB_ACTIVE; 285495f1fda4Szhangyi (F) 285595f1fda4Szhangyi (F) /* 285695f1fda4Szhangyi (F) * Turn on quotas which were not enabled for read-only mounts if 285795f1fda4Szhangyi (F) * filesystem has quota feature, so that they are updated correctly. 285895f1fda4Szhangyi (F) */ 28591751e8a6SLinus Torvalds if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) { 286095f1fda4Szhangyi (F) int ret = ext4_enable_quotas(sb); 286195f1fda4Szhangyi (F) 286295f1fda4Szhangyi (F) if (!ret) 286395f1fda4Szhangyi (F) quota_update = 1; 286495f1fda4Szhangyi (F) else 286595f1fda4Szhangyi (F) ext4_msg(sb, KERN_ERR, 286695f1fda4Szhangyi (F) "Cannot turn on quotas: error %d", ret); 286795f1fda4Szhangyi (F) } 286895f1fda4Szhangyi (F) 286995f1fda4Szhangyi (F) /* Turn on journaled quotas used for old sytle */ 2870a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2871617ba13bSMingming Cao if (EXT4_SB(sb)->s_qf_names[i]) { 2872617ba13bSMingming Cao int ret = ext4_quota_on_mount(sb, i); 287395f1fda4Szhangyi (F) 287495f1fda4Szhangyi (F) if (!ret) 287595f1fda4Szhangyi (F) quota_update = 1; 287695f1fda4Szhangyi (F) else 2877b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 2878b31e1552SEric Sandeen "Cannot turn on journaled " 287995f1fda4Szhangyi (F) "quota: type %d: error %d", i, ret); 2880ac27a0ecSDave Kleikamp } 2881ac27a0ecSDave Kleikamp } 2882ac27a0ecSDave Kleikamp #endif 2883ac27a0ecSDave Kleikamp 2884ac27a0ecSDave Kleikamp while (es->s_last_orphan) { 2885ac27a0ecSDave Kleikamp struct inode *inode; 2886ac27a0ecSDave Kleikamp 2887c65d5c6cSVegard Nossum /* 2888c65d5c6cSVegard Nossum * We may have encountered an error during cleanup; if 2889c65d5c6cSVegard Nossum * so, skip the rest. 2890c65d5c6cSVegard Nossum */ 2891c65d5c6cSVegard Nossum if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2892c65d5c6cSVegard Nossum jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2893c65d5c6cSVegard Nossum es->s_last_orphan = 0; 2894c65d5c6cSVegard Nossum break; 2895c65d5c6cSVegard Nossum } 2896c65d5c6cSVegard Nossum 289797bd42b9SJosef Bacik inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 289897bd42b9SJosef Bacik if (IS_ERR(inode)) { 2899ac27a0ecSDave Kleikamp es->s_last_orphan = 0; 2900ac27a0ecSDave Kleikamp break; 2901ac27a0ecSDave Kleikamp } 2902ac27a0ecSDave Kleikamp 2903617ba13bSMingming Cao list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2904871a2931SChristoph Hellwig dquot_initialize(inode); 2905ac27a0ecSDave Kleikamp if (inode->i_nlink) { 2906566370a2SPaul Taysom if (test_opt(sb, DEBUG)) 2907b31e1552SEric Sandeen ext4_msg(sb, KERN_DEBUG, 2908b31e1552SEric Sandeen "%s: truncating inode %lu to %lld bytes", 290946e665e9SHarvey Harrison __func__, inode->i_ino, inode->i_size); 2910e5f8eab8STheodore Ts'o jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2911ac27a0ecSDave Kleikamp inode->i_ino, inode->i_size); 29125955102cSAl Viro inode_lock(inode); 291355f252c9SLukas Czerner truncate_inode_pages(inode->i_mapping, inode->i_size); 29142c98eb5eSTheodore Ts'o ret = ext4_truncate(inode); 29152c98eb5eSTheodore Ts'o if (ret) 29162c98eb5eSTheodore Ts'o ext4_std_error(inode->i_sb, ret); 29175955102cSAl Viro inode_unlock(inode); 2918ac27a0ecSDave Kleikamp nr_truncates++; 2919ac27a0ecSDave Kleikamp } else { 2920566370a2SPaul Taysom if (test_opt(sb, DEBUG)) 2921b31e1552SEric Sandeen ext4_msg(sb, KERN_DEBUG, 2922b31e1552SEric Sandeen "%s: deleting unreferenced inode %lu", 292346e665e9SHarvey Harrison __func__, inode->i_ino); 2924ac27a0ecSDave Kleikamp jbd_debug(2, "deleting unreferenced inode %lu\n", 2925ac27a0ecSDave Kleikamp inode->i_ino); 2926ac27a0ecSDave Kleikamp nr_orphans++; 2927ac27a0ecSDave Kleikamp } 2928ac27a0ecSDave Kleikamp iput(inode); /* The delete magic happens here! */ 2929ac27a0ecSDave Kleikamp } 2930ac27a0ecSDave Kleikamp 2931ac27a0ecSDave Kleikamp #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 2932ac27a0ecSDave Kleikamp 2933ac27a0ecSDave Kleikamp if (nr_orphans) 2934b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 2935b31e1552SEric Sandeen PLURAL(nr_orphans)); 2936ac27a0ecSDave Kleikamp if (nr_truncates) 2937b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 2938b31e1552SEric Sandeen PLURAL(nr_truncates)); 2939ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 294095f1fda4Szhangyi (F) /* Turn off quotas if they were enabled for orphan cleanup */ 294195f1fda4Szhangyi (F) if (quota_update) { 2942a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 294395f1fda4Szhangyi (F) if (sb_dqopt(sb)->files[i]) 2944287a8095SChristoph Hellwig dquot_quota_off(sb, i); 2945ac27a0ecSDave Kleikamp } 294695f1fda4Szhangyi (F) } 2947ac27a0ecSDave Kleikamp #endif 29481751e8a6SLinus Torvalds sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2949ac27a0ecSDave Kleikamp } 29500b8e58a1SAndreas Dilger 2951cd2291a4SEric Sandeen /* 2952cd2291a4SEric Sandeen * Maximal extent format file size. 2953cd2291a4SEric Sandeen * Resulting logical blkno at s_maxbytes must fit in our on-disk 2954cd2291a4SEric Sandeen * extent format containers, within a sector_t, and within i_blocks 2955cd2291a4SEric Sandeen * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 2956cd2291a4SEric Sandeen * so that won't be a limiting factor. 2957cd2291a4SEric Sandeen * 2958f17722f9SLukas Czerner * However there is other limiting factor. We do store extents in the form 2959f17722f9SLukas Czerner * of starting block and length, hence the resulting length of the extent 2960f17722f9SLukas Czerner * covering maximum file size must fit into on-disk format containers as 2961f17722f9SLukas Czerner * well. Given that length is always by 1 unit bigger than max unit (because 2962f17722f9SLukas Czerner * we count 0 as well) we have to lower the s_maxbytes by one fs block. 2963f17722f9SLukas Czerner * 2964cd2291a4SEric Sandeen * Note, this does *not* consider any metadata overhead for vfs i_blocks. 2965cd2291a4SEric Sandeen */ 2966f287a1a5STheodore Ts'o static loff_t ext4_max_size(int blkbits, int has_huge_files) 2967cd2291a4SEric Sandeen { 2968cd2291a4SEric Sandeen loff_t res; 2969cd2291a4SEric Sandeen loff_t upper_limit = MAX_LFS_FILESIZE; 2970cd2291a4SEric Sandeen 297172deb455SChristoph Hellwig BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); 297272deb455SChristoph Hellwig 297372deb455SChristoph Hellwig if (!has_huge_files) { 2974cd2291a4SEric Sandeen upper_limit = (1LL << 32) - 1; 2975cd2291a4SEric Sandeen 2976cd2291a4SEric Sandeen /* total blocks in file system block size */ 2977cd2291a4SEric Sandeen upper_limit >>= (blkbits - 9); 2978cd2291a4SEric Sandeen upper_limit <<= blkbits; 2979cd2291a4SEric Sandeen } 2980cd2291a4SEric Sandeen 2981f17722f9SLukas Czerner /* 2982f17722f9SLukas Czerner * 32-bit extent-start container, ee_block. We lower the maxbytes 2983f17722f9SLukas Czerner * by one fs block, so ee_len can cover the extent of maximum file 2984f17722f9SLukas Czerner * size 2985f17722f9SLukas Czerner */ 2986f17722f9SLukas Czerner res = (1LL << 32) - 1; 2987cd2291a4SEric Sandeen res <<= blkbits; 2988cd2291a4SEric Sandeen 2989cd2291a4SEric Sandeen /* Sanity check against vm- & vfs- imposed limits */ 2990cd2291a4SEric Sandeen if (res > upper_limit) 2991cd2291a4SEric Sandeen res = upper_limit; 2992cd2291a4SEric Sandeen 2993cd2291a4SEric Sandeen return res; 2994cd2291a4SEric Sandeen } 2995ac27a0ecSDave Kleikamp 2996ac27a0ecSDave Kleikamp /* 2997cd2291a4SEric Sandeen * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 29980fc1b451SAneesh Kumar K.V * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 29990fc1b451SAneesh Kumar K.V * We need to be 1 filesystem block less than the 2^48 sector limit. 3000ac27a0ecSDave Kleikamp */ 3001f287a1a5STheodore Ts'o static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 3002ac27a0ecSDave Kleikamp { 3003617ba13bSMingming Cao loff_t res = EXT4_NDIR_BLOCKS; 30040fc1b451SAneesh Kumar K.V int meta_blocks; 30050fc1b451SAneesh Kumar K.V loff_t upper_limit; 30060b8e58a1SAndreas Dilger /* This is calculated to be the largest file size for a dense, block 30070b8e58a1SAndreas Dilger * mapped file such that the file's total number of 512-byte sectors, 30080b8e58a1SAndreas Dilger * including data and all indirect blocks, does not exceed (2^48 - 1). 30090b8e58a1SAndreas Dilger * 30100b8e58a1SAndreas Dilger * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 30110b8e58a1SAndreas Dilger * number of 512-byte sectors of the file. 30120fc1b451SAneesh Kumar K.V */ 30130fc1b451SAneesh Kumar K.V 301472deb455SChristoph Hellwig if (!has_huge_files) { 30150fc1b451SAneesh Kumar K.V /* 301672deb455SChristoph Hellwig * !has_huge_files or implies that the inode i_block field 301772deb455SChristoph Hellwig * represents total file blocks in 2^32 512-byte sectors == 301872deb455SChristoph Hellwig * size of vfs inode i_blocks * 8 30190fc1b451SAneesh Kumar K.V */ 30200fc1b451SAneesh Kumar K.V upper_limit = (1LL << 32) - 1; 30210fc1b451SAneesh Kumar K.V 30220fc1b451SAneesh Kumar K.V /* total blocks in file system block size */ 30230fc1b451SAneesh Kumar K.V upper_limit >>= (bits - 9); 30240fc1b451SAneesh Kumar K.V 30250fc1b451SAneesh Kumar K.V } else { 30268180a562SAneesh Kumar K.V /* 30278180a562SAneesh Kumar K.V * We use 48 bit ext4_inode i_blocks 30288180a562SAneesh Kumar K.V * With EXT4_HUGE_FILE_FL set the i_blocks 30298180a562SAneesh Kumar K.V * represent total number of blocks in 30308180a562SAneesh Kumar K.V * file system block size 30318180a562SAneesh Kumar K.V */ 30320fc1b451SAneesh Kumar K.V upper_limit = (1LL << 48) - 1; 30330fc1b451SAneesh Kumar K.V 30340fc1b451SAneesh Kumar K.V } 30350fc1b451SAneesh Kumar K.V 30360fc1b451SAneesh Kumar K.V /* indirect blocks */ 30370fc1b451SAneesh Kumar K.V meta_blocks = 1; 30380fc1b451SAneesh Kumar K.V /* double indirect blocks */ 30390fc1b451SAneesh Kumar K.V meta_blocks += 1 + (1LL << (bits-2)); 30400fc1b451SAneesh Kumar K.V /* tripple indirect blocks */ 30410fc1b451SAneesh Kumar K.V meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 30420fc1b451SAneesh Kumar K.V 30430fc1b451SAneesh Kumar K.V upper_limit -= meta_blocks; 30440fc1b451SAneesh Kumar K.V upper_limit <<= bits; 3045ac27a0ecSDave Kleikamp 3046ac27a0ecSDave Kleikamp res += 1LL << (bits-2); 3047ac27a0ecSDave Kleikamp res += 1LL << (2*(bits-2)); 3048ac27a0ecSDave Kleikamp res += 1LL << (3*(bits-2)); 3049ac27a0ecSDave Kleikamp res <<= bits; 3050ac27a0ecSDave Kleikamp if (res > upper_limit) 3051ac27a0ecSDave Kleikamp res = upper_limit; 30520fc1b451SAneesh Kumar K.V 30530fc1b451SAneesh Kumar K.V if (res > MAX_LFS_FILESIZE) 30540fc1b451SAneesh Kumar K.V res = MAX_LFS_FILESIZE; 30550fc1b451SAneesh Kumar K.V 3056ac27a0ecSDave Kleikamp return res; 3057ac27a0ecSDave Kleikamp } 3058ac27a0ecSDave Kleikamp 3059617ba13bSMingming Cao static ext4_fsblk_t descriptor_loc(struct super_block *sb, 306070bbb3e0SAndrew Morton ext4_fsblk_t logical_sb_block, int nr) 3061ac27a0ecSDave Kleikamp { 3062617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 3063fd2d4291SAvantika Mathur ext4_group_t bg, first_meta_bg; 3064ac27a0ecSDave Kleikamp int has_super = 0; 3065ac27a0ecSDave Kleikamp 3066ac27a0ecSDave Kleikamp first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 3067ac27a0ecSDave Kleikamp 3068e2b911c5SDarrick J. Wong if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 306970bbb3e0SAndrew Morton return logical_sb_block + nr + 1; 3070ac27a0ecSDave Kleikamp bg = sbi->s_desc_per_block * nr; 3071617ba13bSMingming Cao if (ext4_bg_has_super(sb, bg)) 3072ac27a0ecSDave Kleikamp has_super = 1; 30730b8e58a1SAndreas Dilger 3074bd63f6b0SDarrick J. Wong /* 3075bd63f6b0SDarrick J. Wong * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 3076bd63f6b0SDarrick J. Wong * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 3077bd63f6b0SDarrick J. Wong * on modern mke2fs or blksize > 1k on older mke2fs) then we must 3078bd63f6b0SDarrick J. Wong * compensate. 3079bd63f6b0SDarrick J. Wong */ 3080bd63f6b0SDarrick J. Wong if (sb->s_blocksize == 1024 && nr == 0 && 308149598e04SJun Piao le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 3082bd63f6b0SDarrick J. Wong has_super++; 3083bd63f6b0SDarrick J. Wong 3084617ba13bSMingming Cao return (has_super + ext4_group_first_block_no(sb, bg)); 3085ac27a0ecSDave Kleikamp } 3086ac27a0ecSDave Kleikamp 3087c9de560dSAlex Tomas /** 3088c9de560dSAlex Tomas * ext4_get_stripe_size: Get the stripe size. 3089c9de560dSAlex Tomas * @sbi: In memory super block info 3090c9de560dSAlex Tomas * 3091c9de560dSAlex Tomas * If we have specified it via mount option, then 3092c9de560dSAlex Tomas * use the mount option value. If the value specified at mount time is 3093c9de560dSAlex Tomas * greater than the blocks per group use the super block value. 3094c9de560dSAlex Tomas * If the super block value is greater than blocks per group return 0. 3095c9de560dSAlex Tomas * Allocator needs it be less than blocks per group. 3096c9de560dSAlex Tomas * 3097c9de560dSAlex Tomas */ 3098c9de560dSAlex Tomas static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 3099c9de560dSAlex Tomas { 3100c9de560dSAlex Tomas unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 3101c9de560dSAlex Tomas unsigned long stripe_width = 3102c9de560dSAlex Tomas le32_to_cpu(sbi->s_es->s_raid_stripe_width); 31033eb08658SDan Ehrenberg int ret; 3104c9de560dSAlex Tomas 3105c9de560dSAlex Tomas if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 31063eb08658SDan Ehrenberg ret = sbi->s_stripe; 31075469d7c3SJan Kara else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 31083eb08658SDan Ehrenberg ret = stripe_width; 31095469d7c3SJan Kara else if (stride && stride <= sbi->s_blocks_per_group) 31103eb08658SDan Ehrenberg ret = stride; 31113eb08658SDan Ehrenberg else 31123eb08658SDan Ehrenberg ret = 0; 3113c9de560dSAlex Tomas 31143eb08658SDan Ehrenberg /* 31153eb08658SDan Ehrenberg * If the stripe width is 1, this makes no sense and 31163eb08658SDan Ehrenberg * we set it to 0 to turn off stripe handling code. 31173eb08658SDan Ehrenberg */ 31183eb08658SDan Ehrenberg if (ret <= 1) 31193eb08658SDan Ehrenberg ret = 0; 3120c9de560dSAlex Tomas 31213eb08658SDan Ehrenberg return ret; 3122c9de560dSAlex Tomas } 3123ac27a0ecSDave Kleikamp 3124a13fb1a4SEric Sandeen /* 3125a13fb1a4SEric Sandeen * Check whether this filesystem can be mounted based on 3126a13fb1a4SEric Sandeen * the features present and the RDONLY/RDWR mount requested. 3127a13fb1a4SEric Sandeen * Returns 1 if this filesystem can be mounted as requested, 3128a13fb1a4SEric Sandeen * 0 if it cannot be. 3129a13fb1a4SEric Sandeen */ 3130a13fb1a4SEric Sandeen static int ext4_feature_set_ok(struct super_block *sb, int readonly) 3131a13fb1a4SEric Sandeen { 3132e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext4_incompat_features(sb)) { 3133a13fb1a4SEric Sandeen ext4_msg(sb, KERN_ERR, 3134a13fb1a4SEric Sandeen "Couldn't mount because of " 3135a13fb1a4SEric Sandeen "unsupported optional features (%x)", 3136a13fb1a4SEric Sandeen (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 3137a13fb1a4SEric Sandeen ~EXT4_FEATURE_INCOMPAT_SUPP)); 3138a13fb1a4SEric Sandeen return 0; 3139a13fb1a4SEric Sandeen } 3140a13fb1a4SEric Sandeen 3141c83ad55eSGabriel Krisman Bertazi #ifndef CONFIG_UNICODE 3142c83ad55eSGabriel Krisman Bertazi if (ext4_has_feature_casefold(sb)) { 3143c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 3144c83ad55eSGabriel Krisman Bertazi "Filesystem with casefold feature cannot be " 3145c83ad55eSGabriel Krisman Bertazi "mounted without CONFIG_UNICODE"); 3146c83ad55eSGabriel Krisman Bertazi return 0; 3147c83ad55eSGabriel Krisman Bertazi } 3148c83ad55eSGabriel Krisman Bertazi #endif 3149c83ad55eSGabriel Krisman Bertazi 3150a13fb1a4SEric Sandeen if (readonly) 3151a13fb1a4SEric Sandeen return 1; 3152a13fb1a4SEric Sandeen 3153e2b911c5SDarrick J. Wong if (ext4_has_feature_readonly(sb)) { 31542cb5cc8bSDarrick J. Wong ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 31551751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 31562cb5cc8bSDarrick J. Wong return 1; 31572cb5cc8bSDarrick J. Wong } 31582cb5cc8bSDarrick J. Wong 3159a13fb1a4SEric Sandeen /* Check that feature set is OK for a read-write mount */ 3160e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 3161a13fb1a4SEric Sandeen ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 3162a13fb1a4SEric Sandeen "unsupported optional features (%x)", 3163a13fb1a4SEric Sandeen (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 3164a13fb1a4SEric Sandeen ~EXT4_FEATURE_RO_COMPAT_SUPP)); 3165a13fb1a4SEric Sandeen return 0; 3166a13fb1a4SEric Sandeen } 3167e2b911c5SDarrick J. Wong if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 3168bab08ab9STheodore Ts'o ext4_msg(sb, KERN_ERR, 3169bab08ab9STheodore Ts'o "Can't support bigalloc feature without " 3170bab08ab9STheodore Ts'o "extents feature\n"); 3171bab08ab9STheodore Ts'o return 0; 3172bab08ab9STheodore Ts'o } 31737c319d32SAditya Kali 31749db176bcSJan Kara #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2) 3175d65d87a0STheodore Ts'o if (!readonly && (ext4_has_feature_quota(sb) || 3176d65d87a0STheodore Ts'o ext4_has_feature_project(sb))) { 31777c319d32SAditya Kali ext4_msg(sb, KERN_ERR, 3178d65d87a0STheodore Ts'o "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2"); 3179689c958cSLi Xi return 0; 3180689c958cSLi Xi } 31817c319d32SAditya Kali #endif /* CONFIG_QUOTA */ 3182a13fb1a4SEric Sandeen return 1; 3183a13fb1a4SEric Sandeen } 3184a13fb1a4SEric Sandeen 318566e61a9eSTheodore Ts'o /* 318666e61a9eSTheodore Ts'o * This function is called once a day if we have errors logged 318766e61a9eSTheodore Ts'o * on the file system 318866e61a9eSTheodore Ts'o */ 3189235699a8SKees Cook static void print_daily_error_info(struct timer_list *t) 319066e61a9eSTheodore Ts'o { 3191235699a8SKees Cook struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report); 3192235699a8SKees Cook struct super_block *sb = sbi->s_sb; 3193235699a8SKees Cook struct ext4_super_block *es = sbi->s_es; 319466e61a9eSTheodore Ts'o 319566e61a9eSTheodore Ts'o if (es->s_error_count) 3196ae0f78deSTheodore Ts'o /* fsck newer than v1.41.13 is needed to clean this condition. */ 3197ae0f78deSTheodore Ts'o ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 319866e61a9eSTheodore Ts'o le32_to_cpu(es->s_error_count)); 319966e61a9eSTheodore Ts'o if (es->s_first_error_time) { 32006a0678a7SArnd Bergmann printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 32016a0678a7SArnd Bergmann sb->s_id, 32026a0678a7SArnd Bergmann ext4_get_tstamp(es, s_first_error_time), 320366e61a9eSTheodore Ts'o (int) sizeof(es->s_first_error_func), 320466e61a9eSTheodore Ts'o es->s_first_error_func, 320566e61a9eSTheodore Ts'o le32_to_cpu(es->s_first_error_line)); 320666e61a9eSTheodore Ts'o if (es->s_first_error_ino) 3207651e1c3bSJoe Perches printk(KERN_CONT ": inode %u", 320866e61a9eSTheodore Ts'o le32_to_cpu(es->s_first_error_ino)); 320966e61a9eSTheodore Ts'o if (es->s_first_error_block) 3210651e1c3bSJoe Perches printk(KERN_CONT ": block %llu", (unsigned long long) 321166e61a9eSTheodore Ts'o le64_to_cpu(es->s_first_error_block)); 3212651e1c3bSJoe Perches printk(KERN_CONT "\n"); 321366e61a9eSTheodore Ts'o } 321466e61a9eSTheodore Ts'o if (es->s_last_error_time) { 32156a0678a7SArnd Bergmann printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 32166a0678a7SArnd Bergmann sb->s_id, 32176a0678a7SArnd Bergmann ext4_get_tstamp(es, s_last_error_time), 321866e61a9eSTheodore Ts'o (int) sizeof(es->s_last_error_func), 321966e61a9eSTheodore Ts'o es->s_last_error_func, 322066e61a9eSTheodore Ts'o le32_to_cpu(es->s_last_error_line)); 322166e61a9eSTheodore Ts'o if (es->s_last_error_ino) 3222651e1c3bSJoe Perches printk(KERN_CONT ": inode %u", 322366e61a9eSTheodore Ts'o le32_to_cpu(es->s_last_error_ino)); 322466e61a9eSTheodore Ts'o if (es->s_last_error_block) 3225651e1c3bSJoe Perches printk(KERN_CONT ": block %llu", (unsigned long long) 322666e61a9eSTheodore Ts'o le64_to_cpu(es->s_last_error_block)); 3227651e1c3bSJoe Perches printk(KERN_CONT "\n"); 322866e61a9eSTheodore Ts'o } 322966e61a9eSTheodore Ts'o mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 323066e61a9eSTheodore Ts'o } 323166e61a9eSTheodore Ts'o 3232bfff6873SLukas Czerner /* Find next suitable group and run ext4_init_inode_table */ 3233bfff6873SLukas Czerner static int ext4_run_li_request(struct ext4_li_request *elr) 3234bfff6873SLukas Czerner { 3235bfff6873SLukas Czerner struct ext4_group_desc *gdp = NULL; 32363d392b26STheodore Ts'o struct super_block *sb = elr->lr_super; 32373d392b26STheodore Ts'o ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 32383d392b26STheodore Ts'o ext4_group_t group = elr->lr_next_group; 3239bfff6873SLukas Czerner unsigned long timeout = 0; 32403d392b26STheodore Ts'o unsigned int prefetch_ios = 0; 3241bfff6873SLukas Czerner int ret = 0; 3242bfff6873SLukas Czerner 32433d392b26STheodore Ts'o if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) { 32443d392b26STheodore Ts'o elr->lr_next_group = ext4_mb_prefetch(sb, group, 32453d392b26STheodore Ts'o EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios); 32463d392b26STheodore Ts'o if (prefetch_ios) 32473d392b26STheodore Ts'o ext4_mb_prefetch_fini(sb, elr->lr_next_group, 32483d392b26STheodore Ts'o prefetch_ios); 32493d392b26STheodore Ts'o trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, 32503d392b26STheodore Ts'o prefetch_ios); 32513d392b26STheodore Ts'o if (group >= elr->lr_next_group) { 32523d392b26STheodore Ts'o ret = 1; 32533d392b26STheodore Ts'o if (elr->lr_first_not_zeroed != ngroups && 32543d392b26STheodore Ts'o !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) { 32553d392b26STheodore Ts'o elr->lr_next_group = elr->lr_first_not_zeroed; 32563d392b26STheodore Ts'o elr->lr_mode = EXT4_LI_MODE_ITABLE; 32573d392b26STheodore Ts'o ret = 0; 32583d392b26STheodore Ts'o } 32593d392b26STheodore Ts'o } 32603d392b26STheodore Ts'o return ret; 32613d392b26STheodore Ts'o } 3262bfff6873SLukas Czerner 32633d392b26STheodore Ts'o for (; group < ngroups; group++) { 3264bfff6873SLukas Czerner gdp = ext4_get_group_desc(sb, group, NULL); 3265bfff6873SLukas Czerner if (!gdp) { 3266bfff6873SLukas Czerner ret = 1; 3267bfff6873SLukas Czerner break; 3268bfff6873SLukas Czerner } 3269bfff6873SLukas Czerner 3270bfff6873SLukas Czerner if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3271bfff6873SLukas Czerner break; 3272bfff6873SLukas Czerner } 3273bfff6873SLukas Czerner 32747f511862STheodore Ts'o if (group >= ngroups) 3275bfff6873SLukas Czerner ret = 1; 3276bfff6873SLukas Czerner 3277bfff6873SLukas Czerner if (!ret) { 3278bfff6873SLukas Czerner timeout = jiffies; 3279bfff6873SLukas Czerner ret = ext4_init_inode_table(sb, group, 3280bfff6873SLukas Czerner elr->lr_timeout ? 0 : 1); 32813d392b26STheodore Ts'o trace_ext4_lazy_itable_init(sb, group); 3282bfff6873SLukas Czerner if (elr->lr_timeout == 0) { 328351ce6511SLukas Czerner timeout = (jiffies - timeout) * 32843d392b26STheodore Ts'o EXT4_SB(elr->lr_super)->s_li_wait_mult; 3285bfff6873SLukas Czerner elr->lr_timeout = timeout; 3286bfff6873SLukas Czerner } 3287bfff6873SLukas Czerner elr->lr_next_sched = jiffies + elr->lr_timeout; 3288bfff6873SLukas Czerner elr->lr_next_group = group + 1; 3289bfff6873SLukas Czerner } 3290bfff6873SLukas Czerner return ret; 3291bfff6873SLukas Czerner } 3292bfff6873SLukas Czerner 3293bfff6873SLukas Czerner /* 3294bfff6873SLukas Czerner * Remove lr_request from the list_request and free the 32954ed5c033SLukas Czerner * request structure. Should be called with li_list_mtx held 3296bfff6873SLukas Czerner */ 3297bfff6873SLukas Czerner static void ext4_remove_li_request(struct ext4_li_request *elr) 3298bfff6873SLukas Czerner { 3299bfff6873SLukas Czerner if (!elr) 3300bfff6873SLukas Czerner return; 3301bfff6873SLukas Czerner 3302bfff6873SLukas Czerner list_del(&elr->lr_request); 33033d392b26STheodore Ts'o EXT4_SB(elr->lr_super)->s_li_request = NULL; 3304bfff6873SLukas Czerner kfree(elr); 3305bfff6873SLukas Czerner } 3306bfff6873SLukas Czerner 3307bfff6873SLukas Czerner static void ext4_unregister_li_request(struct super_block *sb) 3308bfff6873SLukas Czerner { 33091bb933fbSLukas Czerner mutex_lock(&ext4_li_mtx); 33101bb933fbSLukas Czerner if (!ext4_li_info) { 33111bb933fbSLukas Czerner mutex_unlock(&ext4_li_mtx); 3312bfff6873SLukas Czerner return; 33131bb933fbSLukas Czerner } 3314bfff6873SLukas Czerner 3315bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 33161bb933fbSLukas Czerner ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3317bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 33181bb933fbSLukas Czerner mutex_unlock(&ext4_li_mtx); 3319bfff6873SLukas Czerner } 3320bfff6873SLukas Czerner 33218f1f7453SEric Sandeen static struct task_struct *ext4_lazyinit_task; 33228f1f7453SEric Sandeen 3323bfff6873SLukas Czerner /* 3324bfff6873SLukas Czerner * This is the function where ext4lazyinit thread lives. It walks 3325bfff6873SLukas Czerner * through the request list searching for next scheduled filesystem. 3326bfff6873SLukas Czerner * When such a fs is found, run the lazy initialization request 3327bfff6873SLukas Czerner * (ext4_rn_li_request) and keep track of the time spend in this 3328bfff6873SLukas Czerner * function. Based on that time we compute next schedule time of 3329bfff6873SLukas Czerner * the request. When walking through the list is complete, compute 3330bfff6873SLukas Czerner * next waking time and put itself into sleep. 3331bfff6873SLukas Czerner */ 3332bfff6873SLukas Czerner static int ext4_lazyinit_thread(void *arg) 3333bfff6873SLukas Czerner { 3334bfff6873SLukas Czerner struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 3335bfff6873SLukas Czerner struct list_head *pos, *n; 3336bfff6873SLukas Czerner struct ext4_li_request *elr; 33374ed5c033SLukas Czerner unsigned long next_wakeup, cur; 3338bfff6873SLukas Czerner 3339bfff6873SLukas Czerner BUG_ON(NULL == eli); 3340bfff6873SLukas Czerner 3341bfff6873SLukas Czerner cont_thread: 3342bfff6873SLukas Czerner while (true) { 3343bfff6873SLukas Czerner next_wakeup = MAX_JIFFY_OFFSET; 3344bfff6873SLukas Czerner 3345bfff6873SLukas Czerner mutex_lock(&eli->li_list_mtx); 3346bfff6873SLukas Czerner if (list_empty(&eli->li_request_list)) { 3347bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3348bfff6873SLukas Czerner goto exit_thread; 3349bfff6873SLukas Czerner } 3350bfff6873SLukas Czerner list_for_each_safe(pos, n, &eli->li_request_list) { 3351e22834f0SDmitry Monakhov int err = 0; 3352e22834f0SDmitry Monakhov int progress = 0; 3353bfff6873SLukas Czerner elr = list_entry(pos, struct ext4_li_request, 3354bfff6873SLukas Czerner lr_request); 3355bfff6873SLukas Czerner 3356e22834f0SDmitry Monakhov if (time_before(jiffies, elr->lr_next_sched)) { 3357e22834f0SDmitry Monakhov if (time_before(elr->lr_next_sched, next_wakeup)) 3358e22834f0SDmitry Monakhov next_wakeup = elr->lr_next_sched; 3359e22834f0SDmitry Monakhov continue; 3360e22834f0SDmitry Monakhov } 3361e22834f0SDmitry Monakhov if (down_read_trylock(&elr->lr_super->s_umount)) { 3362e22834f0SDmitry Monakhov if (sb_start_write_trylock(elr->lr_super)) { 3363e22834f0SDmitry Monakhov progress = 1; 3364e22834f0SDmitry Monakhov /* 3365e22834f0SDmitry Monakhov * We hold sb->s_umount, sb can not 3366e22834f0SDmitry Monakhov * be removed from the list, it is 3367e22834f0SDmitry Monakhov * now safe to drop li_list_mtx 3368e22834f0SDmitry Monakhov */ 3369e22834f0SDmitry Monakhov mutex_unlock(&eli->li_list_mtx); 3370e22834f0SDmitry Monakhov err = ext4_run_li_request(elr); 3371e22834f0SDmitry Monakhov sb_end_write(elr->lr_super); 3372e22834f0SDmitry Monakhov mutex_lock(&eli->li_list_mtx); 3373e22834f0SDmitry Monakhov n = pos->next; 3374e22834f0SDmitry Monakhov } 3375e22834f0SDmitry Monakhov up_read((&elr->lr_super->s_umount)); 3376e22834f0SDmitry Monakhov } 3377b2c78cd0STheodore Ts'o /* error, remove the lazy_init job */ 3378e22834f0SDmitry Monakhov if (err) { 3379bfff6873SLukas Czerner ext4_remove_li_request(elr); 3380bfff6873SLukas Czerner continue; 3381bfff6873SLukas Czerner } 3382e22834f0SDmitry Monakhov if (!progress) { 3383e22834f0SDmitry Monakhov elr->lr_next_sched = jiffies + 3384e22834f0SDmitry Monakhov (prandom_u32() 3385e22834f0SDmitry Monakhov % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3386b2c78cd0STheodore Ts'o } 3387bfff6873SLukas Czerner if (time_before(elr->lr_next_sched, next_wakeup)) 3388bfff6873SLukas Czerner next_wakeup = elr->lr_next_sched; 3389bfff6873SLukas Czerner } 3390bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3391bfff6873SLukas Czerner 3392a0acae0eSTejun Heo try_to_freeze(); 3393bfff6873SLukas Czerner 33944ed5c033SLukas Czerner cur = jiffies; 33954ed5c033SLukas Czerner if ((time_after_eq(cur, next_wakeup)) || 3396f4245bd4SLukas Czerner (MAX_JIFFY_OFFSET == next_wakeup)) { 3397bfff6873SLukas Czerner cond_resched(); 3398bfff6873SLukas Czerner continue; 3399bfff6873SLukas Czerner } 3400bfff6873SLukas Czerner 34014ed5c033SLukas Czerner schedule_timeout_interruptible(next_wakeup - cur); 34024ed5c033SLukas Czerner 34038f1f7453SEric Sandeen if (kthread_should_stop()) { 34048f1f7453SEric Sandeen ext4_clear_request_list(); 34058f1f7453SEric Sandeen goto exit_thread; 34068f1f7453SEric Sandeen } 3407bfff6873SLukas Czerner } 3408bfff6873SLukas Czerner 3409bfff6873SLukas Czerner exit_thread: 3410bfff6873SLukas Czerner /* 3411bfff6873SLukas Czerner * It looks like the request list is empty, but we need 3412bfff6873SLukas Czerner * to check it under the li_list_mtx lock, to prevent any 3413bfff6873SLukas Czerner * additions into it, and of course we should lock ext4_li_mtx 3414bfff6873SLukas Czerner * to atomically free the list and ext4_li_info, because at 3415bfff6873SLukas Czerner * this point another ext4 filesystem could be registering 3416bfff6873SLukas Czerner * new one. 3417bfff6873SLukas Czerner */ 3418bfff6873SLukas Czerner mutex_lock(&ext4_li_mtx); 3419bfff6873SLukas Czerner mutex_lock(&eli->li_list_mtx); 3420bfff6873SLukas Czerner if (!list_empty(&eli->li_request_list)) { 3421bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3422bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 3423bfff6873SLukas Czerner goto cont_thread; 3424bfff6873SLukas Czerner } 3425bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3426bfff6873SLukas Czerner kfree(ext4_li_info); 3427bfff6873SLukas Czerner ext4_li_info = NULL; 3428bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 3429bfff6873SLukas Czerner 3430bfff6873SLukas Czerner return 0; 3431bfff6873SLukas Czerner } 3432bfff6873SLukas Czerner 3433bfff6873SLukas Czerner static void ext4_clear_request_list(void) 3434bfff6873SLukas Czerner { 3435bfff6873SLukas Czerner struct list_head *pos, *n; 3436bfff6873SLukas Czerner struct ext4_li_request *elr; 3437bfff6873SLukas Czerner 3438bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 3439bfff6873SLukas Czerner list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3440bfff6873SLukas Czerner elr = list_entry(pos, struct ext4_li_request, 3441bfff6873SLukas Czerner lr_request); 3442bfff6873SLukas Czerner ext4_remove_li_request(elr); 3443bfff6873SLukas Czerner } 3444bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 3445bfff6873SLukas Czerner } 3446bfff6873SLukas Czerner 3447bfff6873SLukas Czerner static int ext4_run_lazyinit_thread(void) 3448bfff6873SLukas Czerner { 34498f1f7453SEric Sandeen ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 34508f1f7453SEric Sandeen ext4_li_info, "ext4lazyinit"); 34518f1f7453SEric Sandeen if (IS_ERR(ext4_lazyinit_task)) { 34528f1f7453SEric Sandeen int err = PTR_ERR(ext4_lazyinit_task); 3453bfff6873SLukas Czerner ext4_clear_request_list(); 3454bfff6873SLukas Czerner kfree(ext4_li_info); 3455bfff6873SLukas Czerner ext4_li_info = NULL; 345692b97816STheodore Ts'o printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3457bfff6873SLukas Czerner "initialization thread\n", 3458bfff6873SLukas Czerner err); 3459bfff6873SLukas Czerner return err; 3460bfff6873SLukas Czerner } 3461bfff6873SLukas Czerner ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3462bfff6873SLukas Czerner return 0; 3463bfff6873SLukas Czerner } 3464bfff6873SLukas Czerner 3465bfff6873SLukas Czerner /* 3466bfff6873SLukas Czerner * Check whether it make sense to run itable init. thread or not. 3467bfff6873SLukas Czerner * If there is at least one uninitialized inode table, return 3468bfff6873SLukas Czerner * corresponding group number, else the loop goes through all 3469bfff6873SLukas Czerner * groups and return total number of groups. 3470bfff6873SLukas Czerner */ 3471bfff6873SLukas Czerner static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3472bfff6873SLukas Czerner { 3473bfff6873SLukas Czerner ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3474bfff6873SLukas Czerner struct ext4_group_desc *gdp = NULL; 3475bfff6873SLukas Czerner 34768844618dSTheodore Ts'o if (!ext4_has_group_desc_csum(sb)) 34778844618dSTheodore Ts'o return ngroups; 34788844618dSTheodore Ts'o 3479bfff6873SLukas Czerner for (group = 0; group < ngroups; group++) { 3480bfff6873SLukas Czerner gdp = ext4_get_group_desc(sb, group, NULL); 3481bfff6873SLukas Czerner if (!gdp) 3482bfff6873SLukas Czerner continue; 3483bfff6873SLukas Czerner 348450122847STheodore Ts'o if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3485bfff6873SLukas Czerner break; 3486bfff6873SLukas Czerner } 3487bfff6873SLukas Czerner 3488bfff6873SLukas Czerner return group; 3489bfff6873SLukas Czerner } 3490bfff6873SLukas Czerner 3491bfff6873SLukas Czerner static int ext4_li_info_new(void) 3492bfff6873SLukas Czerner { 3493bfff6873SLukas Czerner struct ext4_lazy_init *eli = NULL; 3494bfff6873SLukas Czerner 3495bfff6873SLukas Czerner eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3496bfff6873SLukas Czerner if (!eli) 3497bfff6873SLukas Czerner return -ENOMEM; 3498bfff6873SLukas Czerner 3499bfff6873SLukas Czerner INIT_LIST_HEAD(&eli->li_request_list); 3500bfff6873SLukas Czerner mutex_init(&eli->li_list_mtx); 3501bfff6873SLukas Czerner 3502bfff6873SLukas Czerner eli->li_state |= EXT4_LAZYINIT_QUIT; 3503bfff6873SLukas Czerner 3504bfff6873SLukas Czerner ext4_li_info = eli; 3505bfff6873SLukas Czerner 3506bfff6873SLukas Czerner return 0; 3507bfff6873SLukas Czerner } 3508bfff6873SLukas Czerner 3509bfff6873SLukas Czerner static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3510bfff6873SLukas Czerner ext4_group_t start) 3511bfff6873SLukas Czerner { 3512bfff6873SLukas Czerner struct ext4_li_request *elr; 3513bfff6873SLukas Czerner 3514bfff6873SLukas Czerner elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3515bfff6873SLukas Czerner if (!elr) 3516bfff6873SLukas Czerner return NULL; 3517bfff6873SLukas Czerner 3518bfff6873SLukas Czerner elr->lr_super = sb; 35193d392b26STheodore Ts'o elr->lr_first_not_zeroed = start; 35203d392b26STheodore Ts'o if (test_opt(sb, PREFETCH_BLOCK_BITMAPS)) 35213d392b26STheodore Ts'o elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP; 35223d392b26STheodore Ts'o else { 35233d392b26STheodore Ts'o elr->lr_mode = EXT4_LI_MODE_ITABLE; 3524bfff6873SLukas Czerner elr->lr_next_group = start; 35253d392b26STheodore Ts'o } 3526bfff6873SLukas Czerner 3527bfff6873SLukas Czerner /* 3528bfff6873SLukas Czerner * Randomize first schedule time of the request to 3529bfff6873SLukas Czerner * spread the inode table initialization requests 3530bfff6873SLukas Czerner * better. 3531bfff6873SLukas Czerner */ 3532dd1f723bSTheodore Ts'o elr->lr_next_sched = jiffies + (prandom_u32() % 3533dd1f723bSTheodore Ts'o (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3534bfff6873SLukas Czerner return elr; 3535bfff6873SLukas Czerner } 3536bfff6873SLukas Czerner 35377f511862STheodore Ts'o int ext4_register_li_request(struct super_block *sb, 3538bfff6873SLukas Czerner ext4_group_t first_not_zeroed) 3539bfff6873SLukas Czerner { 3540bfff6873SLukas Czerner struct ext4_sb_info *sbi = EXT4_SB(sb); 35417f511862STheodore Ts'o struct ext4_li_request *elr = NULL; 354249598e04SJun Piao ext4_group_t ngroups = sbi->s_groups_count; 35436c5a6cb9SAndrew Morton int ret = 0; 3544bfff6873SLukas Czerner 35457f511862STheodore Ts'o mutex_lock(&ext4_li_mtx); 354651ce6511SLukas Czerner if (sbi->s_li_request != NULL) { 354751ce6511SLukas Czerner /* 354851ce6511SLukas Czerner * Reset timeout so it can be computed again, because 354951ce6511SLukas Czerner * s_li_wait_mult might have changed. 355051ce6511SLukas Czerner */ 355151ce6511SLukas Czerner sbi->s_li_request->lr_timeout = 0; 35527f511862STheodore Ts'o goto out; 355351ce6511SLukas Czerner } 3554bfff6873SLukas Czerner 35553d392b26STheodore Ts'o if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) && 35563d392b26STheodore Ts'o (first_not_zeroed == ngroups || sb_rdonly(sb) || 35573d392b26STheodore Ts'o !test_opt(sb, INIT_INODE_TABLE))) 35587f511862STheodore Ts'o goto out; 3559bfff6873SLukas Czerner 3560bfff6873SLukas Czerner elr = ext4_li_request_new(sb, first_not_zeroed); 35617f511862STheodore Ts'o if (!elr) { 35627f511862STheodore Ts'o ret = -ENOMEM; 35637f511862STheodore Ts'o goto out; 35647f511862STheodore Ts'o } 3565bfff6873SLukas Czerner 3566bfff6873SLukas Czerner if (NULL == ext4_li_info) { 3567bfff6873SLukas Czerner ret = ext4_li_info_new(); 3568bfff6873SLukas Czerner if (ret) 3569bfff6873SLukas Czerner goto out; 3570bfff6873SLukas Czerner } 3571bfff6873SLukas Czerner 3572bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 3573bfff6873SLukas Czerner list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3574bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 3575bfff6873SLukas Czerner 3576bfff6873SLukas Czerner sbi->s_li_request = elr; 357746e4690bSTao Ma /* 357846e4690bSTao Ma * set elr to NULL here since it has been inserted to 357946e4690bSTao Ma * the request_list and the removal and free of it is 358046e4690bSTao Ma * handled by ext4_clear_request_list from now on. 358146e4690bSTao Ma */ 358246e4690bSTao Ma elr = NULL; 3583bfff6873SLukas Czerner 3584bfff6873SLukas Czerner if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3585bfff6873SLukas Czerner ret = ext4_run_lazyinit_thread(); 3586bfff6873SLukas Czerner if (ret) 3587bfff6873SLukas Czerner goto out; 3588bfff6873SLukas Czerner } 3589bfff6873SLukas Czerner out: 3590bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 3591beed5ecbSNicolas Kaiser if (ret) 3592bfff6873SLukas Czerner kfree(elr); 3593bfff6873SLukas Czerner return ret; 3594bfff6873SLukas Czerner } 3595bfff6873SLukas Czerner 3596bfff6873SLukas Czerner /* 3597bfff6873SLukas Czerner * We do not need to lock anything since this is called on 3598bfff6873SLukas Czerner * module unload. 3599bfff6873SLukas Czerner */ 3600bfff6873SLukas Czerner static void ext4_destroy_lazyinit_thread(void) 3601bfff6873SLukas Czerner { 3602bfff6873SLukas Czerner /* 3603bfff6873SLukas Czerner * If thread exited earlier 3604bfff6873SLukas Czerner * there's nothing to be done. 3605bfff6873SLukas Czerner */ 36068f1f7453SEric Sandeen if (!ext4_li_info || !ext4_lazyinit_task) 3607bfff6873SLukas Czerner return; 3608bfff6873SLukas Czerner 36098f1f7453SEric Sandeen kthread_stop(ext4_lazyinit_task); 3610bfff6873SLukas Czerner } 3611bfff6873SLukas Czerner 361225ed6e8aSDarrick J. Wong static int set_journal_csum_feature_set(struct super_block *sb) 361325ed6e8aSDarrick J. Wong { 361425ed6e8aSDarrick J. Wong int ret = 1; 361525ed6e8aSDarrick J. Wong int compat, incompat; 361625ed6e8aSDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 361725ed6e8aSDarrick J. Wong 36189aa5d32bSDmitry Monakhov if (ext4_has_metadata_csum(sb)) { 3619db9ee220SDarrick J. Wong /* journal checksum v3 */ 362025ed6e8aSDarrick J. Wong compat = 0; 3621db9ee220SDarrick J. Wong incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 362225ed6e8aSDarrick J. Wong } else { 362325ed6e8aSDarrick J. Wong /* journal checksum v1 */ 362425ed6e8aSDarrick J. Wong compat = JBD2_FEATURE_COMPAT_CHECKSUM; 362525ed6e8aSDarrick J. Wong incompat = 0; 362625ed6e8aSDarrick J. Wong } 362725ed6e8aSDarrick J. Wong 3628feb8c6d3SDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 3629feb8c6d3SDarrick J. Wong JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3630feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_CSUM_V3 | 3631feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_CSUM_V2); 363225ed6e8aSDarrick J. Wong if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 363325ed6e8aSDarrick J. Wong ret = jbd2_journal_set_features(sbi->s_journal, 363425ed6e8aSDarrick J. Wong compat, 0, 363525ed6e8aSDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 363625ed6e8aSDarrick J. Wong incompat); 363725ed6e8aSDarrick J. Wong } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 363825ed6e8aSDarrick J. Wong ret = jbd2_journal_set_features(sbi->s_journal, 363925ed6e8aSDarrick J. Wong compat, 0, 364025ed6e8aSDarrick J. Wong incompat); 364125ed6e8aSDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 0, 0, 364225ed6e8aSDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 364325ed6e8aSDarrick J. Wong } else { 3644feb8c6d3SDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3645feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 364625ed6e8aSDarrick J. Wong } 364725ed6e8aSDarrick J. Wong 364825ed6e8aSDarrick J. Wong return ret; 364925ed6e8aSDarrick J. Wong } 365025ed6e8aSDarrick J. Wong 3651952fc18eSTheodore Ts'o /* 3652952fc18eSTheodore Ts'o * Note: calculating the overhead so we can be compatible with 3653952fc18eSTheodore Ts'o * historical BSD practice is quite difficult in the face of 3654952fc18eSTheodore Ts'o * clusters/bigalloc. This is because multiple metadata blocks from 3655952fc18eSTheodore Ts'o * different block group can end up in the same allocation cluster. 3656952fc18eSTheodore Ts'o * Calculating the exact overhead in the face of clustered allocation 3657952fc18eSTheodore Ts'o * requires either O(all block bitmaps) in memory or O(number of block 3658952fc18eSTheodore Ts'o * groups**2) in time. We will still calculate the superblock for 3659952fc18eSTheodore Ts'o * older file systems --- and if we come across with a bigalloc file 3660952fc18eSTheodore Ts'o * system with zero in s_overhead_clusters the estimate will be close to 3661952fc18eSTheodore Ts'o * correct especially for very large cluster sizes --- but for newer 3662952fc18eSTheodore Ts'o * file systems, it's better to calculate this figure once at mkfs 3663952fc18eSTheodore Ts'o * time, and store it in the superblock. If the superblock value is 3664952fc18eSTheodore Ts'o * present (even for non-bigalloc file systems), we will use it. 3665952fc18eSTheodore Ts'o */ 3666952fc18eSTheodore Ts'o static int count_overhead(struct super_block *sb, ext4_group_t grp, 3667952fc18eSTheodore Ts'o char *buf) 3668952fc18eSTheodore Ts'o { 3669952fc18eSTheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 3670952fc18eSTheodore Ts'o struct ext4_group_desc *gdp; 3671952fc18eSTheodore Ts'o ext4_fsblk_t first_block, last_block, b; 3672952fc18eSTheodore Ts'o ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3673952fc18eSTheodore Ts'o int s, j, count = 0; 3674952fc18eSTheodore Ts'o 3675e2b911c5SDarrick J. Wong if (!ext4_has_feature_bigalloc(sb)) 36760548bbb8STheodore Ts'o return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 36770548bbb8STheodore Ts'o sbi->s_itb_per_group + 2); 36780548bbb8STheodore Ts'o 3679952fc18eSTheodore Ts'o first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3680952fc18eSTheodore Ts'o (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3681952fc18eSTheodore Ts'o last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3682952fc18eSTheodore Ts'o for (i = 0; i < ngroups; i++) { 3683952fc18eSTheodore Ts'o gdp = ext4_get_group_desc(sb, i, NULL); 3684952fc18eSTheodore Ts'o b = ext4_block_bitmap(sb, gdp); 3685952fc18eSTheodore Ts'o if (b >= first_block && b <= last_block) { 3686952fc18eSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3687952fc18eSTheodore Ts'o count++; 3688952fc18eSTheodore Ts'o } 3689952fc18eSTheodore Ts'o b = ext4_inode_bitmap(sb, gdp); 3690952fc18eSTheodore Ts'o if (b >= first_block && b <= last_block) { 3691952fc18eSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3692952fc18eSTheodore Ts'o count++; 3693952fc18eSTheodore Ts'o } 3694952fc18eSTheodore Ts'o b = ext4_inode_table(sb, gdp); 3695952fc18eSTheodore Ts'o if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3696952fc18eSTheodore Ts'o for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3697952fc18eSTheodore Ts'o int c = EXT4_B2C(sbi, b - first_block); 3698952fc18eSTheodore Ts'o ext4_set_bit(c, buf); 3699952fc18eSTheodore Ts'o count++; 3700952fc18eSTheodore Ts'o } 3701952fc18eSTheodore Ts'o if (i != grp) 3702952fc18eSTheodore Ts'o continue; 3703952fc18eSTheodore Ts'o s = 0; 3704952fc18eSTheodore Ts'o if (ext4_bg_has_super(sb, grp)) { 3705952fc18eSTheodore Ts'o ext4_set_bit(s++, buf); 3706952fc18eSTheodore Ts'o count++; 3707952fc18eSTheodore Ts'o } 3708c48ae41bSTheodore Ts'o j = ext4_bg_num_gdb(sb, grp); 3709c48ae41bSTheodore Ts'o if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 3710c48ae41bSTheodore Ts'o ext4_error(sb, "Invalid number of block group " 3711c48ae41bSTheodore Ts'o "descriptor blocks: %d", j); 3712c48ae41bSTheodore Ts'o j = EXT4_BLOCKS_PER_GROUP(sb) - s; 3713952fc18eSTheodore Ts'o } 3714c48ae41bSTheodore Ts'o count += j; 3715c48ae41bSTheodore Ts'o for (; j > 0; j--) 3716c48ae41bSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3717952fc18eSTheodore Ts'o } 3718952fc18eSTheodore Ts'o if (!count) 3719952fc18eSTheodore Ts'o return 0; 3720952fc18eSTheodore Ts'o return EXT4_CLUSTERS_PER_GROUP(sb) - 3721952fc18eSTheodore Ts'o ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3722952fc18eSTheodore Ts'o } 3723952fc18eSTheodore Ts'o 3724952fc18eSTheodore Ts'o /* 3725952fc18eSTheodore Ts'o * Compute the overhead and stash it in sbi->s_overhead 3726952fc18eSTheodore Ts'o */ 3727952fc18eSTheodore Ts'o int ext4_calculate_overhead(struct super_block *sb) 3728952fc18eSTheodore Ts'o { 3729952fc18eSTheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 3730952fc18eSTheodore Ts'o struct ext4_super_block *es = sbi->s_es; 37313c816dedSEric Whitney struct inode *j_inode; 37323c816dedSEric Whitney unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 3733952fc18eSTheodore Ts'o ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3734952fc18eSTheodore Ts'o ext4_fsblk_t overhead = 0; 37354fdb5543SDmitry Monakhov char *buf = (char *) get_zeroed_page(GFP_NOFS); 3736952fc18eSTheodore Ts'o 3737952fc18eSTheodore Ts'o if (!buf) 3738952fc18eSTheodore Ts'o return -ENOMEM; 3739952fc18eSTheodore Ts'o 3740952fc18eSTheodore Ts'o /* 3741952fc18eSTheodore Ts'o * Compute the overhead (FS structures). This is constant 3742952fc18eSTheodore Ts'o * for a given filesystem unless the number of block groups 3743952fc18eSTheodore Ts'o * changes so we cache the previous value until it does. 3744952fc18eSTheodore Ts'o */ 3745952fc18eSTheodore Ts'o 3746952fc18eSTheodore Ts'o /* 3747952fc18eSTheodore Ts'o * All of the blocks before first_data_block are overhead 3748952fc18eSTheodore Ts'o */ 3749952fc18eSTheodore Ts'o overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3750952fc18eSTheodore Ts'o 3751952fc18eSTheodore Ts'o /* 3752952fc18eSTheodore Ts'o * Add the overhead found in each block group 3753952fc18eSTheodore Ts'o */ 3754952fc18eSTheodore Ts'o for (i = 0; i < ngroups; i++) { 3755952fc18eSTheodore Ts'o int blks; 3756952fc18eSTheodore Ts'o 3757952fc18eSTheodore Ts'o blks = count_overhead(sb, i, buf); 3758952fc18eSTheodore Ts'o overhead += blks; 3759952fc18eSTheodore Ts'o if (blks) 3760952fc18eSTheodore Ts'o memset(buf, 0, PAGE_SIZE); 3761952fc18eSTheodore Ts'o cond_resched(); 3762952fc18eSTheodore Ts'o } 37633c816dedSEric Whitney 37643c816dedSEric Whitney /* 37653c816dedSEric Whitney * Add the internal journal blocks whether the journal has been 37663c816dedSEric Whitney * loaded or not 37673c816dedSEric Whitney */ 3768ee7ed3aaSChunguang Xu if (sbi->s_journal && !sbi->s_journal_bdev) 3769810da240SLukas Czerner overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen); 3770f1eec3b0SRitesh Harjani else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) { 3771f1eec3b0SRitesh Harjani /* j_inum for internal journal is non-zero */ 37723c816dedSEric Whitney j_inode = ext4_get_journal_inode(sb, j_inum); 37733c816dedSEric Whitney if (j_inode) { 37743c816dedSEric Whitney j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 37753c816dedSEric Whitney overhead += EXT4_NUM_B2C(sbi, j_blocks); 37763c816dedSEric Whitney iput(j_inode); 37773c816dedSEric Whitney } else { 37783c816dedSEric Whitney ext4_msg(sb, KERN_ERR, "can't get journal size"); 37793c816dedSEric Whitney } 37803c816dedSEric Whitney } 3781952fc18eSTheodore Ts'o sbi->s_overhead = overhead; 3782952fc18eSTheodore Ts'o smp_wmb(); 3783952fc18eSTheodore Ts'o free_page((unsigned long) buf); 3784952fc18eSTheodore Ts'o return 0; 3785952fc18eSTheodore Ts'o } 3786952fc18eSTheodore Ts'o 3787b5799018STheodore Ts'o static void ext4_set_resv_clusters(struct super_block *sb) 378827dd4385SLukas Czerner { 378927dd4385SLukas Czerner ext4_fsblk_t resv_clusters; 3790b5799018STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 379127dd4385SLukas Czerner 379227dd4385SLukas Czerner /* 379330fac0f7SJan Kara * There's no need to reserve anything when we aren't using extents. 379430fac0f7SJan Kara * The space estimates are exact, there are no unwritten extents, 379530fac0f7SJan Kara * hole punching doesn't need new metadata... This is needed especially 379630fac0f7SJan Kara * to keep ext2/3 backward compatibility. 379730fac0f7SJan Kara */ 3798e2b911c5SDarrick J. Wong if (!ext4_has_feature_extents(sb)) 3799b5799018STheodore Ts'o return; 380030fac0f7SJan Kara /* 380127dd4385SLukas Czerner * By default we reserve 2% or 4096 clusters, whichever is smaller. 380227dd4385SLukas Czerner * This should cover the situations where we can not afford to run 380327dd4385SLukas Czerner * out of space like for example punch hole, or converting 3804556615dcSLukas Czerner * unwritten extents in delalloc path. In most cases such 380527dd4385SLukas Czerner * allocation would require 1, or 2 blocks, higher numbers are 380627dd4385SLukas Czerner * very rare. 380727dd4385SLukas Czerner */ 3808b5799018STheodore Ts'o resv_clusters = (ext4_blocks_count(sbi->s_es) >> 3809b5799018STheodore Ts'o sbi->s_cluster_bits); 381027dd4385SLukas Czerner 381127dd4385SLukas Czerner do_div(resv_clusters, 50); 381227dd4385SLukas Czerner resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 381327dd4385SLukas Czerner 3814b5799018STheodore Ts'o atomic64_set(&sbi->s_resv_clusters, resv_clusters); 381527dd4385SLukas Czerner } 381627dd4385SLukas Czerner 3817617ba13bSMingming Cao static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3818ac27a0ecSDave Kleikamp { 38195e405595SDan Williams struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev); 3820d4c402d9SCurt Wohlgemuth char *orig_data = kstrdup(data, GFP_KERNEL); 38211d0c3924STheodore Ts'o struct buffer_head *bh, **group_desc; 3822617ba13bSMingming Cao struct ext4_super_block *es = NULL; 38235aee0f8aSTheodore Ts'o struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 38247c990728SSuraj Jitindar Singh struct flex_groups **flex_groups; 3825617ba13bSMingming Cao ext4_fsblk_t block; 3826617ba13bSMingming Cao ext4_fsblk_t sb_block = get_sb_block(&data); 382770bbb3e0SAndrew Morton ext4_fsblk_t logical_sb_block; 3828ac27a0ecSDave Kleikamp unsigned long offset = 0; 3829ac27a0ecSDave Kleikamp unsigned long journal_devnum = 0; 3830ac27a0ecSDave Kleikamp unsigned long def_mount_opts; 3831ac27a0ecSDave Kleikamp struct inode *root; 38320390131bSFrank Mayhar const char *descr; 3833dcc7dae3SCyrill Gorcunov int ret = -ENOMEM; 3834281b5995STheodore Ts'o int blocksize, clustersize; 38354ec11028STheodore Ts'o unsigned int db_count; 38364ec11028STheodore Ts'o unsigned int i; 3837ef5fd681SKaixu Xia int needs_recovery, has_huge_files; 3838bd81d8eeSLaurent Vivier __u64 blocks_count; 383907aa2ea1SLukas Czerner int err = 0; 3840b3881f74STheodore Ts'o unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3841bfff6873SLukas Czerner ext4_group_t first_not_zeroed; 3842ac27a0ecSDave Kleikamp 38435aee0f8aSTheodore Ts'o if ((data && !orig_data) || !sbi) 38445aee0f8aSTheodore Ts'o goto out_free_base; 3845705895b6SPekka Enberg 3846aed9eb1bSColin Ian King sbi->s_daxdev = dax_dev; 3847705895b6SPekka Enberg sbi->s_blockgroup_lock = 3848705895b6SPekka Enberg kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 38495aee0f8aSTheodore Ts'o if (!sbi->s_blockgroup_lock) 38505aee0f8aSTheodore Ts'o goto out_free_base; 38515aee0f8aSTheodore Ts'o 3852ac27a0ecSDave Kleikamp sb->s_fs_info = sbi; 38532c0544b2STheodore Ts'o sbi->s_sb = sb; 3854240799cdSTheodore Ts'o sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3855d9c9bef1SMiklos Szeredi sbi->s_sb_block = sb_block; 3856f613dfcbSTheodore Ts'o if (sb->s_bdev->bd_part) 3857f613dfcbSTheodore Ts'o sbi->s_sectors_written_start = 3858dbae2c55SMichael Callahan part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]); 3859ac27a0ecSDave Kleikamp 38609f6200bbSTheodore Ts'o /* Cleanup superblock name */ 3861ec3904dcSRasmus Villemoes strreplace(sb->s_id, '/', '!'); 38629f6200bbSTheodore Ts'o 386307aa2ea1SLukas Czerner /* -EINVAL is default */ 3864dcc7dae3SCyrill Gorcunov ret = -EINVAL; 3865617ba13bSMingming Cao blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3866ac27a0ecSDave Kleikamp if (!blocksize) { 3867b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3868ac27a0ecSDave Kleikamp goto out_fail; 3869ac27a0ecSDave Kleikamp } 3870ac27a0ecSDave Kleikamp 3871ac27a0ecSDave Kleikamp /* 3872617ba13bSMingming Cao * The ext4 superblock will not be buffer aligned for other than 1kB 3873ac27a0ecSDave Kleikamp * block sizes. We need to calculate the offset from buffer start. 3874ac27a0ecSDave Kleikamp */ 3875617ba13bSMingming Cao if (blocksize != EXT4_MIN_BLOCK_SIZE) { 387670bbb3e0SAndrew Morton logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 387770bbb3e0SAndrew Morton offset = do_div(logical_sb_block, blocksize); 3878ac27a0ecSDave Kleikamp } else { 387970bbb3e0SAndrew Morton logical_sb_block = sb_block; 3880ac27a0ecSDave Kleikamp } 3881ac27a0ecSDave Kleikamp 3882a8ac900bSGioh Kim if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) { 3883b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "unable to read superblock"); 3884ac27a0ecSDave Kleikamp goto out_fail; 3885ac27a0ecSDave Kleikamp } 3886ac27a0ecSDave Kleikamp /* 3887ac27a0ecSDave Kleikamp * Note: s_es must be initialized as soon as possible because 3888617ba13bSMingming Cao * some ext4 macro-instructions depend on its value 3889ac27a0ecSDave Kleikamp */ 38902716b802STheodore Ts'o es = (struct ext4_super_block *) (bh->b_data + offset); 3891ac27a0ecSDave Kleikamp sbi->s_es = es; 3892ac27a0ecSDave Kleikamp sb->s_magic = le16_to_cpu(es->s_magic); 3893617ba13bSMingming Cao if (sb->s_magic != EXT4_SUPER_MAGIC) 3894617ba13bSMingming Cao goto cantfind_ext4; 3895afc32f7eSTheodore Ts'o sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3896ac27a0ecSDave Kleikamp 3897feb0ab32SDarrick J. Wong /* Warn if metadata_csum and gdt_csum are both set. */ 3898e2b911c5SDarrick J. Wong if (ext4_has_feature_metadata_csum(sb) && 3899e2b911c5SDarrick J. Wong ext4_has_feature_gdt_csum(sb)) 3900363307e6SJakub Wilk ext4_warning(sb, "metadata_csum and uninit_bg are " 3901feb0ab32SDarrick J. Wong "redundant flags; please run fsck."); 3902feb0ab32SDarrick J. Wong 3903d25425f8SDarrick J. Wong /* Check for a known checksum algorithm */ 3904d25425f8SDarrick J. Wong if (!ext4_verify_csum_type(sb, es)) { 3905d25425f8SDarrick J. Wong ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3906d25425f8SDarrick J. Wong "unknown checksum algorithm."); 3907d25425f8SDarrick J. Wong silent = 1; 3908d25425f8SDarrick J. Wong goto cantfind_ext4; 3909d25425f8SDarrick J. Wong } 3910d25425f8SDarrick J. Wong 39110441984aSDarrick J. Wong /* Load the checksum driver */ 39120441984aSDarrick J. Wong sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 39130441984aSDarrick J. Wong if (IS_ERR(sbi->s_chksum_driver)) { 39140441984aSDarrick J. Wong ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 39150441984aSDarrick J. Wong ret = PTR_ERR(sbi->s_chksum_driver); 39160441984aSDarrick J. Wong sbi->s_chksum_driver = NULL; 39170441984aSDarrick J. Wong goto failed_mount; 39180441984aSDarrick J. Wong } 39190441984aSDarrick J. Wong 3920a9c47317SDarrick J. Wong /* Check superblock checksum */ 3921a9c47317SDarrick J. Wong if (!ext4_superblock_csum_verify(sb, es)) { 3922a9c47317SDarrick J. Wong ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3923a9c47317SDarrick J. Wong "invalid superblock checksum. Run e2fsck?"); 3924a9c47317SDarrick J. Wong silent = 1; 39256a797d27SDarrick J. Wong ret = -EFSBADCRC; 3926a9c47317SDarrick J. Wong goto cantfind_ext4; 3927a9c47317SDarrick J. Wong } 3928a9c47317SDarrick J. Wong 3929a9c47317SDarrick J. Wong /* Precompute checksum seed for all metadata */ 3930e2b911c5SDarrick J. Wong if (ext4_has_feature_csum_seed(sb)) 39318c81bd8fSDarrick J. Wong sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 3932dec214d0STahsin Erdogan else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) 3933a9c47317SDarrick J. Wong sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 3934a9c47317SDarrick J. Wong sizeof(es->s_uuid)); 3935a9c47317SDarrick J. Wong 3936ac27a0ecSDave Kleikamp /* Set defaults before we parse the mount options */ 3937ac27a0ecSDave Kleikamp def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3938fd8c37ecSTheodore Ts'o set_opt(sb, INIT_INODE_TABLE); 3939617ba13bSMingming Cao if (def_mount_opts & EXT4_DEFM_DEBUG) 3940fd8c37ecSTheodore Ts'o set_opt(sb, DEBUG); 394187f26807STheodore Ts'o if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 3942fd8c37ecSTheodore Ts'o set_opt(sb, GRPID); 3943617ba13bSMingming Cao if (def_mount_opts & EXT4_DEFM_UID16) 3944fd8c37ecSTheodore Ts'o set_opt(sb, NO_UID32); 3945ea663336SEric Sandeen /* xattr user namespace & acls are now defaulted on */ 3946fd8c37ecSTheodore Ts'o set_opt(sb, XATTR_USER); 394703010a33STheodore Ts'o #ifdef CONFIG_EXT4_FS_POSIX_ACL 3948fd8c37ecSTheodore Ts'o set_opt(sb, POSIX_ACL); 39492e7842b8SHugh Dickins #endif 395098c1a759SDarrick J. Wong /* don't forget to enable journal_csum when metadata_csum is enabled. */ 395198c1a759SDarrick J. Wong if (ext4_has_metadata_csum(sb)) 395298c1a759SDarrick J. Wong set_opt(sb, JOURNAL_CHECKSUM); 395398c1a759SDarrick J. Wong 3954617ba13bSMingming Cao if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 3955fd8c37ecSTheodore Ts'o set_opt(sb, JOURNAL_DATA); 3956617ba13bSMingming Cao else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 3957fd8c37ecSTheodore Ts'o set_opt(sb, ORDERED_DATA); 3958617ba13bSMingming Cao else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 3959fd8c37ecSTheodore Ts'o set_opt(sb, WRITEBACK_DATA); 3960ac27a0ecSDave Kleikamp 3961617ba13bSMingming Cao if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 3962fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_PANIC); 3963bb4f397aSAneesh Kumar K.V else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 3964fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_CONT); 3965bb4f397aSAneesh Kumar K.V else 3966fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_RO); 396745f1a9c3SDarrick J. Wong /* block_validity enabled by default; disable with noblock_validity */ 3968fd8c37ecSTheodore Ts'o set_opt(sb, BLOCK_VALIDITY); 39698b67f04aSTheodore Ts'o if (def_mount_opts & EXT4_DEFM_DISCARD) 3970fd8c37ecSTheodore Ts'o set_opt(sb, DISCARD); 3971ac27a0ecSDave Kleikamp 397208cefc7aSEric W. Biederman sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 397308cefc7aSEric W. Biederman sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 397430773840STheodore Ts'o sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 397530773840STheodore Ts'o sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 397630773840STheodore Ts'o sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3977ac27a0ecSDave Kleikamp 39788b67f04aSTheodore Ts'o if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 3979fd8c37ecSTheodore Ts'o set_opt(sb, BARRIER); 3980ac27a0ecSDave Kleikamp 39811e2462f9SMingming Cao /* 3982dd919b98SAneesh Kumar K.V * enable delayed allocation by default 3983dd919b98SAneesh Kumar K.V * Use -o nodelalloc to turn it off 3984dd919b98SAneesh Kumar K.V */ 3985bc0b75f7STheodore Ts'o if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 39868b67f04aSTheodore Ts'o ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 3987fd8c37ecSTheodore Ts'o set_opt(sb, DELALLOC); 3988dd919b98SAneesh Kumar K.V 398951ce6511SLukas Czerner /* 399051ce6511SLukas Czerner * set default s_li_wait_mult for lazyinit, for the case there is 399151ce6511SLukas Czerner * no mount option specified. 399251ce6511SLukas Czerner */ 399351ce6511SLukas Czerner sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 399451ce6511SLukas Czerner 39954f97a681STheodore Ts'o blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 3996626b035bSRitesh Harjani 3997626b035bSRitesh Harjani if (blocksize == PAGE_SIZE) 3998626b035bSRitesh Harjani set_opt(sb, DIOREAD_NOLOCK); 3999626b035bSRitesh Harjani 40004f97a681STheodore Ts'o if (blocksize < EXT4_MIN_BLOCK_SIZE || 40014f97a681STheodore Ts'o blocksize > EXT4_MAX_BLOCK_SIZE) { 40024f97a681STheodore Ts'o ext4_msg(sb, KERN_ERR, 40034f97a681STheodore Ts'o "Unsupported filesystem blocksize %d (%d log_block_size)", 40044f97a681STheodore Ts'o blocksize, le32_to_cpu(es->s_log_block_size)); 40054f97a681STheodore Ts'o goto failed_mount; 40064f97a681STheodore Ts'o } 40074f97a681STheodore Ts'o 40089803387cSTheodore Ts'o if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 40099803387cSTheodore Ts'o sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 40109803387cSTheodore Ts'o sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 40119803387cSTheodore Ts'o } else { 40129803387cSTheodore Ts'o sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 40139803387cSTheodore Ts'o sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 40149803387cSTheodore Ts'o if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 40159803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 40169803387cSTheodore Ts'o sbi->s_first_ino); 40179803387cSTheodore Ts'o goto failed_mount; 40189803387cSTheodore Ts'o } 40199803387cSTheodore Ts'o if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 40209803387cSTheodore Ts'o (!is_power_of_2(sbi->s_inode_size)) || 40219803387cSTheodore Ts'o (sbi->s_inode_size > blocksize)) { 40229803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 40239803387cSTheodore Ts'o "unsupported inode size: %d", 40249803387cSTheodore Ts'o sbi->s_inode_size); 40254f97a681STheodore Ts'o ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize); 40269803387cSTheodore Ts'o goto failed_mount; 40279803387cSTheodore Ts'o } 40289803387cSTheodore Ts'o /* 40299803387cSTheodore Ts'o * i_atime_extra is the last extra field available for 40309803387cSTheodore Ts'o * [acm]times in struct ext4_inode. Checking for that 40319803387cSTheodore Ts'o * field should suffice to ensure we have extra space 40329803387cSTheodore Ts'o * for all three. 40339803387cSTheodore Ts'o */ 40349803387cSTheodore Ts'o if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) + 40359803387cSTheodore Ts'o sizeof(((struct ext4_inode *)0)->i_atime_extra)) { 40369803387cSTheodore Ts'o sb->s_time_gran = 1; 40379803387cSTheodore Ts'o sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX; 40389803387cSTheodore Ts'o } else { 40399803387cSTheodore Ts'o sb->s_time_gran = NSEC_PER_SEC; 40409803387cSTheodore Ts'o sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX; 40419803387cSTheodore Ts'o } 40429803387cSTheodore Ts'o sb->s_time_min = EXT4_TIMESTAMP_MIN; 40439803387cSTheodore Ts'o } 40449803387cSTheodore Ts'o if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 40459803387cSTheodore Ts'o sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 40469803387cSTheodore Ts'o EXT4_GOOD_OLD_INODE_SIZE; 40479803387cSTheodore Ts'o if (ext4_has_feature_extra_isize(sb)) { 40489803387cSTheodore Ts'o unsigned v, max = (sbi->s_inode_size - 40499803387cSTheodore Ts'o EXT4_GOOD_OLD_INODE_SIZE); 40509803387cSTheodore Ts'o 40519803387cSTheodore Ts'o v = le16_to_cpu(es->s_want_extra_isize); 40529803387cSTheodore Ts'o if (v > max) { 40539803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 40549803387cSTheodore Ts'o "bad s_want_extra_isize: %d", v); 40559803387cSTheodore Ts'o goto failed_mount; 40569803387cSTheodore Ts'o } 40579803387cSTheodore Ts'o if (sbi->s_want_extra_isize < v) 40589803387cSTheodore Ts'o sbi->s_want_extra_isize = v; 40599803387cSTheodore Ts'o 40609803387cSTheodore Ts'o v = le16_to_cpu(es->s_min_extra_isize); 40619803387cSTheodore Ts'o if (v > max) { 40629803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 40639803387cSTheodore Ts'o "bad s_min_extra_isize: %d", v); 40649803387cSTheodore Ts'o goto failed_mount; 40659803387cSTheodore Ts'o } 40669803387cSTheodore Ts'o if (sbi->s_want_extra_isize < v) 40679803387cSTheodore Ts'o sbi->s_want_extra_isize = v; 40689803387cSTheodore Ts'o } 40699803387cSTheodore Ts'o } 40709803387cSTheodore Ts'o 40715aee0f8aSTheodore Ts'o if (sbi->s_es->s_mount_opts[0]) { 40725aee0f8aSTheodore Ts'o char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, 40735aee0f8aSTheodore Ts'o sizeof(sbi->s_es->s_mount_opts), 40745aee0f8aSTheodore Ts'o GFP_KERNEL); 40755aee0f8aSTheodore Ts'o if (!s_mount_opts) 40765aee0f8aSTheodore Ts'o goto failed_mount; 40775aee0f8aSTheodore Ts'o if (!parse_options(s_mount_opts, sb, &journal_devnum, 40785aee0f8aSTheodore Ts'o &journal_ioprio, 0)) { 40798b67f04aSTheodore Ts'o ext4_msg(sb, KERN_WARNING, 40808b67f04aSTheodore Ts'o "failed to parse options in superblock: %s", 40815aee0f8aSTheodore Ts'o s_mount_opts); 40825aee0f8aSTheodore Ts'o } 40835aee0f8aSTheodore Ts'o kfree(s_mount_opts); 40848b67f04aSTheodore Ts'o } 40855a916be1STheodore Ts'o sbi->s_def_mount_opt = sbi->s_mount_opt; 4086b3881f74STheodore Ts'o if (!parse_options((char *) data, sb, &journal_devnum, 4087661aa520SEric Sandeen &journal_ioprio, 0)) 4088ac27a0ecSDave Kleikamp goto failed_mount; 4089ac27a0ecSDave Kleikamp 4090c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 4091c83ad55eSGabriel Krisman Bertazi if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) { 4092c83ad55eSGabriel Krisman Bertazi const struct ext4_sb_encodings *encoding_info; 4093c83ad55eSGabriel Krisman Bertazi struct unicode_map *encoding; 4094c83ad55eSGabriel Krisman Bertazi __u16 encoding_flags; 4095c83ad55eSGabriel Krisman Bertazi 4096c83ad55eSGabriel Krisman Bertazi if (ext4_has_feature_encrypt(sb)) { 4097c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 4098c83ad55eSGabriel Krisman Bertazi "Can't mount with encoding and encryption"); 4099c83ad55eSGabriel Krisman Bertazi goto failed_mount; 4100c83ad55eSGabriel Krisman Bertazi } 4101c83ad55eSGabriel Krisman Bertazi 4102c83ad55eSGabriel Krisman Bertazi if (ext4_sb_read_encoding(es, &encoding_info, 4103c83ad55eSGabriel Krisman Bertazi &encoding_flags)) { 4104c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 4105c83ad55eSGabriel Krisman Bertazi "Encoding requested by superblock is unknown"); 4106c83ad55eSGabriel Krisman Bertazi goto failed_mount; 4107c83ad55eSGabriel Krisman Bertazi } 4108c83ad55eSGabriel Krisman Bertazi 4109c83ad55eSGabriel Krisman Bertazi encoding = utf8_load(encoding_info->version); 4110c83ad55eSGabriel Krisman Bertazi if (IS_ERR(encoding)) { 4111c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 4112c83ad55eSGabriel Krisman Bertazi "can't mount with superblock charset: %s-%s " 4113c83ad55eSGabriel Krisman Bertazi "not supported by the kernel. flags: 0x%x.", 4114c83ad55eSGabriel Krisman Bertazi encoding_info->name, encoding_info->version, 4115c83ad55eSGabriel Krisman Bertazi encoding_flags); 4116c83ad55eSGabriel Krisman Bertazi goto failed_mount; 4117c83ad55eSGabriel Krisman Bertazi } 4118c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " 4119c83ad55eSGabriel Krisman Bertazi "%s-%s with flags 0x%hx", encoding_info->name, 4120c83ad55eSGabriel Krisman Bertazi encoding_info->version?:"\b", encoding_flags); 4121c83ad55eSGabriel Krisman Bertazi 4122c83ad55eSGabriel Krisman Bertazi sbi->s_encoding = encoding; 4123c83ad55eSGabriel Krisman Bertazi sbi->s_encoding_flags = encoding_flags; 4124c83ad55eSGabriel Krisman Bertazi } 4125c83ad55eSGabriel Krisman Bertazi #endif 4126c83ad55eSGabriel Krisman Bertazi 412756889787STheodore Ts'o if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 4128244adf64STheodore Ts'o printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n"); 4129781c036bSKaixu Xia /* can't mount with both data=journal and dioread_nolock. */ 4130244adf64STheodore Ts'o clear_opt(sb, DIOREAD_NOLOCK); 413156889787STheodore Ts'o if (test_opt2(sb, EXPLICIT_DELALLOC)) { 413256889787STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't mount with " 413356889787STheodore Ts'o "both data=journal and delalloc"); 413456889787STheodore Ts'o goto failed_mount; 413556889787STheodore Ts'o } 4136fc626fe3SIra Weiny if (test_opt(sb, DAX_ALWAYS)) { 4137923ae0ffSRoss Zwisler ext4_msg(sb, KERN_ERR, "can't mount with " 4138923ae0ffSRoss Zwisler "both data=journal and dax"); 4139923ae0ffSRoss Zwisler goto failed_mount; 4140923ae0ffSRoss Zwisler } 414173b92a2aSSergey Karamov if (ext4_has_feature_encrypt(sb)) { 414273b92a2aSSergey Karamov ext4_msg(sb, KERN_WARNING, 414373b92a2aSSergey Karamov "encrypted files will use data=ordered " 414473b92a2aSSergey Karamov "instead of data journaling mode"); 414573b92a2aSSergey Karamov } 414656889787STheodore Ts'o if (test_opt(sb, DELALLOC)) 414756889787STheodore Ts'o clear_opt(sb, DELALLOC); 4148001e4a87STejun Heo } else { 4149001e4a87STejun Heo sb->s_iflags |= SB_I_CGROUPWB; 415056889787STheodore Ts'o } 415156889787STheodore Ts'o 41521751e8a6SLinus Torvalds sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 41531751e8a6SLinus Torvalds (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 4154ac27a0ecSDave Kleikamp 4155617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 4156e2b911c5SDarrick J. Wong (ext4_has_compat_features(sb) || 4157e2b911c5SDarrick J. Wong ext4_has_ro_compat_features(sb) || 4158e2b911c5SDarrick J. Wong ext4_has_incompat_features(sb))) 4159b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 4160b31e1552SEric Sandeen "feature flags set on rev 0 fs, " 4161b31e1552SEric Sandeen "running e2fsck is recommended"); 4162469108ffSTheodore Tso 4163ed3654ebSTheodore Ts'o if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 4164ed3654ebSTheodore Ts'o set_opt2(sb, HURD_COMPAT); 4165e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb)) { 4166ed3654ebSTheodore Ts'o ext4_msg(sb, KERN_ERR, 4167ed3654ebSTheodore Ts'o "The Hurd can't support 64-bit file systems"); 4168ed3654ebSTheodore Ts'o goto failed_mount; 4169ed3654ebSTheodore Ts'o } 4170dec214d0STahsin Erdogan 4171dec214d0STahsin Erdogan /* 4172dec214d0STahsin Erdogan * ea_inode feature uses l_i_version field which is not 4173dec214d0STahsin Erdogan * available in HURD_COMPAT mode. 4174dec214d0STahsin Erdogan */ 4175dec214d0STahsin Erdogan if (ext4_has_feature_ea_inode(sb)) { 4176dec214d0STahsin Erdogan ext4_msg(sb, KERN_ERR, 4177dec214d0STahsin Erdogan "ea_inode feature is not supported for Hurd"); 4178dec214d0STahsin Erdogan goto failed_mount; 4179dec214d0STahsin Erdogan } 4180ed3654ebSTheodore Ts'o } 4181ed3654ebSTheodore Ts'o 41822035e776STheodore Ts'o if (IS_EXT2_SB(sb)) { 41832035e776STheodore Ts'o if (ext2_feature_set_ok(sb)) 41842035e776STheodore Ts'o ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 41852035e776STheodore Ts'o "using the ext4 subsystem"); 41862035e776STheodore Ts'o else { 41870d9366d6SEric Sandeen /* 41880d9366d6SEric Sandeen * If we're probing be silent, if this looks like 41890d9366d6SEric Sandeen * it's actually an ext[34] filesystem. 41900d9366d6SEric Sandeen */ 41910d9366d6SEric Sandeen if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 41920d9366d6SEric Sandeen goto failed_mount; 41932035e776STheodore Ts'o ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 41942035e776STheodore Ts'o "to feature incompatibilities"); 41952035e776STheodore Ts'o goto failed_mount; 41962035e776STheodore Ts'o } 41972035e776STheodore Ts'o } 41982035e776STheodore Ts'o 41992035e776STheodore Ts'o if (IS_EXT3_SB(sb)) { 42002035e776STheodore Ts'o if (ext3_feature_set_ok(sb)) 42012035e776STheodore Ts'o ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 42022035e776STheodore Ts'o "using the ext4 subsystem"); 42032035e776STheodore Ts'o else { 42040d9366d6SEric Sandeen /* 42050d9366d6SEric Sandeen * If we're probing be silent, if this looks like 42060d9366d6SEric Sandeen * it's actually an ext4 filesystem. 42070d9366d6SEric Sandeen */ 42080d9366d6SEric Sandeen if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 42090d9366d6SEric Sandeen goto failed_mount; 42102035e776STheodore Ts'o ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 42112035e776STheodore Ts'o "to feature incompatibilities"); 42122035e776STheodore Ts'o goto failed_mount; 42132035e776STheodore Ts'o } 42142035e776STheodore Ts'o } 42152035e776STheodore Ts'o 4216469108ffSTheodore Tso /* 4217ac27a0ecSDave Kleikamp * Check feature flags regardless of the revision level, since we 4218ac27a0ecSDave Kleikamp * previously didn't change the revision level when setting the flags, 4219ac27a0ecSDave Kleikamp * so there is a chance incompat flags are set on a rev 0 filesystem. 4220ac27a0ecSDave Kleikamp */ 4221bc98a42cSDavid Howells if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 4222ac27a0ecSDave Kleikamp goto failed_mount; 4223a13fb1a4SEric Sandeen 42248cdf3372STheodore Ts'o if (le32_to_cpu(es->s_log_block_size) > 42258cdf3372STheodore Ts'o (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 42268cdf3372STheodore Ts'o ext4_msg(sb, KERN_ERR, 42278cdf3372STheodore Ts'o "Invalid log block size: %u", 42288cdf3372STheodore Ts'o le32_to_cpu(es->s_log_block_size)); 4229ac27a0ecSDave Kleikamp goto failed_mount; 4230ac27a0ecSDave Kleikamp } 4231bfe0a5f4STheodore Ts'o if (le32_to_cpu(es->s_log_cluster_size) > 4232bfe0a5f4STheodore Ts'o (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 4233bfe0a5f4STheodore Ts'o ext4_msg(sb, KERN_ERR, 4234bfe0a5f4STheodore Ts'o "Invalid log cluster size: %u", 4235bfe0a5f4STheodore Ts'o le32_to_cpu(es->s_log_cluster_size)); 4236bfe0a5f4STheodore Ts'o goto failed_mount; 4237bfe0a5f4STheodore Ts'o } 4238ac27a0ecSDave Kleikamp 42395b9554dcSTheodore Ts'o if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { 42405b9554dcSTheodore Ts'o ext4_msg(sb, KERN_ERR, 42415b9554dcSTheodore Ts'o "Number of reserved GDT blocks insanely large: %d", 42425b9554dcSTheodore Ts'o le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 42435b9554dcSTheodore Ts'o goto failed_mount; 42445b9554dcSTheodore Ts'o } 42455b9554dcSTheodore Ts'o 4246a8ab6d38SIra Weiny if (bdev_dax_supported(sb->s_bdev, blocksize)) 4247a8ab6d38SIra Weiny set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags); 4248a8ab6d38SIra Weiny 4249fc626fe3SIra Weiny if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) { 4250559db4c6SRoss Zwisler if (ext4_has_feature_inline_data(sb)) { 4251559db4c6SRoss Zwisler ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 4252559db4c6SRoss Zwisler " that may contain inline data"); 4253361d24d4SEric Sandeen goto failed_mount; 4254559db4c6SRoss Zwisler } 4255a8ab6d38SIra Weiny if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) { 425624f3478dSDan Williams ext4_msg(sb, KERN_ERR, 4257361d24d4SEric Sandeen "DAX unsupported by block device."); 4258361d24d4SEric Sandeen goto failed_mount; 425924f3478dSDan Williams } 4260923ae0ffSRoss Zwisler } 4261923ae0ffSRoss Zwisler 4262e2b911c5SDarrick J. Wong if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 42636ddb2447STheodore Ts'o ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 42646ddb2447STheodore Ts'o es->s_encryption_level); 42656ddb2447STheodore Ts'o goto failed_mount; 42666ddb2447STheodore Ts'o } 42676ddb2447STheodore Ts'o 4268ac27a0ecSDave Kleikamp if (sb->s_blocksize != blocksize) { 4269ce40733cSAneesh Kumar K.V /* Validate the filesystem blocksize */ 4270ce40733cSAneesh Kumar K.V if (!sb_set_blocksize(sb, blocksize)) { 4271b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "bad block size %d", 4272ce40733cSAneesh Kumar K.V blocksize); 4273ac27a0ecSDave Kleikamp goto failed_mount; 4274ac27a0ecSDave Kleikamp } 4275ac27a0ecSDave Kleikamp 4276ac27a0ecSDave Kleikamp brelse(bh); 427770bbb3e0SAndrew Morton logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 427870bbb3e0SAndrew Morton offset = do_div(logical_sb_block, blocksize); 4279a8ac900bSGioh Kim bh = sb_bread_unmovable(sb, logical_sb_block); 4280ac27a0ecSDave Kleikamp if (!bh) { 4281b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4282b31e1552SEric Sandeen "Can't read superblock on 2nd try"); 4283ac27a0ecSDave Kleikamp goto failed_mount; 4284ac27a0ecSDave Kleikamp } 42852716b802STheodore Ts'o es = (struct ext4_super_block *)(bh->b_data + offset); 4286ac27a0ecSDave Kleikamp sbi->s_es = es; 4287617ba13bSMingming Cao if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 4288b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4289b31e1552SEric Sandeen "Magic mismatch, very weird!"); 4290ac27a0ecSDave Kleikamp goto failed_mount; 4291ac27a0ecSDave Kleikamp } 4292ac27a0ecSDave Kleikamp } 4293ac27a0ecSDave Kleikamp 4294e2b911c5SDarrick J. Wong has_huge_files = ext4_has_feature_huge_file(sb); 4295f287a1a5STheodore Ts'o sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 4296f287a1a5STheodore Ts'o has_huge_files); 4297f287a1a5STheodore Ts'o sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 4298ac27a0ecSDave Kleikamp 42990d1ee42fSAlexandre Ratchov sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 4300e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb)) { 43018fadc143SAlexandre Ratchov if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 43020d1ee42fSAlexandre Ratchov sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 4303d8ea6cf8Svignesh babu !is_power_of_2(sbi->s_desc_size)) { 4304b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4305b31e1552SEric Sandeen "unsupported descriptor size %lu", 43060d1ee42fSAlexandre Ratchov sbi->s_desc_size); 43070d1ee42fSAlexandre Ratchov goto failed_mount; 43080d1ee42fSAlexandre Ratchov } 43090d1ee42fSAlexandre Ratchov } else 43100d1ee42fSAlexandre Ratchov sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 43110b8e58a1SAndreas Dilger 4312ac27a0ecSDave Kleikamp sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 4313ac27a0ecSDave Kleikamp sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 43140b8e58a1SAndreas Dilger 4315617ba13bSMingming Cao sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 4316ac27a0ecSDave Kleikamp if (sbi->s_inodes_per_block == 0) 4317617ba13bSMingming Cao goto cantfind_ext4; 4318cd6bb35bSTheodore Ts'o if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 4319cd6bb35bSTheodore Ts'o sbi->s_inodes_per_group > blocksize * 8) { 4320cd6bb35bSTheodore Ts'o ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 4321b9c538daSJosh Triplett sbi->s_inodes_per_group); 4322cd6bb35bSTheodore Ts'o goto failed_mount; 4323cd6bb35bSTheodore Ts'o } 4324ac27a0ecSDave Kleikamp sbi->s_itb_per_group = sbi->s_inodes_per_group / 4325ac27a0ecSDave Kleikamp sbi->s_inodes_per_block; 43260d1ee42fSAlexandre Ratchov sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 4327ac27a0ecSDave Kleikamp sbi->s_sbh = bh; 4328ac27a0ecSDave Kleikamp sbi->s_mount_state = le16_to_cpu(es->s_state); 4329e57aa839SFengguang Wu sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 4330e57aa839SFengguang Wu sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 43310b8e58a1SAndreas Dilger 4332ac27a0ecSDave Kleikamp for (i = 0; i < 4; i++) 4333ac27a0ecSDave Kleikamp sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 4334ac27a0ecSDave Kleikamp sbi->s_def_hash_version = es->s_def_hash_version; 4335e2b911c5SDarrick J. Wong if (ext4_has_feature_dir_index(sb)) { 4336f99b2589STheodore Ts'o i = le32_to_cpu(es->s_flags); 4337f99b2589STheodore Ts'o if (i & EXT2_FLAGS_UNSIGNED_HASH) 4338f99b2589STheodore Ts'o sbi->s_hash_unsigned = 3; 4339f99b2589STheodore Ts'o else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 4340f99b2589STheodore Ts'o #ifdef __CHAR_UNSIGNED__ 4341bc98a42cSDavid Howells if (!sb_rdonly(sb)) 434223301410STheodore Ts'o es->s_flags |= 434323301410STheodore Ts'o cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 4344f99b2589STheodore Ts'o sbi->s_hash_unsigned = 3; 4345f99b2589STheodore Ts'o #else 4346bc98a42cSDavid Howells if (!sb_rdonly(sb)) 434723301410STheodore Ts'o es->s_flags |= 434823301410STheodore Ts'o cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 4349f99b2589STheodore Ts'o #endif 4350f99b2589STheodore Ts'o } 435123301410STheodore Ts'o } 4352ac27a0ecSDave Kleikamp 4353281b5995STheodore Ts'o /* Handle clustersize */ 4354281b5995STheodore Ts'o clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 4355ef5fd681SKaixu Xia if (ext4_has_feature_bigalloc(sb)) { 4356281b5995STheodore Ts'o if (clustersize < blocksize) { 4357281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, 4358281b5995STheodore Ts'o "cluster size (%d) smaller than " 4359281b5995STheodore Ts'o "block size (%d)", clustersize, blocksize); 4360281b5995STheodore Ts'o goto failed_mount; 4361281b5995STheodore Ts'o } 4362281b5995STheodore Ts'o sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 4363281b5995STheodore Ts'o le32_to_cpu(es->s_log_block_size); 4364281b5995STheodore Ts'o sbi->s_clusters_per_group = 4365281b5995STheodore Ts'o le32_to_cpu(es->s_clusters_per_group); 4366281b5995STheodore Ts'o if (sbi->s_clusters_per_group > blocksize * 8) { 4367281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, 4368281b5995STheodore Ts'o "#clusters per group too big: %lu", 4369281b5995STheodore Ts'o sbi->s_clusters_per_group); 4370281b5995STheodore Ts'o goto failed_mount; 4371281b5995STheodore Ts'o } 4372281b5995STheodore Ts'o if (sbi->s_blocks_per_group != 4373281b5995STheodore Ts'o (sbi->s_clusters_per_group * (clustersize / blocksize))) { 4374281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 4375281b5995STheodore Ts'o "clusters per group (%lu) inconsistent", 4376281b5995STheodore Ts'o sbi->s_blocks_per_group, 4377281b5995STheodore Ts'o sbi->s_clusters_per_group); 4378281b5995STheodore Ts'o goto failed_mount; 4379281b5995STheodore Ts'o } 4380281b5995STheodore Ts'o } else { 4381281b5995STheodore Ts'o if (clustersize != blocksize) { 4382bfe0a5f4STheodore Ts'o ext4_msg(sb, KERN_ERR, 4383bfe0a5f4STheodore Ts'o "fragment/cluster size (%d) != " 4384bfe0a5f4STheodore Ts'o "block size (%d)", clustersize, blocksize); 4385bfe0a5f4STheodore Ts'o goto failed_mount; 4386281b5995STheodore Ts'o } 4387ac27a0ecSDave Kleikamp if (sbi->s_blocks_per_group > blocksize * 8) { 4388b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4389b31e1552SEric Sandeen "#blocks per group too big: %lu", 4390ac27a0ecSDave Kleikamp sbi->s_blocks_per_group); 4391ac27a0ecSDave Kleikamp goto failed_mount; 4392ac27a0ecSDave Kleikamp } 4393281b5995STheodore Ts'o sbi->s_clusters_per_group = sbi->s_blocks_per_group; 4394281b5995STheodore Ts'o sbi->s_cluster_bits = 0; 4395281b5995STheodore Ts'o } 4396281b5995STheodore Ts'o sbi->s_cluster_ratio = clustersize / blocksize; 4397281b5995STheodore Ts'o 4398960fd856STheodore Ts'o /* Do we have standard group size of clustersize * 8 blocks ? */ 4399960fd856STheodore Ts'o if (sbi->s_blocks_per_group == clustersize << 3) 4400960fd856STheodore Ts'o set_opt2(sb, STD_GROUP_SIZE); 4401960fd856STheodore Ts'o 4402bf43d84bSEric Sandeen /* 4403bf43d84bSEric Sandeen * Test whether we have more sectors than will fit in sector_t, 4404bf43d84bSEric Sandeen * and whether the max offset is addressable by the page cache. 4405bf43d84bSEric Sandeen */ 44065a9ae68aSDarrick J. Wong err = generic_check_addressable(sb->s_blocksize_bits, 440730ca22c7SPatrick J. LoPresti ext4_blocks_count(es)); 44085a9ae68aSDarrick J. Wong if (err) { 4409b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "filesystem" 4410bf43d84bSEric Sandeen " too large to mount safely on this system"); 4411ac27a0ecSDave Kleikamp goto failed_mount; 4412ac27a0ecSDave Kleikamp } 4413ac27a0ecSDave Kleikamp 4414617ba13bSMingming Cao if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 4415617ba13bSMingming Cao goto cantfind_ext4; 4416e7c95593SEric Sandeen 44170f2ddca6SFrom: Thiemo Nagel /* check blocks count against device size */ 44180f2ddca6SFrom: Thiemo Nagel blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 44190f2ddca6SFrom: Thiemo Nagel if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4420b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4421b31e1552SEric Sandeen "exceeds size of device (%llu blocks)", 44220f2ddca6SFrom: Thiemo Nagel ext4_blocks_count(es), blocks_count); 44230f2ddca6SFrom: Thiemo Nagel goto failed_mount; 44240f2ddca6SFrom: Thiemo Nagel } 44250f2ddca6SFrom: Thiemo Nagel 44264ec11028STheodore Ts'o /* 44274ec11028STheodore Ts'o * It makes no sense for the first data block to be beyond the end 44284ec11028STheodore Ts'o * of the filesystem. 44294ec11028STheodore Ts'o */ 44304ec11028STheodore Ts'o if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4431b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4432b31e1552SEric Sandeen "block %u is beyond end of filesystem (%llu)", 4433e7c95593SEric Sandeen le32_to_cpu(es->s_first_data_block), 44344ec11028STheodore Ts'o ext4_blocks_count(es)); 4435e7c95593SEric Sandeen goto failed_mount; 4436e7c95593SEric Sandeen } 4437bfe0a5f4STheodore Ts'o if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4438bfe0a5f4STheodore Ts'o (sbi->s_cluster_ratio == 1)) { 4439bfe0a5f4STheodore Ts'o ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4440bfe0a5f4STheodore Ts'o "block is 0 with a 1k block and cluster size"); 4441bfe0a5f4STheodore Ts'o goto failed_mount; 4442bfe0a5f4STheodore Ts'o } 4443bfe0a5f4STheodore Ts'o 4444bd81d8eeSLaurent Vivier blocks_count = (ext4_blocks_count(es) - 4445bd81d8eeSLaurent Vivier le32_to_cpu(es->s_first_data_block) + 4446bd81d8eeSLaurent Vivier EXT4_BLOCKS_PER_GROUP(sb) - 1); 4447bd81d8eeSLaurent Vivier do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 44484ec11028STheodore Ts'o if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4449df41460aSJosh Triplett ext4_msg(sb, KERN_WARNING, "groups count too large: %llu " 44504ec11028STheodore Ts'o "(block count %llu, first data block %u, " 4451df41460aSJosh Triplett "blocks per group %lu)", blocks_count, 44524ec11028STheodore Ts'o ext4_blocks_count(es), 44534ec11028STheodore Ts'o le32_to_cpu(es->s_first_data_block), 44544ec11028STheodore Ts'o EXT4_BLOCKS_PER_GROUP(sb)); 44554ec11028STheodore Ts'o goto failed_mount; 44564ec11028STheodore Ts'o } 4457bd81d8eeSLaurent Vivier sbi->s_groups_count = blocks_count; 4458fb0a387dSEric Sandeen sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 4459fb0a387dSEric Sandeen (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 44609e463084STheodore Ts'o if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 44619e463084STheodore Ts'o le32_to_cpu(es->s_inodes_count)) { 44629e463084STheodore Ts'o ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 44639e463084STheodore Ts'o le32_to_cpu(es->s_inodes_count), 44649e463084STheodore Ts'o ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 44659e463084STheodore Ts'o ret = -EINVAL; 44669e463084STheodore Ts'o goto failed_mount; 44679e463084STheodore Ts'o } 4468617ba13bSMingming Cao db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4469617ba13bSMingming Cao EXT4_DESC_PER_BLOCK(sb); 44703a4b77cdSEryu Guan if (ext4_has_feature_meta_bg(sb)) { 44712ba3e6e8STheodore Ts'o if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 44723a4b77cdSEryu Guan ext4_msg(sb, KERN_WARNING, 44733a4b77cdSEryu Guan "first meta block group too large: %u " 44743a4b77cdSEryu Guan "(group descriptor block count %u)", 44753a4b77cdSEryu Guan le32_to_cpu(es->s_first_meta_bg), db_count); 44763a4b77cdSEryu Guan goto failed_mount; 44773a4b77cdSEryu Guan } 44783a4b77cdSEryu Guan } 44791d0c3924STheodore Ts'o rcu_assign_pointer(sbi->s_group_desc, 44801d0c3924STheodore Ts'o kvmalloc_array(db_count, 4481f18a5f21STheodore Ts'o sizeof(struct buffer_head *), 44821d0c3924STheodore Ts'o GFP_KERNEL)); 4483ac27a0ecSDave Kleikamp if (sbi->s_group_desc == NULL) { 4484b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "not enough memory"); 44852cde417dSTheodore Ts'o ret = -ENOMEM; 4486ac27a0ecSDave Kleikamp goto failed_mount; 4487ac27a0ecSDave Kleikamp } 4488ac27a0ecSDave Kleikamp 4489705895b6SPekka Enberg bgl_lock_init(sbi->s_blockgroup_lock); 4490ac27a0ecSDave Kleikamp 449185c8f176SAndrew Perepechko /* Pre-read the descriptors into the buffer cache */ 449285c8f176SAndrew Perepechko for (i = 0; i < db_count; i++) { 449385c8f176SAndrew Perepechko block = descriptor_loc(sb, logical_sb_block, i); 4494d87f6392SRoman Gushchin sb_breadahead_unmovable(sb, block); 449585c8f176SAndrew Perepechko } 449685c8f176SAndrew Perepechko 4497ac27a0ecSDave Kleikamp for (i = 0; i < db_count; i++) { 44981d0c3924STheodore Ts'o struct buffer_head *bh; 44991d0c3924STheodore Ts'o 450070bbb3e0SAndrew Morton block = descriptor_loc(sb, logical_sb_block, i); 45011d0c3924STheodore Ts'o bh = sb_bread_unmovable(sb, block); 45021d0c3924STheodore Ts'o if (!bh) { 4503b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4504b31e1552SEric Sandeen "can't read group descriptor %d", i); 4505ac27a0ecSDave Kleikamp db_count = i; 4506ac27a0ecSDave Kleikamp goto failed_mount2; 4507ac27a0ecSDave Kleikamp } 45081d0c3924STheodore Ts'o rcu_read_lock(); 45091d0c3924STheodore Ts'o rcu_dereference(sbi->s_group_desc)[i] = bh; 45101d0c3924STheodore Ts'o rcu_read_unlock(); 4511ac27a0ecSDave Kleikamp } 451244de022cSTheodore Ts'o sbi->s_gdb_count = db_count; 4513829fa70dSTheodore Ts'o if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { 4514b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 45156a797d27SDarrick J. Wong ret = -EFSCORRUPTED; 4516f9ae9cf5STheodore Ts'o goto failed_mount2; 4517ac27a0ecSDave Kleikamp } 4518772cb7c8SJose R. Santos 4519235699a8SKees Cook timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 452004496411STao Ma 4521a75ae78fSDmitry Monakhov /* Register extent status tree shrinker */ 4522eb68d0e2SZheng Liu if (ext4_es_register_shrinker(sbi)) 4523ce7e010aSTheodore Ts'o goto failed_mount3; 4524ce7e010aSTheodore Ts'o 4525c9de560dSAlex Tomas sbi->s_stripe = ext4_get_stripe_size(sbi); 452667a5da56SZheng Liu sbi->s_extent_max_zeroout_kb = 32; 4527c9de560dSAlex Tomas 4528f9ae9cf5STheodore Ts'o /* 4529f9ae9cf5STheodore Ts'o * set up enough so that it can read an inode 4530f9ae9cf5STheodore Ts'o */ 4531f9ae9cf5STheodore Ts'o sb->s_op = &ext4_sops; 4532617ba13bSMingming Cao sb->s_export_op = &ext4_export_ops; 4533617ba13bSMingming Cao sb->s_xattr = ext4_xattr_handlers; 4534643fa961SChandan Rajendra #ifdef CONFIG_FS_ENCRYPTION 4535a7550b30SJaegeuk Kim sb->s_cop = &ext4_cryptops; 4536ffcc4182SEric Biggers #endif 4537c93d8f88SEric Biggers #ifdef CONFIG_FS_VERITY 4538c93d8f88SEric Biggers sb->s_vop = &ext4_verityops; 4539c93d8f88SEric Biggers #endif 4540ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4541617ba13bSMingming Cao sb->dq_op = &ext4_quota_operations; 4542e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) 45431fa5efe3SJan Kara sb->s_qcop = &dquot_quotactl_sysfile_ops; 4544262b4662SJan Kara else 4545262b4662SJan Kara sb->s_qcop = &ext4_qctl_operations; 4546689c958cSLi Xi sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4547ac27a0ecSDave Kleikamp #endif 454885787090SChristoph Hellwig memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 4549f2fa2ffcSAneesh Kumar K.V 4550ac27a0ecSDave Kleikamp INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 45513b9d4ed2STheodore Ts'o mutex_init(&sbi->s_orphan_lock); 4552ac27a0ecSDave Kleikamp 4553ac27a0ecSDave Kleikamp sb->s_root = NULL; 4554ac27a0ecSDave Kleikamp 4555ac27a0ecSDave Kleikamp needs_recovery = (es->s_last_orphan != 0 || 4556e2b911c5SDarrick J. Wong ext4_has_feature_journal_needs_recovery(sb)); 4557ac27a0ecSDave Kleikamp 4558bc98a42cSDavid Howells if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) 4559c5e06d10SJohann Lombardi if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 456050460fe8SDarrick J. Wong goto failed_mount3a; 4561c5e06d10SJohann Lombardi 4562ac27a0ecSDave Kleikamp /* 4563ac27a0ecSDave Kleikamp * The first inode we look at is the journal inode. Don't try 4564ac27a0ecSDave Kleikamp * root first: it may be modified in the journal! 4565ac27a0ecSDave Kleikamp */ 4566e2b911c5SDarrick J. Wong if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 45674753d8a2STheodore Ts'o err = ext4_load_journal(sb, es, journal_devnum); 45684753d8a2STheodore Ts'o if (err) 456950460fe8SDarrick J. Wong goto failed_mount3a; 4570bc98a42cSDavid Howells } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 4571e2b911c5SDarrick J. Wong ext4_has_feature_journal_needs_recovery(sb)) { 4572b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "required journal recovery " 4573b31e1552SEric Sandeen "suppressed and not mounted read-only"); 4574744692dcSJiaying Zhang goto failed_mount_wq; 4575ac27a0ecSDave Kleikamp } else { 45761e381f60SDmitry Monakhov /* Nojournal mode, all journal mount options are illegal */ 45771e381f60SDmitry Monakhov if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 45781e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 45791e381f60SDmitry Monakhov "journal_checksum, fs mounted w/o journal"); 45801e381f60SDmitry Monakhov goto failed_mount_wq; 45811e381f60SDmitry Monakhov } 45821e381f60SDmitry Monakhov if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 45831e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 45841e381f60SDmitry Monakhov "journal_async_commit, fs mounted w/o journal"); 45851e381f60SDmitry Monakhov goto failed_mount_wq; 45861e381f60SDmitry Monakhov } 45871e381f60SDmitry Monakhov if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 45881e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 45891e381f60SDmitry Monakhov "commit=%lu, fs mounted w/o journal", 45901e381f60SDmitry Monakhov sbi->s_commit_interval / HZ); 45911e381f60SDmitry Monakhov goto failed_mount_wq; 45921e381f60SDmitry Monakhov } 45931e381f60SDmitry Monakhov if (EXT4_MOUNT_DATA_FLAGS & 45941e381f60SDmitry Monakhov (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 45951e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 45961e381f60SDmitry Monakhov "data=, fs mounted w/o journal"); 45971e381f60SDmitry Monakhov goto failed_mount_wq; 45981e381f60SDmitry Monakhov } 459950b29d8fSDebabrata Banerjee sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; 46001e381f60SDmitry Monakhov clear_opt(sb, JOURNAL_CHECKSUM); 4601fd8c37ecSTheodore Ts'o clear_opt(sb, DATA_FLAGS); 46020390131bSFrank Mayhar sbi->s_journal = NULL; 46030390131bSFrank Mayhar needs_recovery = 0; 46040390131bSFrank Mayhar goto no_journal; 4605ac27a0ecSDave Kleikamp } 4606ac27a0ecSDave Kleikamp 4607e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb) && 4608eb40a09cSJose R. Santos !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4609eb40a09cSJose R. Santos JBD2_FEATURE_INCOMPAT_64BIT)) { 4610b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 4611744692dcSJiaying Zhang goto failed_mount_wq; 4612eb40a09cSJose R. Santos } 4613eb40a09cSJose R. Santos 461425ed6e8aSDarrick J. Wong if (!set_journal_csum_feature_set(sb)) { 461525ed6e8aSDarrick J. Wong ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 461625ed6e8aSDarrick J. Wong "feature set"); 461725ed6e8aSDarrick J. Wong goto failed_mount_wq; 4618d4da6c9cSLinus Torvalds } 4619818d276cSGirish Shilamkar 4620ac27a0ecSDave Kleikamp /* We have now updated the journal if required, so we can 4621ac27a0ecSDave Kleikamp * validate the data journaling mode. */ 4622ac27a0ecSDave Kleikamp switch (test_opt(sb, DATA_FLAGS)) { 4623ac27a0ecSDave Kleikamp case 0: 4624ac27a0ecSDave Kleikamp /* No mode set, assume a default based on the journal 462563f57933SAndrew Morton * capabilities: ORDERED_DATA if the journal can 462663f57933SAndrew Morton * cope, else JOURNAL_DATA 462763f57933SAndrew Morton */ 4628dab291afSMingming Cao if (jbd2_journal_check_available_features 462927f394a7STyson Nottingham (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4630fd8c37ecSTheodore Ts'o set_opt(sb, ORDERED_DATA); 463127f394a7STyson Nottingham sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 463227f394a7STyson Nottingham } else { 4633fd8c37ecSTheodore Ts'o set_opt(sb, JOURNAL_DATA); 463427f394a7STyson Nottingham sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 463527f394a7STyson Nottingham } 4636ac27a0ecSDave Kleikamp break; 4637ac27a0ecSDave Kleikamp 4638617ba13bSMingming Cao case EXT4_MOUNT_ORDERED_DATA: 4639617ba13bSMingming Cao case EXT4_MOUNT_WRITEBACK_DATA: 4640dab291afSMingming Cao if (!jbd2_journal_check_available_features 4641dab291afSMingming Cao (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4642b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Journal does not support " 4643b31e1552SEric Sandeen "requested data journaling mode"); 4644744692dcSJiaying Zhang goto failed_mount_wq; 4645ac27a0ecSDave Kleikamp } 4646ac27a0ecSDave Kleikamp default: 4647ac27a0ecSDave Kleikamp break; 4648ac27a0ecSDave Kleikamp } 4649ab04df78SJan Kara 4650ab04df78SJan Kara if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 4651ab04df78SJan Kara test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4652ab04df78SJan Kara ext4_msg(sb, KERN_ERR, "can't mount with " 4653ab04df78SJan Kara "journal_async_commit in data=ordered mode"); 4654ab04df78SJan Kara goto failed_mount_wq; 4655ab04df78SJan Kara } 4656ab04df78SJan Kara 4657b3881f74STheodore Ts'o set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4658ac27a0ecSDave Kleikamp 465918aadd47SBobi Jam sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; 466018aadd47SBobi Jam 4661ce7e010aSTheodore Ts'o no_journal: 4662cdb7ee4cSTahsin Erdogan if (!test_opt(sb, NO_MBCACHE)) { 466347387409STahsin Erdogan sbi->s_ea_block_cache = ext4_xattr_create_cache(); 466447387409STahsin Erdogan if (!sbi->s_ea_block_cache) { 4665cdb7ee4cSTahsin Erdogan ext4_msg(sb, KERN_ERR, 4666cdb7ee4cSTahsin Erdogan "Failed to create ea_block_cache"); 46679c191f70ST Makphaibulchoke goto failed_mount_wq; 46689c191f70ST Makphaibulchoke } 46699c191f70ST Makphaibulchoke 4670dec214d0STahsin Erdogan if (ext4_has_feature_ea_inode(sb)) { 4671dec214d0STahsin Erdogan sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 4672dec214d0STahsin Erdogan if (!sbi->s_ea_inode_cache) { 4673dec214d0STahsin Erdogan ext4_msg(sb, KERN_ERR, 4674dec214d0STahsin Erdogan "Failed to create ea_inode_cache"); 4675dec214d0STahsin Erdogan goto failed_mount_wq; 4676dec214d0STahsin Erdogan } 4677dec214d0STahsin Erdogan } 4678cdb7ee4cSTahsin Erdogan } 4679dec214d0STahsin Erdogan 4680c93d8f88SEric Biggers if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) { 4681c93d8f88SEric Biggers ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity"); 4682c93d8f88SEric Biggers goto failed_mount_wq; 4683c93d8f88SEric Biggers } 4684c93d8f88SEric Biggers 4685bc98a42cSDavid Howells if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) && 4686e2b911c5SDarrick J. Wong !ext4_has_feature_encrypt(sb)) { 4687e2b911c5SDarrick J. Wong ext4_set_feature_encrypt(sb); 46886ddb2447STheodore Ts'o ext4_commit_super(sb, 1); 46896ddb2447STheodore Ts'o } 46906ddb2447STheodore Ts'o 4691fd89d5f2STejun Heo /* 4692952fc18eSTheodore Ts'o * Get the # of file system overhead blocks from the 4693952fc18eSTheodore Ts'o * superblock if present. 4694952fc18eSTheodore Ts'o */ 4695952fc18eSTheodore Ts'o if (es->s_overhead_clusters) 4696952fc18eSTheodore Ts'o sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 4697952fc18eSTheodore Ts'o else { 469807aa2ea1SLukas Czerner err = ext4_calculate_overhead(sb); 469907aa2ea1SLukas Czerner if (err) 4700952fc18eSTheodore Ts'o goto failed_mount_wq; 4701952fc18eSTheodore Ts'o } 4702952fc18eSTheodore Ts'o 4703952fc18eSTheodore Ts'o /* 4704fd89d5f2STejun Heo * The maximum number of concurrent works can be high and 4705fd89d5f2STejun Heo * concurrency isn't really necessary. Limit it to 1. 4706fd89d5f2STejun Heo */ 47072e8fa54eSJan Kara EXT4_SB(sb)->rsv_conversion_wq = 47082e8fa54eSJan Kara alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 47092e8fa54eSJan Kara if (!EXT4_SB(sb)->rsv_conversion_wq) { 47102e8fa54eSJan Kara printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 471107aa2ea1SLukas Czerner ret = -ENOMEM; 47122e8fa54eSJan Kara goto failed_mount4; 47132e8fa54eSJan Kara } 47142e8fa54eSJan Kara 4715ac27a0ecSDave Kleikamp /* 4716dab291afSMingming Cao * The jbd2_journal_load will have done any necessary log recovery, 4717ac27a0ecSDave Kleikamp * so we can safely mount the rest of the filesystem now. 4718ac27a0ecSDave Kleikamp */ 4719ac27a0ecSDave Kleikamp 47208a363970STheodore Ts'o root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 47211d1fe1eeSDavid Howells if (IS_ERR(root)) { 4722b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "get root inode failed"); 47231d1fe1eeSDavid Howells ret = PTR_ERR(root); 472432a9bb57SManish Katiyar root = NULL; 4725ac27a0ecSDave Kleikamp goto failed_mount4; 4726ac27a0ecSDave Kleikamp } 4727ac27a0ecSDave Kleikamp if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 4728b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 472994bf608aSAl Viro iput(root); 4730ac27a0ecSDave Kleikamp goto failed_mount4; 4731ac27a0ecSDave Kleikamp } 4732b886ee3eSGabriel Krisman Bertazi 4733b886ee3eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 4734b886ee3eSGabriel Krisman Bertazi if (sbi->s_encoding) 4735b886ee3eSGabriel Krisman Bertazi sb->s_d_op = &ext4_dentry_ops; 4736b886ee3eSGabriel Krisman Bertazi #endif 4737b886ee3eSGabriel Krisman Bertazi 473848fde701SAl Viro sb->s_root = d_make_root(root); 47391d1fe1eeSDavid Howells if (!sb->s_root) { 4740b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "get root dentry failed"); 47411d1fe1eeSDavid Howells ret = -ENOMEM; 47421d1fe1eeSDavid Howells goto failed_mount4; 47431d1fe1eeSDavid Howells } 4744ac27a0ecSDave Kleikamp 4745c89128a0SJaegeuk Kim ret = ext4_setup_super(sb, es, sb_rdonly(sb)); 4746c89128a0SJaegeuk Kim if (ret == -EROFS) { 47471751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 4748c89128a0SJaegeuk Kim ret = 0; 4749c89128a0SJaegeuk Kim } else if (ret) 4750c89128a0SJaegeuk Kim goto failed_mount4a; 4751ef7f3835SKalpak Shah 4752b5799018STheodore Ts'o ext4_set_resv_clusters(sb); 475327dd4385SLukas Czerner 47540f5bde1dSJan Kara if (test_opt(sb, BLOCK_VALIDITY)) { 47556fd058f7STheodore Ts'o err = ext4_setup_system_zone(sb); 47566fd058f7STheodore Ts'o if (err) { 4757b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "failed to initialize system " 4758fbe845ddSCurt Wohlgemuth "zone (%d)", err); 4759f9ae9cf5STheodore Ts'o goto failed_mount4a; 4760f9ae9cf5STheodore Ts'o } 47610f5bde1dSJan Kara } 4762f9ae9cf5STheodore Ts'o 4763f9ae9cf5STheodore Ts'o ext4_ext_init(sb); 4764f9ae9cf5STheodore Ts'o err = ext4_mb_init(sb); 4765f9ae9cf5STheodore Ts'o if (err) { 4766f9ae9cf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 4767f9ae9cf5STheodore Ts'o err); 4768dcf2d804STao Ma goto failed_mount5; 4769c2774d84SAneesh Kumar K.V } 4770c2774d84SAneesh Kumar K.V 4771d5e03cbbSTheodore Ts'o block = ext4_count_free_clusters(sb); 4772d5e03cbbSTheodore Ts'o ext4_free_blocks_count_set(sbi->s_es, 4773d5e03cbbSTheodore Ts'o EXT4_C2B(sbi, block)); 47744274f516STheodore Ts'o ext4_superblock_csum_set(sb); 4775908c7f19STejun Heo err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 4776908c7f19STejun Heo GFP_KERNEL); 4777d5e03cbbSTheodore Ts'o if (!err) { 4778d5e03cbbSTheodore Ts'o unsigned long freei = ext4_count_free_inodes(sb); 4779d5e03cbbSTheodore Ts'o sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 47804274f516STheodore Ts'o ext4_superblock_csum_set(sb); 4781908c7f19STejun Heo err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 4782908c7f19STejun Heo GFP_KERNEL); 4783d5e03cbbSTheodore Ts'o } 4784d5e03cbbSTheodore Ts'o if (!err) 4785d5e03cbbSTheodore Ts'o err = percpu_counter_init(&sbi->s_dirs_counter, 4786908c7f19STejun Heo ext4_count_dirs(sb), GFP_KERNEL); 4787d5e03cbbSTheodore Ts'o if (!err) 4788908c7f19STejun Heo err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 4789908c7f19STejun Heo GFP_KERNEL); 4790c8585c6fSDaeho Jeong if (!err) 4791bbd55937SEric Biggers err = percpu_init_rwsem(&sbi->s_writepages_rwsem); 4792c8585c6fSDaeho Jeong 4793d5e03cbbSTheodore Ts'o if (err) { 4794d5e03cbbSTheodore Ts'o ext4_msg(sb, KERN_ERR, "insufficient memory"); 4795d5e03cbbSTheodore Ts'o goto failed_mount6; 4796d5e03cbbSTheodore Ts'o } 4797d5e03cbbSTheodore Ts'o 4798e2b911c5SDarrick J. Wong if (ext4_has_feature_flex_bg(sb)) 4799d5e03cbbSTheodore Ts'o if (!ext4_fill_flex_info(sb)) { 4800d5e03cbbSTheodore Ts'o ext4_msg(sb, KERN_ERR, 4801d5e03cbbSTheodore Ts'o "unable to initialize " 4802d5e03cbbSTheodore Ts'o "flex_bg meta info!"); 4803d5e03cbbSTheodore Ts'o goto failed_mount6; 4804d5e03cbbSTheodore Ts'o } 4805d5e03cbbSTheodore Ts'o 4806bfff6873SLukas Czerner err = ext4_register_li_request(sb, first_not_zeroed); 4807bfff6873SLukas Czerner if (err) 4808dcf2d804STao Ma goto failed_mount6; 4809bfff6873SLukas Czerner 4810b5799018STheodore Ts'o err = ext4_register_sysfs(sb); 4811dcf2d804STao Ma if (err) 4812dcf2d804STao Ma goto failed_mount7; 48133197ebdbSTheodore Ts'o 48149b2ff357SJan Kara #ifdef CONFIG_QUOTA 48159b2ff357SJan Kara /* Enable quota usage during mount. */ 4816bc98a42cSDavid Howells if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 48179b2ff357SJan Kara err = ext4_enable_quotas(sb); 48189b2ff357SJan Kara if (err) 48199b2ff357SJan Kara goto failed_mount8; 48209b2ff357SJan Kara } 48219b2ff357SJan Kara #endif /* CONFIG_QUOTA */ 48229b2ff357SJan Kara 4823bc71726cSzhangyi (F) /* 4824bc71726cSzhangyi (F) * Save the original bdev mapping's wb_err value which could be 4825bc71726cSzhangyi (F) * used to detect the metadata async write error. 4826bc71726cSzhangyi (F) */ 4827bc71726cSzhangyi (F) spin_lock_init(&sbi->s_bdev_wb_lock); 4828bc71726cSzhangyi (F) if (!sb_rdonly(sb)) 4829bc71726cSzhangyi (F) errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err, 4830bc71726cSzhangyi (F) &sbi->s_bdev_wb_err); 4831bc71726cSzhangyi (F) sb->s_bdev->bd_super = sb; 4832617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 4833617ba13bSMingming Cao ext4_orphan_cleanup(sb, es); 4834617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 48350390131bSFrank Mayhar if (needs_recovery) { 4836b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "recovery complete"); 483711215630SJan Kara err = ext4_mark_recovery_complete(sb, es); 483811215630SJan Kara if (err) 483911215630SJan Kara goto failed_mount8; 48400390131bSFrank Mayhar } 48410390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal) { 48420390131bSFrank Mayhar if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 48430390131bSFrank Mayhar descr = " journalled data mode"; 48440390131bSFrank Mayhar else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 48450390131bSFrank Mayhar descr = " ordered data mode"; 48460390131bSFrank Mayhar else 48470390131bSFrank Mayhar descr = " writeback data mode"; 48480390131bSFrank Mayhar } else 48490390131bSFrank Mayhar descr = "out journal"; 48500390131bSFrank Mayhar 485179add3a3SLukas Czerner if (test_opt(sb, DISCARD)) { 485279add3a3SLukas Czerner struct request_queue *q = bdev_get_queue(sb->s_bdev); 485379add3a3SLukas Czerner if (!blk_queue_discard(q)) 485479add3a3SLukas Czerner ext4_msg(sb, KERN_WARNING, 485579add3a3SLukas Czerner "mounting with \"discard\" option, but " 485679add3a3SLukas Czerner "the device does not support discard"); 485779add3a3SLukas Czerner } 485879add3a3SLukas Czerner 4859e294a537STheodore Ts'o if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 4860d4c402d9SCurt Wohlgemuth ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 48615aee0f8aSTheodore Ts'o "Opts: %.*s%s%s", descr, 48625aee0f8aSTheodore Ts'o (int) sizeof(sbi->s_es->s_mount_opts), 48635aee0f8aSTheodore Ts'o sbi->s_es->s_mount_opts, 48648b67f04aSTheodore Ts'o *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 4865ac27a0ecSDave Kleikamp 486666e61a9eSTheodore Ts'o if (es->s_error_count) 486766e61a9eSTheodore Ts'o mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 4868ac27a0ecSDave Kleikamp 4869efbed4dcSTheodore Ts'o /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 4870efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 4871efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 4872efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 48731cf006edSDmitry Monakhov atomic_set(&sbi->s_warning_count, 0); 48741cf006edSDmitry Monakhov atomic_set(&sbi->s_msg_count, 0); 4875efbed4dcSTheodore Ts'o 4876d4c402d9SCurt Wohlgemuth kfree(orig_data); 4877ac27a0ecSDave Kleikamp return 0; 4878ac27a0ecSDave Kleikamp 4879617ba13bSMingming Cao cantfind_ext4: 4880ac27a0ecSDave Kleikamp if (!silent) 4881b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 4882ac27a0ecSDave Kleikamp goto failed_mount; 4883ac27a0ecSDave Kleikamp 488472ba7450STheodore Ts'o failed_mount8: 4885ebd173beSTheodore Ts'o ext4_unregister_sysfs(sb); 4886cb8d53d2SEric Biggers kobject_put(&sbi->s_kobj); 4887dcf2d804STao Ma failed_mount7: 4888dcf2d804STao Ma ext4_unregister_li_request(sb); 4889dcf2d804STao Ma failed_mount6: 4890f9ae9cf5STheodore Ts'o ext4_mb_release(sb); 48917c990728SSuraj Jitindar Singh rcu_read_lock(); 48927c990728SSuraj Jitindar Singh flex_groups = rcu_dereference(sbi->s_flex_groups); 48937c990728SSuraj Jitindar Singh if (flex_groups) { 48947c990728SSuraj Jitindar Singh for (i = 0; i < sbi->s_flex_groups_allocated; i++) 48957c990728SSuraj Jitindar Singh kvfree(flex_groups[i]); 48967c990728SSuraj Jitindar Singh kvfree(flex_groups); 48977c990728SSuraj Jitindar Singh } 48987c990728SSuraj Jitindar Singh rcu_read_unlock(); 4899d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_freeclusters_counter); 4900d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_freeinodes_counter); 4901d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_dirs_counter); 4902d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 4903bbd55937SEric Biggers percpu_free_rwsem(&sbi->s_writepages_rwsem); 490400764937SAzat Khuzhin failed_mount5: 4905f9ae9cf5STheodore Ts'o ext4_ext_release(sb); 4906f9ae9cf5STheodore Ts'o ext4_release_system_zone(sb); 4907f9ae9cf5STheodore Ts'o failed_mount4a: 490894bf608aSAl Viro dput(sb->s_root); 490932a9bb57SManish Katiyar sb->s_root = NULL; 491094bf608aSAl Viro failed_mount4: 4911b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "mount failed"); 49122e8fa54eSJan Kara if (EXT4_SB(sb)->rsv_conversion_wq) 49132e8fa54eSJan Kara destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 49144c0425ffSMingming Cao failed_mount_wq: 4915dec214d0STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 4916dec214d0STahsin Erdogan sbi->s_ea_inode_cache = NULL; 491750c15df6SChengguang Xu 491847387409STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 491947387409STahsin Erdogan sbi->s_ea_block_cache = NULL; 492050c15df6SChengguang Xu 49210390131bSFrank Mayhar if (sbi->s_journal) { 4922dab291afSMingming Cao jbd2_journal_destroy(sbi->s_journal); 492347b4a50bSJan Kara sbi->s_journal = NULL; 49240390131bSFrank Mayhar } 492550460fe8SDarrick J. Wong failed_mount3a: 4926d3922a77SZheng Liu ext4_es_unregister_shrinker(sbi); 4927eb68d0e2SZheng Liu failed_mount3: 49289105bb14SAl Viro del_timer_sync(&sbi->s_err_report); 4929c5e06d10SJohann Lombardi if (sbi->s_mmp_tsk) 4930c5e06d10SJohann Lombardi kthread_stop(sbi->s_mmp_tsk); 4931ac27a0ecSDave Kleikamp failed_mount2: 49321d0c3924STheodore Ts'o rcu_read_lock(); 49331d0c3924STheodore Ts'o group_desc = rcu_dereference(sbi->s_group_desc); 4934ac27a0ecSDave Kleikamp for (i = 0; i < db_count; i++) 49351d0c3924STheodore Ts'o brelse(group_desc[i]); 49361d0c3924STheodore Ts'o kvfree(group_desc); 49371d0c3924STheodore Ts'o rcu_read_unlock(); 4938ac27a0ecSDave Kleikamp failed_mount: 49390441984aSDarrick J. Wong if (sbi->s_chksum_driver) 49400441984aSDarrick J. Wong crypto_free_shash(sbi->s_chksum_driver); 4941c83ad55eSGabriel Krisman Bertazi 4942c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 4943c83ad55eSGabriel Krisman Bertazi utf8_unload(sbi->s_encoding); 4944c83ad55eSGabriel Krisman Bertazi #endif 4945c83ad55eSGabriel Krisman Bertazi 4946ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4947a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 49480ba33facSTheodore Ts'o kfree(get_qf_name(sb, sbi, i)); 4949ac27a0ecSDave Kleikamp #endif 4950ed318a6cSEric Biggers fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx); 4951617ba13bSMingming Cao ext4_blkdev_remove(sbi); 4952ac27a0ecSDave Kleikamp brelse(bh); 4953ac27a0ecSDave Kleikamp out_fail: 4954ac27a0ecSDave Kleikamp sb->s_fs_info = NULL; 4955f6830165SManish Katiyar kfree(sbi->s_blockgroup_lock); 49565aee0f8aSTheodore Ts'o out_free_base: 4957ac27a0ecSDave Kleikamp kfree(sbi); 4958d4c402d9SCurt Wohlgemuth kfree(orig_data); 49595e405595SDan Williams fs_put_dax(dax_dev); 496007aa2ea1SLukas Czerner return err ? err : ret; 4961ac27a0ecSDave Kleikamp } 4962ac27a0ecSDave Kleikamp 4963ac27a0ecSDave Kleikamp /* 4964ac27a0ecSDave Kleikamp * Setup any per-fs journal parameters now. We'll do this both on 4965ac27a0ecSDave Kleikamp * initial mount, once the journal has been initialised but before we've 4966ac27a0ecSDave Kleikamp * done any recovery; and again on any subsequent remount. 4967ac27a0ecSDave Kleikamp */ 4968617ba13bSMingming Cao static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 4969ac27a0ecSDave Kleikamp { 4970617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 4971ac27a0ecSDave Kleikamp 4972ac27a0ecSDave Kleikamp journal->j_commit_interval = sbi->s_commit_interval; 497330773840STheodore Ts'o journal->j_min_batch_time = sbi->s_min_batch_time; 497430773840STheodore Ts'o journal->j_max_batch_time = sbi->s_max_batch_time; 4975ac27a0ecSDave Kleikamp 4976a931da6aSTheodore Ts'o write_lock(&journal->j_state_lock); 4977ac27a0ecSDave Kleikamp if (test_opt(sb, BARRIER)) 4978dab291afSMingming Cao journal->j_flags |= JBD2_BARRIER; 4979ac27a0ecSDave Kleikamp else 4980dab291afSMingming Cao journal->j_flags &= ~JBD2_BARRIER; 49815bf5683aSHidehiro Kawai if (test_opt(sb, DATA_ERR_ABORT)) 49825bf5683aSHidehiro Kawai journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 49835bf5683aSHidehiro Kawai else 49845bf5683aSHidehiro Kawai journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 4985a931da6aSTheodore Ts'o write_unlock(&journal->j_state_lock); 4986ac27a0ecSDave Kleikamp } 4987ac27a0ecSDave Kleikamp 4988c6cb7e77SEric Whitney static struct inode *ext4_get_journal_inode(struct super_block *sb, 4989ac27a0ecSDave Kleikamp unsigned int journal_inum) 4990ac27a0ecSDave Kleikamp { 4991ac27a0ecSDave Kleikamp struct inode *journal_inode; 4992ac27a0ecSDave Kleikamp 4993c6cb7e77SEric Whitney /* 4994c6cb7e77SEric Whitney * Test for the existence of a valid inode on disk. Bad things 4995c6cb7e77SEric Whitney * happen if we iget() an unused inode, as the subsequent iput() 4996c6cb7e77SEric Whitney * will try to delete it. 4997c6cb7e77SEric Whitney */ 49988a363970STheodore Ts'o journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 49991d1fe1eeSDavid Howells if (IS_ERR(journal_inode)) { 5000b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "no journal found"); 5001ac27a0ecSDave Kleikamp return NULL; 5002ac27a0ecSDave Kleikamp } 5003ac27a0ecSDave Kleikamp if (!journal_inode->i_nlink) { 5004ac27a0ecSDave Kleikamp make_bad_inode(journal_inode); 5005ac27a0ecSDave Kleikamp iput(journal_inode); 5006b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 5007ac27a0ecSDave Kleikamp return NULL; 5008ac27a0ecSDave Kleikamp } 5009ac27a0ecSDave Kleikamp 5010e5f8eab8STheodore Ts'o jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 5011ac27a0ecSDave Kleikamp journal_inode, journal_inode->i_size); 50121d1fe1eeSDavid Howells if (!S_ISREG(journal_inode->i_mode)) { 5013b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "invalid journal inode"); 5014ac27a0ecSDave Kleikamp iput(journal_inode); 5015ac27a0ecSDave Kleikamp return NULL; 5016ac27a0ecSDave Kleikamp } 5017c6cb7e77SEric Whitney return journal_inode; 5018c6cb7e77SEric Whitney } 5019c6cb7e77SEric Whitney 5020c6cb7e77SEric Whitney static journal_t *ext4_get_journal(struct super_block *sb, 5021c6cb7e77SEric Whitney unsigned int journal_inum) 5022c6cb7e77SEric Whitney { 5023c6cb7e77SEric Whitney struct inode *journal_inode; 5024c6cb7e77SEric Whitney journal_t *journal; 5025c6cb7e77SEric Whitney 502611215630SJan Kara if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 502711215630SJan Kara return NULL; 5028c6cb7e77SEric Whitney 5029c6cb7e77SEric Whitney journal_inode = ext4_get_journal_inode(sb, journal_inum); 5030c6cb7e77SEric Whitney if (!journal_inode) 5031c6cb7e77SEric Whitney return NULL; 5032ac27a0ecSDave Kleikamp 5033dab291afSMingming Cao journal = jbd2_journal_init_inode(journal_inode); 5034ac27a0ecSDave Kleikamp if (!journal) { 5035b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 5036ac27a0ecSDave Kleikamp iput(journal_inode); 5037ac27a0ecSDave Kleikamp return NULL; 5038ac27a0ecSDave Kleikamp } 5039ac27a0ecSDave Kleikamp journal->j_private = sb; 5040617ba13bSMingming Cao ext4_init_journal_params(sb, journal); 5041ac27a0ecSDave Kleikamp return journal; 5042ac27a0ecSDave Kleikamp } 5043ac27a0ecSDave Kleikamp 5044617ba13bSMingming Cao static journal_t *ext4_get_dev_journal(struct super_block *sb, 5045ac27a0ecSDave Kleikamp dev_t j_dev) 5046ac27a0ecSDave Kleikamp { 5047ac27a0ecSDave Kleikamp struct buffer_head *bh; 5048ac27a0ecSDave Kleikamp journal_t *journal; 5049617ba13bSMingming Cao ext4_fsblk_t start; 5050617ba13bSMingming Cao ext4_fsblk_t len; 5051ac27a0ecSDave Kleikamp int hblock, blocksize; 5052617ba13bSMingming Cao ext4_fsblk_t sb_block; 5053ac27a0ecSDave Kleikamp unsigned long offset; 5054617ba13bSMingming Cao struct ext4_super_block *es; 5055ac27a0ecSDave Kleikamp struct block_device *bdev; 5056ac27a0ecSDave Kleikamp 505711215630SJan Kara if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 505811215630SJan Kara return NULL; 50590390131bSFrank Mayhar 5060b31e1552SEric Sandeen bdev = ext4_blkdev_get(j_dev, sb); 5061ac27a0ecSDave Kleikamp if (bdev == NULL) 5062ac27a0ecSDave Kleikamp return NULL; 5063ac27a0ecSDave Kleikamp 5064ac27a0ecSDave Kleikamp blocksize = sb->s_blocksize; 5065e1defc4fSMartin K. Petersen hblock = bdev_logical_block_size(bdev); 5066ac27a0ecSDave Kleikamp if (blocksize < hblock) { 5067b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 5068b31e1552SEric Sandeen "blocksize too small for journal device"); 5069ac27a0ecSDave Kleikamp goto out_bdev; 5070ac27a0ecSDave Kleikamp } 5071ac27a0ecSDave Kleikamp 5072617ba13bSMingming Cao sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 5073617ba13bSMingming Cao offset = EXT4_MIN_BLOCK_SIZE % blocksize; 5074ac27a0ecSDave Kleikamp set_blocksize(bdev, blocksize); 5075ac27a0ecSDave Kleikamp if (!(bh = __bread(bdev, sb_block, blocksize))) { 5076b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 5077b31e1552SEric Sandeen "external journal"); 5078ac27a0ecSDave Kleikamp goto out_bdev; 5079ac27a0ecSDave Kleikamp } 5080ac27a0ecSDave Kleikamp 50812716b802STheodore Ts'o es = (struct ext4_super_block *) (bh->b_data + offset); 5082617ba13bSMingming Cao if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 5083ac27a0ecSDave Kleikamp !(le32_to_cpu(es->s_feature_incompat) & 5084617ba13bSMingming Cao EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 5085b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "external journal has " 5086b31e1552SEric Sandeen "bad superblock"); 5087ac27a0ecSDave Kleikamp brelse(bh); 5088ac27a0ecSDave Kleikamp goto out_bdev; 5089ac27a0ecSDave Kleikamp } 5090ac27a0ecSDave Kleikamp 5091df4763beSDarrick J. Wong if ((le32_to_cpu(es->s_feature_ro_compat) & 5092df4763beSDarrick J. Wong EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 5093df4763beSDarrick J. Wong es->s_checksum != ext4_superblock_csum(sb, es)) { 5094df4763beSDarrick J. Wong ext4_msg(sb, KERN_ERR, "external journal has " 5095df4763beSDarrick J. Wong "corrupt superblock"); 5096df4763beSDarrick J. Wong brelse(bh); 5097df4763beSDarrick J. Wong goto out_bdev; 5098df4763beSDarrick J. Wong } 5099df4763beSDarrick J. Wong 5100617ba13bSMingming Cao if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 5101b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 5102ac27a0ecSDave Kleikamp brelse(bh); 5103ac27a0ecSDave Kleikamp goto out_bdev; 5104ac27a0ecSDave Kleikamp } 5105ac27a0ecSDave Kleikamp 5106bd81d8eeSLaurent Vivier len = ext4_blocks_count(es); 5107ac27a0ecSDave Kleikamp start = sb_block + 1; 5108ac27a0ecSDave Kleikamp brelse(bh); /* we're done with the superblock */ 5109ac27a0ecSDave Kleikamp 5110dab291afSMingming Cao journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 5111ac27a0ecSDave Kleikamp start, len, blocksize); 5112ac27a0ecSDave Kleikamp if (!journal) { 5113b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "failed to create device journal"); 5114ac27a0ecSDave Kleikamp goto out_bdev; 5115ac27a0ecSDave Kleikamp } 5116ac27a0ecSDave Kleikamp journal->j_private = sb; 5117dfec8a14SMike Christie ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer); 5118ac27a0ecSDave Kleikamp wait_on_buffer(journal->j_sb_buffer); 5119ac27a0ecSDave Kleikamp if (!buffer_uptodate(journal->j_sb_buffer)) { 5120b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 5121ac27a0ecSDave Kleikamp goto out_journal; 5122ac27a0ecSDave Kleikamp } 5123ac27a0ecSDave Kleikamp if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 5124b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "External journal has more than one " 5125b31e1552SEric Sandeen "user (unsupported) - %d", 5126ac27a0ecSDave Kleikamp be32_to_cpu(journal->j_superblock->s_nr_users)); 5127ac27a0ecSDave Kleikamp goto out_journal; 5128ac27a0ecSDave Kleikamp } 5129ee7ed3aaSChunguang Xu EXT4_SB(sb)->s_journal_bdev = bdev; 5130617ba13bSMingming Cao ext4_init_journal_params(sb, journal); 5131ac27a0ecSDave Kleikamp return journal; 51320b8e58a1SAndreas Dilger 5133ac27a0ecSDave Kleikamp out_journal: 5134dab291afSMingming Cao jbd2_journal_destroy(journal); 5135ac27a0ecSDave Kleikamp out_bdev: 5136617ba13bSMingming Cao ext4_blkdev_put(bdev); 5137ac27a0ecSDave Kleikamp return NULL; 5138ac27a0ecSDave Kleikamp } 5139ac27a0ecSDave Kleikamp 5140617ba13bSMingming Cao static int ext4_load_journal(struct super_block *sb, 5141617ba13bSMingming Cao struct ext4_super_block *es, 5142ac27a0ecSDave Kleikamp unsigned long journal_devnum) 5143ac27a0ecSDave Kleikamp { 5144ac27a0ecSDave Kleikamp journal_t *journal; 5145ac27a0ecSDave Kleikamp unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 5146ac27a0ecSDave Kleikamp dev_t journal_dev; 5147ac27a0ecSDave Kleikamp int err = 0; 5148ac27a0ecSDave Kleikamp int really_read_only; 5149273108faSLukas Czerner int journal_dev_ro; 5150ac27a0ecSDave Kleikamp 515111215630SJan Kara if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 515211215630SJan Kara return -EFSCORRUPTED; 51530390131bSFrank Mayhar 5154ac27a0ecSDave Kleikamp if (journal_devnum && 5155ac27a0ecSDave Kleikamp journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5156b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "external journal device major/minor " 5157b31e1552SEric Sandeen "numbers have changed"); 5158ac27a0ecSDave Kleikamp journal_dev = new_decode_dev(journal_devnum); 5159ac27a0ecSDave Kleikamp } else 5160ac27a0ecSDave Kleikamp journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 5161ac27a0ecSDave Kleikamp 5162273108faSLukas Czerner if (journal_inum && journal_dev) { 5163273108faSLukas Czerner ext4_msg(sb, KERN_ERR, 5164273108faSLukas Czerner "filesystem has both journal inode and journal device!"); 5165273108faSLukas Czerner return -EINVAL; 5166273108faSLukas Czerner } 5167273108faSLukas Czerner 5168273108faSLukas Czerner if (journal_inum) { 5169273108faSLukas Czerner journal = ext4_get_journal(sb, journal_inum); 5170273108faSLukas Czerner if (!journal) 5171273108faSLukas Czerner return -EINVAL; 5172273108faSLukas Czerner } else { 5173273108faSLukas Czerner journal = ext4_get_dev_journal(sb, journal_dev); 5174273108faSLukas Czerner if (!journal) 5175273108faSLukas Czerner return -EINVAL; 5176273108faSLukas Czerner } 5177273108faSLukas Czerner 5178273108faSLukas Czerner journal_dev_ro = bdev_read_only(journal->j_dev); 5179273108faSLukas Czerner really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro; 5180273108faSLukas Czerner 5181273108faSLukas Czerner if (journal_dev_ro && !sb_rdonly(sb)) { 5182273108faSLukas Czerner ext4_msg(sb, KERN_ERR, 5183273108faSLukas Czerner "journal device read-only, try mounting with '-o ro'"); 5184273108faSLukas Czerner err = -EROFS; 5185273108faSLukas Czerner goto err_out; 5186273108faSLukas Czerner } 5187ac27a0ecSDave Kleikamp 5188ac27a0ecSDave Kleikamp /* 5189ac27a0ecSDave Kleikamp * Are we loading a blank journal or performing recovery after a 5190ac27a0ecSDave Kleikamp * crash? For recovery, we need to check in advance whether we 5191ac27a0ecSDave Kleikamp * can get read-write access to the device. 5192ac27a0ecSDave Kleikamp */ 5193e2b911c5SDarrick J. Wong if (ext4_has_feature_journal_needs_recovery(sb)) { 5194bc98a42cSDavid Howells if (sb_rdonly(sb)) { 5195b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "INFO: recovery " 5196b31e1552SEric Sandeen "required on readonly filesystem"); 5197ac27a0ecSDave Kleikamp if (really_read_only) { 5198b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "write access " 5199d98bf8cdSSimon Ruderich "unavailable, cannot proceed " 5200d98bf8cdSSimon Ruderich "(try mounting with noload)"); 5201273108faSLukas Czerner err = -EROFS; 5202273108faSLukas Czerner goto err_out; 5203ac27a0ecSDave Kleikamp } 5204b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "write access will " 5205b31e1552SEric Sandeen "be enabled during recovery"); 5206ac27a0ecSDave Kleikamp } 5207ac27a0ecSDave Kleikamp } 5208ac27a0ecSDave Kleikamp 520990576c0bSTheodore Ts'o if (!(journal->j_flags & JBD2_BARRIER)) 5210b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "barriers disabled"); 52114776004fSTheodore Ts'o 5212e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal_needs_recovery(sb)) 5213dab291afSMingming Cao err = jbd2_journal_wipe(journal, !really_read_only); 52141c13d5c0STheodore Ts'o if (!err) { 52151c13d5c0STheodore Ts'o char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 52161c13d5c0STheodore Ts'o if (save) 52171c13d5c0STheodore Ts'o memcpy(save, ((char *) es) + 52181c13d5c0STheodore Ts'o EXT4_S_ERR_START, EXT4_S_ERR_LEN); 5219dab291afSMingming Cao err = jbd2_journal_load(journal); 52201c13d5c0STheodore Ts'o if (save) 52211c13d5c0STheodore Ts'o memcpy(((char *) es) + EXT4_S_ERR_START, 52221c13d5c0STheodore Ts'o save, EXT4_S_ERR_LEN); 52231c13d5c0STheodore Ts'o kfree(save); 52241c13d5c0STheodore Ts'o } 5225ac27a0ecSDave Kleikamp 5226ac27a0ecSDave Kleikamp if (err) { 5227b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "error loading journal"); 5228273108faSLukas Czerner goto err_out; 5229ac27a0ecSDave Kleikamp } 5230ac27a0ecSDave Kleikamp 5231617ba13bSMingming Cao EXT4_SB(sb)->s_journal = journal; 523211215630SJan Kara err = ext4_clear_journal_err(sb, es); 523311215630SJan Kara if (err) { 523411215630SJan Kara EXT4_SB(sb)->s_journal = NULL; 523511215630SJan Kara jbd2_journal_destroy(journal); 523611215630SJan Kara return err; 523711215630SJan Kara } 5238ac27a0ecSDave Kleikamp 5239c41303ceSMaciej Żenczykowski if (!really_read_only && journal_devnum && 5240ac27a0ecSDave Kleikamp journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5241ac27a0ecSDave Kleikamp es->s_journal_dev = cpu_to_le32(journal_devnum); 5242ac27a0ecSDave Kleikamp 5243ac27a0ecSDave Kleikamp /* Make sure we flush the recovery flag to disk. */ 5244e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 5245ac27a0ecSDave Kleikamp } 5246ac27a0ecSDave Kleikamp 5247ac27a0ecSDave Kleikamp return 0; 5248273108faSLukas Czerner 5249273108faSLukas Czerner err_out: 5250273108faSLukas Czerner jbd2_journal_destroy(journal); 5251273108faSLukas Czerner return err; 5252ac27a0ecSDave Kleikamp } 5253ac27a0ecSDave Kleikamp 5254e2d67052STheodore Ts'o static int ext4_commit_super(struct super_block *sb, int sync) 5255ac27a0ecSDave Kleikamp { 5256e2d67052STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 5257617ba13bSMingming Cao struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 5258c4be0c1dSTakashi Sato int error = 0; 5259ac27a0ecSDave Kleikamp 5260bdfe0cbdSTheodore Ts'o if (!sbh || block_device_ejected(sb)) 5261c4be0c1dSTakashi Sato return error; 5262a17712c8SJon Derrick 5263a17712c8SJon Derrick /* 526471290b36STheodore Ts'o * If the file system is mounted read-only, don't update the 526571290b36STheodore Ts'o * superblock write time. This avoids updating the superblock 526671290b36STheodore Ts'o * write time when we are mounting the root file system 526771290b36STheodore Ts'o * read/only but we need to replay the journal; at that point, 526871290b36STheodore Ts'o * for people who are east of GMT and who make their clock 526971290b36STheodore Ts'o * tick in localtime for Windows bug-for-bug compatibility, 527071290b36STheodore Ts'o * the clock is set in the future, and this will cause e2fsck 527171290b36STheodore Ts'o * to complain and force a full file system check. 527271290b36STheodore Ts'o */ 52731751e8a6SLinus Torvalds if (!(sb->s_flags & SB_RDONLY)) 52746a0678a7SArnd Bergmann ext4_update_tstamp(es, s_wtime); 5275f613dfcbSTheodore Ts'o if (sb->s_bdev->bd_part) 5276afc32f7eSTheodore Ts'o es->s_kbytes_written = 5277afc32f7eSTheodore Ts'o cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 5278dbae2c55SMichael Callahan ((part_stat_read(sb->s_bdev->bd_part, 5279dbae2c55SMichael Callahan sectors[STAT_WRITE]) - 5280afc32f7eSTheodore Ts'o EXT4_SB(sb)->s_sectors_written_start) >> 1)); 5281f613dfcbSTheodore Ts'o else 5282f613dfcbSTheodore Ts'o es->s_kbytes_written = 5283f613dfcbSTheodore Ts'o cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 5284d5e03cbbSTheodore Ts'o if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter)) 528557042651STheodore Ts'o ext4_free_blocks_count_set(es, 528657042651STheodore Ts'o EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 528757042651STheodore Ts'o &EXT4_SB(sb)->s_freeclusters_counter))); 5288d5e03cbbSTheodore Ts'o if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter)) 52897f93cff9STheodore Ts'o es->s_free_inodes_count = 52907f93cff9STheodore Ts'o cpu_to_le32(percpu_counter_sum_positive( 52915d1b1b3fSAneesh Kumar K.V &EXT4_SB(sb)->s_freeinodes_counter)); 5292ac27a0ecSDave Kleikamp BUFFER_TRACE(sbh, "marking dirty"); 529306db49e6STheodore Ts'o ext4_superblock_csum_set(sb); 52941566a48aSTheodore Ts'o if (sync) 52954743f839SPranay Kr. Srivastava lock_buffer(sbh); 5296e8680786STheodore Ts'o if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 52974743f839SPranay Kr. Srivastava /* 52984743f839SPranay Kr. Srivastava * Oh, dear. A previous attempt to write the 52994743f839SPranay Kr. Srivastava * superblock failed. This could happen because the 53004743f839SPranay Kr. Srivastava * USB device was yanked out. Or it could happen to 53014743f839SPranay Kr. Srivastava * be a transient write error and maybe the block will 53024743f839SPranay Kr. Srivastava * be remapped. Nothing we can do but to retry the 53034743f839SPranay Kr. Srivastava * write and hope for the best. 53044743f839SPranay Kr. Srivastava */ 53054743f839SPranay Kr. Srivastava ext4_msg(sb, KERN_ERR, "previous I/O error to " 53064743f839SPranay Kr. Srivastava "superblock detected"); 53074743f839SPranay Kr. Srivastava clear_buffer_write_io_error(sbh); 53084743f839SPranay Kr. Srivastava set_buffer_uptodate(sbh); 53094743f839SPranay Kr. Srivastava } 5310ac27a0ecSDave Kleikamp mark_buffer_dirty(sbh); 5311914258bfSTheodore Ts'o if (sync) { 53121566a48aSTheodore Ts'o unlock_buffer(sbh); 5313564bc402SDaeho Jeong error = __sync_dirty_buffer(sbh, 531400473374SJan Kara REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0)); 5315c89128a0SJaegeuk Kim if (buffer_write_io_error(sbh)) { 5316b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "I/O error while writing " 5317b31e1552SEric Sandeen "superblock"); 5318914258bfSTheodore Ts'o clear_buffer_write_io_error(sbh); 5319914258bfSTheodore Ts'o set_buffer_uptodate(sbh); 5320914258bfSTheodore Ts'o } 5321914258bfSTheodore Ts'o } 5322c4be0c1dSTakashi Sato return error; 5323ac27a0ecSDave Kleikamp } 5324ac27a0ecSDave Kleikamp 5325ac27a0ecSDave Kleikamp /* 5326ac27a0ecSDave Kleikamp * Have we just finished recovery? If so, and if we are mounting (or 5327ac27a0ecSDave Kleikamp * remounting) the filesystem readonly, then we will end up with a 5328ac27a0ecSDave Kleikamp * consistent fs on disk. Record that fact. 5329ac27a0ecSDave Kleikamp */ 533011215630SJan Kara static int ext4_mark_recovery_complete(struct super_block *sb, 5331617ba13bSMingming Cao struct ext4_super_block *es) 5332ac27a0ecSDave Kleikamp { 533311215630SJan Kara int err; 5334617ba13bSMingming Cao journal_t *journal = EXT4_SB(sb)->s_journal; 5335ac27a0ecSDave Kleikamp 5336e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal(sb)) { 533711215630SJan Kara if (journal != NULL) { 533811215630SJan Kara ext4_error(sb, "Journal got removed while the fs was " 533911215630SJan Kara "mounted!"); 534011215630SJan Kara return -EFSCORRUPTED; 534111215630SJan Kara } 534211215630SJan Kara return 0; 53430390131bSFrank Mayhar } 5344dab291afSMingming Cao jbd2_journal_lock_updates(journal); 534511215630SJan Kara err = jbd2_journal_flush(journal); 534611215630SJan Kara if (err < 0) 53477ffe1ea8SHidehiro Kawai goto out; 53487ffe1ea8SHidehiro Kawai 5349bc98a42cSDavid Howells if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) { 5350e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 5351e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 5352ac27a0ecSDave Kleikamp } 53537ffe1ea8SHidehiro Kawai out: 5354dab291afSMingming Cao jbd2_journal_unlock_updates(journal); 535511215630SJan Kara return err; 5356ac27a0ecSDave Kleikamp } 5357ac27a0ecSDave Kleikamp 5358ac27a0ecSDave Kleikamp /* 5359ac27a0ecSDave Kleikamp * If we are mounting (or read-write remounting) a filesystem whose journal 5360ac27a0ecSDave Kleikamp * has recorded an error from a previous lifetime, move that error to the 5361ac27a0ecSDave Kleikamp * main filesystem now. 5362ac27a0ecSDave Kleikamp */ 536311215630SJan Kara static int ext4_clear_journal_err(struct super_block *sb, 5364617ba13bSMingming Cao struct ext4_super_block *es) 5365ac27a0ecSDave Kleikamp { 5366ac27a0ecSDave Kleikamp journal_t *journal; 5367ac27a0ecSDave Kleikamp int j_errno; 5368ac27a0ecSDave Kleikamp const char *errstr; 5369ac27a0ecSDave Kleikamp 537011215630SJan Kara if (!ext4_has_feature_journal(sb)) { 537111215630SJan Kara ext4_error(sb, "Journal got removed while the fs was mounted!"); 537211215630SJan Kara return -EFSCORRUPTED; 537311215630SJan Kara } 53740390131bSFrank Mayhar 5375617ba13bSMingming Cao journal = EXT4_SB(sb)->s_journal; 5376ac27a0ecSDave Kleikamp 5377ac27a0ecSDave Kleikamp /* 5378ac27a0ecSDave Kleikamp * Now check for any error status which may have been recorded in the 5379617ba13bSMingming Cao * journal by a prior ext4_error() or ext4_abort() 5380ac27a0ecSDave Kleikamp */ 5381ac27a0ecSDave Kleikamp 5382dab291afSMingming Cao j_errno = jbd2_journal_errno(journal); 5383ac27a0ecSDave Kleikamp if (j_errno) { 5384ac27a0ecSDave Kleikamp char nbuf[16]; 5385ac27a0ecSDave Kleikamp 5386617ba13bSMingming Cao errstr = ext4_decode_error(sb, j_errno, nbuf); 538712062dddSEric Sandeen ext4_warning(sb, "Filesystem error recorded " 5388ac27a0ecSDave Kleikamp "from previous mount: %s", errstr); 538912062dddSEric Sandeen ext4_warning(sb, "Marking fs in need of filesystem check."); 5390ac27a0ecSDave Kleikamp 5391617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 5392617ba13bSMingming Cao es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 5393e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 5394ac27a0ecSDave Kleikamp 5395dab291afSMingming Cao jbd2_journal_clear_err(journal); 5396d796c52eSTheodore Ts'o jbd2_journal_update_sb_errno(journal); 5397ac27a0ecSDave Kleikamp } 539811215630SJan Kara return 0; 5399ac27a0ecSDave Kleikamp } 5400ac27a0ecSDave Kleikamp 5401ac27a0ecSDave Kleikamp /* 5402ac27a0ecSDave Kleikamp * Force the running and committing transactions to commit, 5403ac27a0ecSDave Kleikamp * and wait on the commit. 5404ac27a0ecSDave Kleikamp */ 5405617ba13bSMingming Cao int ext4_force_commit(struct super_block *sb) 5406ac27a0ecSDave Kleikamp { 5407ac27a0ecSDave Kleikamp journal_t *journal; 5408ac27a0ecSDave Kleikamp 5409bc98a42cSDavid Howells if (sb_rdonly(sb)) 5410ac27a0ecSDave Kleikamp return 0; 5411ac27a0ecSDave Kleikamp 5412617ba13bSMingming Cao journal = EXT4_SB(sb)->s_journal; 5413b1deefc9SGuo Chao return ext4_journal_force_commit(journal); 5414ac27a0ecSDave Kleikamp } 5415ac27a0ecSDave Kleikamp 5416617ba13bSMingming Cao static int ext4_sync_fs(struct super_block *sb, int wait) 5417ac27a0ecSDave Kleikamp { 541814ce0cb4STheodore Ts'o int ret = 0; 54199eddacf9SJan Kara tid_t target; 542006a407f1SDmitry Monakhov bool needs_barrier = false; 54218d5d02e6SMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 5422ac27a0ecSDave Kleikamp 542349598e04SJun Piao if (unlikely(ext4_forced_shutdown(sbi))) 54240db1ff22STheodore Ts'o return 0; 54250db1ff22STheodore Ts'o 54269bffad1eSTheodore Ts'o trace_ext4_sync_fs(sb, wait); 54272e8fa54eSJan Kara flush_workqueue(sbi->rsv_conversion_wq); 5428a1177825SJan Kara /* 5429a1177825SJan Kara * Writeback quota in non-journalled quota case - journalled quota has 5430a1177825SJan Kara * no dirty dquots 5431a1177825SJan Kara */ 5432a1177825SJan Kara dquot_writeback_dquots(sb, -1); 543306a407f1SDmitry Monakhov /* 543406a407f1SDmitry Monakhov * Data writeback is possible w/o journal transaction, so barrier must 543506a407f1SDmitry Monakhov * being sent at the end of the function. But we can skip it if 543606a407f1SDmitry Monakhov * transaction_commit will do it for us. 543706a407f1SDmitry Monakhov */ 5438bda32530STheodore Ts'o if (sbi->s_journal) { 543906a407f1SDmitry Monakhov target = jbd2_get_latest_transaction(sbi->s_journal); 544006a407f1SDmitry Monakhov if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 544106a407f1SDmitry Monakhov !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 544206a407f1SDmitry Monakhov needs_barrier = true; 544306a407f1SDmitry Monakhov 54448d5d02e6SMingming Cao if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 5445ac27a0ecSDave Kleikamp if (wait) 5446bda32530STheodore Ts'o ret = jbd2_log_wait_commit(sbi->s_journal, 5447bda32530STheodore Ts'o target); 54480390131bSFrank Mayhar } 5449bda32530STheodore Ts'o } else if (wait && test_opt(sb, BARRIER)) 5450bda32530STheodore Ts'o needs_barrier = true; 545106a407f1SDmitry Monakhov if (needs_barrier) { 545206a407f1SDmitry Monakhov int err; 54539398554fSChristoph Hellwig err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL); 545406a407f1SDmitry Monakhov if (!ret) 545506a407f1SDmitry Monakhov ret = err; 545606a407f1SDmitry Monakhov } 545706a407f1SDmitry Monakhov 545806a407f1SDmitry Monakhov return ret; 545906a407f1SDmitry Monakhov } 546006a407f1SDmitry Monakhov 5461ac27a0ecSDave Kleikamp /* 5462ac27a0ecSDave Kleikamp * LVM calls this function before a (read-only) snapshot is created. This 5463ac27a0ecSDave Kleikamp * gives us a chance to flush the journal completely and mark the fs clean. 5464be4f27d3SYongqiang Yang * 5465be4f27d3SYongqiang Yang * Note that only this function cannot bring a filesystem to be in a clean 54668e8ad8a5SJan Kara * state independently. It relies on upper layer to stop all data & metadata 54678e8ad8a5SJan Kara * modifications. 5468ac27a0ecSDave Kleikamp */ 5469c4be0c1dSTakashi Sato static int ext4_freeze(struct super_block *sb) 5470ac27a0ecSDave Kleikamp { 5471c4be0c1dSTakashi Sato int error = 0; 5472c4be0c1dSTakashi Sato journal_t *journal; 5473ac27a0ecSDave Kleikamp 5474bc98a42cSDavid Howells if (sb_rdonly(sb)) 54759ca92389STheodore Ts'o return 0; 54769ca92389STheodore Ts'o 5477c4be0c1dSTakashi Sato journal = EXT4_SB(sb)->s_journal; 5478ac27a0ecSDave Kleikamp 5479bb044576STheodore Ts'o if (journal) { 5480ac27a0ecSDave Kleikamp /* Now we set up the journal barrier. */ 5481dab291afSMingming Cao jbd2_journal_lock_updates(journal); 54827ffe1ea8SHidehiro Kawai 54837ffe1ea8SHidehiro Kawai /* 5484bb044576STheodore Ts'o * Don't clear the needs_recovery flag if we failed to 5485bb044576STheodore Ts'o * flush the journal. 54867ffe1ea8SHidehiro Kawai */ 5487c4be0c1dSTakashi Sato error = jbd2_journal_flush(journal); 54886b0310fbSEric Sandeen if (error < 0) 54896b0310fbSEric Sandeen goto out; 5490ac27a0ecSDave Kleikamp 5491ac27a0ecSDave Kleikamp /* Journal blocked and flushed, clear needs_recovery flag. */ 5492e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 5493c642dc9eSEric Sandeen } 5494c642dc9eSEric Sandeen 5495e2d67052STheodore Ts'o error = ext4_commit_super(sb, 1); 54966b0310fbSEric Sandeen out: 5497bb044576STheodore Ts'o if (journal) 54988e8ad8a5SJan Kara /* we rely on upper layer to stop further updates */ 5499bb044576STheodore Ts'o jbd2_journal_unlock_updates(journal); 55006b0310fbSEric Sandeen return error; 5501ac27a0ecSDave Kleikamp } 5502ac27a0ecSDave Kleikamp 5503ac27a0ecSDave Kleikamp /* 5504ac27a0ecSDave Kleikamp * Called by LVM after the snapshot is done. We need to reset the RECOVER 5505ac27a0ecSDave Kleikamp * flag here, even though the filesystem is not technically dirty yet. 5506ac27a0ecSDave Kleikamp */ 5507c4be0c1dSTakashi Sato static int ext4_unfreeze(struct super_block *sb) 5508ac27a0ecSDave Kleikamp { 5509bc98a42cSDavid Howells if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) 55109ca92389STheodore Ts'o return 0; 55119ca92389STheodore Ts'o 5512c642dc9eSEric Sandeen if (EXT4_SB(sb)->s_journal) { 55139ca92389STheodore Ts'o /* Reset the needs_recovery flag before the fs is unlocked. */ 5514e2b911c5SDarrick J. Wong ext4_set_feature_journal_needs_recovery(sb); 5515c642dc9eSEric Sandeen } 5516c642dc9eSEric Sandeen 5517e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 5518c4be0c1dSTakashi Sato return 0; 5519ac27a0ecSDave Kleikamp } 5520ac27a0ecSDave Kleikamp 5521673c6100STheodore Ts'o /* 5522673c6100STheodore Ts'o * Structure to save mount options for ext4_remount's benefit 5523673c6100STheodore Ts'o */ 5524673c6100STheodore Ts'o struct ext4_mount_options { 5525673c6100STheodore Ts'o unsigned long s_mount_opt; 5526a2595b8aSTheodore Ts'o unsigned long s_mount_opt2; 552708cefc7aSEric W. Biederman kuid_t s_resuid; 552808cefc7aSEric W. Biederman kgid_t s_resgid; 5529673c6100STheodore Ts'o unsigned long s_commit_interval; 5530673c6100STheodore Ts'o u32 s_min_batch_time, s_max_batch_time; 5531673c6100STheodore Ts'o #ifdef CONFIG_QUOTA 5532673c6100STheodore Ts'o int s_jquota_fmt; 5533a2d4a646SJan Kara char *s_qf_names[EXT4_MAXQUOTAS]; 5534673c6100STheodore Ts'o #endif 5535673c6100STheodore Ts'o }; 5536673c6100STheodore Ts'o 5537617ba13bSMingming Cao static int ext4_remount(struct super_block *sb, int *flags, char *data) 5538ac27a0ecSDave Kleikamp { 5539617ba13bSMingming Cao struct ext4_super_block *es; 5540617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 5541f25391ebSLukas Czerner unsigned long old_sb_flags, vfs_flags; 5542617ba13bSMingming Cao struct ext4_mount_options old_opts; 5543c79d967dSChristoph Hellwig int enable_quota = 0; 55448a266467STheodore Ts'o ext4_group_t g; 5545b3881f74STheodore Ts'o unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 5546c5e06d10SJohann Lombardi int err = 0; 5547ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 554803dafb5fSChen Gang int i, j; 554933458eabSTheodore Ts'o char *to_free[EXT4_MAXQUOTAS]; 5550ac27a0ecSDave Kleikamp #endif 5551d4c402d9SCurt Wohlgemuth char *orig_data = kstrdup(data, GFP_KERNEL); 5552ac27a0ecSDave Kleikamp 555321ac738eSChengguang Xu if (data && !orig_data) 555421ac738eSChengguang Xu return -ENOMEM; 555521ac738eSChengguang Xu 5556ac27a0ecSDave Kleikamp /* Store the original options */ 5557ac27a0ecSDave Kleikamp old_sb_flags = sb->s_flags; 5558ac27a0ecSDave Kleikamp old_opts.s_mount_opt = sbi->s_mount_opt; 5559a2595b8aSTheodore Ts'o old_opts.s_mount_opt2 = sbi->s_mount_opt2; 5560ac27a0ecSDave Kleikamp old_opts.s_resuid = sbi->s_resuid; 5561ac27a0ecSDave Kleikamp old_opts.s_resgid = sbi->s_resgid; 5562ac27a0ecSDave Kleikamp old_opts.s_commit_interval = sbi->s_commit_interval; 556330773840STheodore Ts'o old_opts.s_min_batch_time = sbi->s_min_batch_time; 556430773840STheodore Ts'o old_opts.s_max_batch_time = sbi->s_max_batch_time; 5565ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 5566ac27a0ecSDave Kleikamp old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 5567a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 556803dafb5fSChen Gang if (sbi->s_qf_names[i]) { 556933458eabSTheodore Ts'o char *qf_name = get_qf_name(sb, sbi, i); 557033458eabSTheodore Ts'o 557133458eabSTheodore Ts'o old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 557203dafb5fSChen Gang if (!old_opts.s_qf_names[i]) { 557303dafb5fSChen Gang for (j = 0; j < i; j++) 557403dafb5fSChen Gang kfree(old_opts.s_qf_names[j]); 55753e36a163SWei Yongjun kfree(orig_data); 557603dafb5fSChen Gang return -ENOMEM; 557703dafb5fSChen Gang } 557803dafb5fSChen Gang } else 557903dafb5fSChen Gang old_opts.s_qf_names[i] = NULL; 5580ac27a0ecSDave Kleikamp #endif 5581b3881f74STheodore Ts'o if (sbi->s_journal && sbi->s_journal->j_task->io_context) 5582b3881f74STheodore Ts'o journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 5583ac27a0ecSDave Kleikamp 5584f25391ebSLukas Czerner /* 5585f25391ebSLukas Czerner * Some options can be enabled by ext4 and/or by VFS mount flag 5586f25391ebSLukas Czerner * either way we need to make sure it matches in both *flags and 5587f25391ebSLukas Czerner * s_flags. Copy those selected flags from *flags to s_flags 5588f25391ebSLukas Czerner */ 5589f25391ebSLukas Czerner vfs_flags = SB_LAZYTIME | SB_I_VERSION; 5590f25391ebSLukas Czerner sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags); 5591f25391ebSLukas Czerner 5592661aa520SEric Sandeen if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 5593ac27a0ecSDave Kleikamp err = -EINVAL; 5594ac27a0ecSDave Kleikamp goto restore_opts; 5595ac27a0ecSDave Kleikamp } 5596ac27a0ecSDave Kleikamp 55976b992ff2SDarrick J. Wong if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 55986b992ff2SDarrick J. Wong test_opt(sb, JOURNAL_CHECKSUM)) { 55996b992ff2SDarrick J. Wong ext4_msg(sb, KERN_ERR, "changing journal_checksum " 56002d5b86e0SEric Sandeen "during remount not supported; ignoring"); 56012d5b86e0SEric Sandeen sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 5602c6d3d56dSDarrick J. Wong } 5603c6d3d56dSDarrick J. Wong 56046ae6514bSPiotr Sarna if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 56056ae6514bSPiotr Sarna if (test_opt2(sb, EXPLICIT_DELALLOC)) { 56066ae6514bSPiotr Sarna ext4_msg(sb, KERN_ERR, "can't mount with " 56076ae6514bSPiotr Sarna "both data=journal and delalloc"); 56086ae6514bSPiotr Sarna err = -EINVAL; 56096ae6514bSPiotr Sarna goto restore_opts; 56106ae6514bSPiotr Sarna } 56116ae6514bSPiotr Sarna if (test_opt(sb, DIOREAD_NOLOCK)) { 56126ae6514bSPiotr Sarna ext4_msg(sb, KERN_ERR, "can't mount with " 56136ae6514bSPiotr Sarna "both data=journal and dioread_nolock"); 56146ae6514bSPiotr Sarna err = -EINVAL; 56156ae6514bSPiotr Sarna goto restore_opts; 56166ae6514bSPiotr Sarna } 5617ab04df78SJan Kara } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 5618ab04df78SJan Kara if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5619ab04df78SJan Kara ext4_msg(sb, KERN_ERR, "can't mount with " 5620ab04df78SJan Kara "journal_async_commit in data=ordered mode"); 5621ab04df78SJan Kara err = -EINVAL; 5622ab04df78SJan Kara goto restore_opts; 5623ab04df78SJan Kara } 5624923ae0ffSRoss Zwisler } 5625923ae0ffSRoss Zwisler 5626cdb7ee4cSTahsin Erdogan if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 5627cdb7ee4cSTahsin Erdogan ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 5628cdb7ee4cSTahsin Erdogan err = -EINVAL; 5629cdb7ee4cSTahsin Erdogan goto restore_opts; 5630cdb7ee4cSTahsin Erdogan } 5631cdb7ee4cSTahsin Erdogan 56324ab2f15bSTheodore Ts'o if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 563354d3adbcSTheodore Ts'o ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user"); 5634ac27a0ecSDave Kleikamp 56351751e8a6SLinus Torvalds sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 56361751e8a6SLinus Torvalds (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 5637ac27a0ecSDave Kleikamp 5638ac27a0ecSDave Kleikamp es = sbi->s_es; 5639ac27a0ecSDave Kleikamp 5640b3881f74STheodore Ts'o if (sbi->s_journal) { 5641617ba13bSMingming Cao ext4_init_journal_params(sb, sbi->s_journal); 5642b3881f74STheodore Ts'o set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 5643b3881f74STheodore Ts'o } 5644ac27a0ecSDave Kleikamp 56451751e8a6SLinus Torvalds if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) { 56464ab2f15bSTheodore Ts'o if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 5647ac27a0ecSDave Kleikamp err = -EROFS; 5648ac27a0ecSDave Kleikamp goto restore_opts; 5649ac27a0ecSDave Kleikamp } 5650ac27a0ecSDave Kleikamp 56511751e8a6SLinus Torvalds if (*flags & SB_RDONLY) { 565238c03b34STheodore Ts'o err = sync_filesystem(sb); 565338c03b34STheodore Ts'o if (err < 0) 565438c03b34STheodore Ts'o goto restore_opts; 56550f0dd62fSChristoph Hellwig err = dquot_suspend(sb, -1); 56560f0dd62fSChristoph Hellwig if (err < 0) 5657c79d967dSChristoph Hellwig goto restore_opts; 5658c79d967dSChristoph Hellwig 5659ac27a0ecSDave Kleikamp /* 5660ac27a0ecSDave Kleikamp * First of all, the unconditional stuff we have to do 5661ac27a0ecSDave Kleikamp * to disable replay of the journal when we next remount 5662ac27a0ecSDave Kleikamp */ 56631751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 5664ac27a0ecSDave Kleikamp 5665ac27a0ecSDave Kleikamp /* 5666ac27a0ecSDave Kleikamp * OK, test if we are remounting a valid rw partition 5667ac27a0ecSDave Kleikamp * readonly, and if so set the rdonly flag and then 5668ac27a0ecSDave Kleikamp * mark the partition as valid again. 5669ac27a0ecSDave Kleikamp */ 5670617ba13bSMingming Cao if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 5671617ba13bSMingming Cao (sbi->s_mount_state & EXT4_VALID_FS)) 5672ac27a0ecSDave Kleikamp es->s_state = cpu_to_le16(sbi->s_mount_state); 5673ac27a0ecSDave Kleikamp 567411215630SJan Kara if (sbi->s_journal) { 567511215630SJan Kara /* 567611215630SJan Kara * We let remount-ro finish even if marking fs 567711215630SJan Kara * as clean failed... 567811215630SJan Kara */ 5679617ba13bSMingming Cao ext4_mark_recovery_complete(sb, es); 568011215630SJan Kara } 56812dca60d9STheodore Ts'o if (sbi->s_mmp_tsk) 56822dca60d9STheodore Ts'o kthread_stop(sbi->s_mmp_tsk); 5683ac27a0ecSDave Kleikamp } else { 5684a13fb1a4SEric Sandeen /* Make sure we can mount this feature set readwrite */ 5685e2b911c5SDarrick J. Wong if (ext4_has_feature_readonly(sb) || 56862cb5cc8bSDarrick J. Wong !ext4_feature_set_ok(sb, 0)) { 5687ac27a0ecSDave Kleikamp err = -EROFS; 5688ac27a0ecSDave Kleikamp goto restore_opts; 5689ac27a0ecSDave Kleikamp } 5690ead6596bSEric Sandeen /* 56918a266467STheodore Ts'o * Make sure the group descriptor checksums 56920b8e58a1SAndreas Dilger * are sane. If they aren't, refuse to remount r/w. 56938a266467STheodore Ts'o */ 56948a266467STheodore Ts'o for (g = 0; g < sbi->s_groups_count; g++) { 56958a266467STheodore Ts'o struct ext4_group_desc *gdp = 56968a266467STheodore Ts'o ext4_get_group_desc(sb, g, NULL); 56978a266467STheodore Ts'o 5698feb0ab32SDarrick J. Wong if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 5699b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 5700b31e1552SEric Sandeen "ext4_remount: Checksum for group %u failed (%u!=%u)", 5701e2b911c5SDarrick J. Wong g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 57028a266467STheodore Ts'o le16_to_cpu(gdp->bg_checksum)); 57036a797d27SDarrick J. Wong err = -EFSBADCRC; 57048a266467STheodore Ts'o goto restore_opts; 57058a266467STheodore Ts'o } 57068a266467STheodore Ts'o } 57078a266467STheodore Ts'o 57088a266467STheodore Ts'o /* 5709ead6596bSEric Sandeen * If we have an unprocessed orphan list hanging 5710ead6596bSEric Sandeen * around from a previously readonly bdev mount, 5711ead6596bSEric Sandeen * require a full umount/remount for now. 5712ead6596bSEric Sandeen */ 5713ead6596bSEric Sandeen if (es->s_last_orphan) { 5714b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "Couldn't " 5715ead6596bSEric Sandeen "remount RDWR because of unprocessed " 5716ead6596bSEric Sandeen "orphan inode list. Please " 5717b31e1552SEric Sandeen "umount/remount instead"); 5718ead6596bSEric Sandeen err = -EINVAL; 5719ead6596bSEric Sandeen goto restore_opts; 5720ead6596bSEric Sandeen } 5721ead6596bSEric Sandeen 5722ac27a0ecSDave Kleikamp /* 5723bc71726cSzhangyi (F) * Update the original bdev mapping's wb_err value 5724bc71726cSzhangyi (F) * which could be used to detect the metadata async 5725bc71726cSzhangyi (F) * write error. 5726bc71726cSzhangyi (F) */ 5727bc71726cSzhangyi (F) errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err, 5728bc71726cSzhangyi (F) &sbi->s_bdev_wb_err); 5729bc71726cSzhangyi (F) 5730bc71726cSzhangyi (F) /* 5731ac27a0ecSDave Kleikamp * Mounting a RDONLY partition read-write, so reread 5732ac27a0ecSDave Kleikamp * and store the current valid flag. (It may have 5733ac27a0ecSDave Kleikamp * been changed by e2fsck since we originally mounted 5734ac27a0ecSDave Kleikamp * the partition.) 5735ac27a0ecSDave Kleikamp */ 573611215630SJan Kara if (sbi->s_journal) { 573711215630SJan Kara err = ext4_clear_journal_err(sb, es); 573811215630SJan Kara if (err) 573911215630SJan Kara goto restore_opts; 574011215630SJan Kara } 5741ac27a0ecSDave Kleikamp sbi->s_mount_state = le16_to_cpu(es->s_state); 5742c89128a0SJaegeuk Kim 5743c89128a0SJaegeuk Kim err = ext4_setup_super(sb, es, 0); 5744c89128a0SJaegeuk Kim if (err) 5745c89128a0SJaegeuk Kim goto restore_opts; 5746c89128a0SJaegeuk Kim 57471751e8a6SLinus Torvalds sb->s_flags &= ~SB_RDONLY; 5748e2b911c5SDarrick J. Wong if (ext4_has_feature_mmp(sb)) 5749c5e06d10SJohann Lombardi if (ext4_multi_mount_protect(sb, 5750c5e06d10SJohann Lombardi le64_to_cpu(es->s_mmp_block))) { 5751c5e06d10SJohann Lombardi err = -EROFS; 5752c5e06d10SJohann Lombardi goto restore_opts; 5753c5e06d10SJohann Lombardi } 5754c79d967dSChristoph Hellwig enable_quota = 1; 5755ac27a0ecSDave Kleikamp } 5756ac27a0ecSDave Kleikamp } 5757bfff6873SLukas Czerner 5758bfff6873SLukas Czerner /* 5759bfff6873SLukas Czerner * Reinitialize lazy itable initialization thread based on 5760bfff6873SLukas Czerner * current settings 5761bfff6873SLukas Czerner */ 5762bc98a42cSDavid Howells if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 5763bfff6873SLukas Czerner ext4_unregister_li_request(sb); 5764bfff6873SLukas Czerner else { 5765bfff6873SLukas Czerner ext4_group_t first_not_zeroed; 5766bfff6873SLukas Czerner first_not_zeroed = ext4_has_uninit_itable(sb); 5767bfff6873SLukas Czerner ext4_register_li_request(sb, first_not_zeroed); 5768bfff6873SLukas Czerner } 5769bfff6873SLukas Czerner 57700f5bde1dSJan Kara /* 57710f5bde1dSJan Kara * Handle creation of system zone data early because it can fail. 57720f5bde1dSJan Kara * Releasing of existing data is done when we are sure remount will 57730f5bde1dSJan Kara * succeed. 57740f5bde1dSJan Kara */ 5775*dd0db94fSChunguang Xu if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) { 5776d176b1f6SJan Kara err = ext4_setup_system_zone(sb); 5777d176b1f6SJan Kara if (err) 5778d176b1f6SJan Kara goto restore_opts; 57790f5bde1dSJan Kara } 5780d176b1f6SJan Kara 5781c89128a0SJaegeuk Kim if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 5782c89128a0SJaegeuk Kim err = ext4_commit_super(sb, 1); 5783c89128a0SJaegeuk Kim if (err) 5784c89128a0SJaegeuk Kim goto restore_opts; 5785c89128a0SJaegeuk Kim } 57860390131bSFrank Mayhar 5787ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 5788ac27a0ecSDave Kleikamp /* Release old quota file names */ 5789a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 5790ac27a0ecSDave Kleikamp kfree(old_opts.s_qf_names[i]); 57917c319d32SAditya Kali if (enable_quota) { 57927c319d32SAditya Kali if (sb_any_quota_suspended(sb)) 57930f0dd62fSChristoph Hellwig dquot_resume(sb, -1); 5794e2b911c5SDarrick J. Wong else if (ext4_has_feature_quota(sb)) { 57957c319d32SAditya Kali err = ext4_enable_quotas(sb); 579607724f98STheodore Ts'o if (err) 57977c319d32SAditya Kali goto restore_opts; 57987c319d32SAditya Kali } 57997c319d32SAditya Kali } 58007c319d32SAditya Kali #endif 5801*dd0db94fSChunguang Xu if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 58020f5bde1dSJan Kara ext4_release_system_zone(sb); 5803d4c402d9SCurt Wohlgemuth 5804f25391ebSLukas Czerner /* 5805f25391ebSLukas Czerner * Some options can be enabled by ext4 and/or by VFS mount flag 5806f25391ebSLukas Czerner * either way we need to make sure it matches in both *flags and 5807f25391ebSLukas Czerner * s_flags. Copy those selected flags from s_flags to *flags 5808f25391ebSLukas Czerner */ 5809f25391ebSLukas Czerner *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags); 5810f25391ebSLukas Czerner 5811d4c402d9SCurt Wohlgemuth ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 5812d4c402d9SCurt Wohlgemuth kfree(orig_data); 5813ac27a0ecSDave Kleikamp return 0; 58140b8e58a1SAndreas Dilger 5815ac27a0ecSDave Kleikamp restore_opts: 5816ac27a0ecSDave Kleikamp sb->s_flags = old_sb_flags; 5817ac27a0ecSDave Kleikamp sbi->s_mount_opt = old_opts.s_mount_opt; 5818a2595b8aSTheodore Ts'o sbi->s_mount_opt2 = old_opts.s_mount_opt2; 5819ac27a0ecSDave Kleikamp sbi->s_resuid = old_opts.s_resuid; 5820ac27a0ecSDave Kleikamp sbi->s_resgid = old_opts.s_resgid; 5821ac27a0ecSDave Kleikamp sbi->s_commit_interval = old_opts.s_commit_interval; 582230773840STheodore Ts'o sbi->s_min_batch_time = old_opts.s_min_batch_time; 582330773840STheodore Ts'o sbi->s_max_batch_time = old_opts.s_max_batch_time; 5824*dd0db94fSChunguang Xu if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 58250f5bde1dSJan Kara ext4_release_system_zone(sb); 5826ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 5827ac27a0ecSDave Kleikamp sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 5828a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 582933458eabSTheodore Ts'o to_free[i] = get_qf_name(sb, sbi, i); 583033458eabSTheodore Ts'o rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 5831ac27a0ecSDave Kleikamp } 583233458eabSTheodore Ts'o synchronize_rcu(); 583333458eabSTheodore Ts'o for (i = 0; i < EXT4_MAXQUOTAS; i++) 583433458eabSTheodore Ts'o kfree(to_free[i]); 5835ac27a0ecSDave Kleikamp #endif 5836d4c402d9SCurt Wohlgemuth kfree(orig_data); 5837ac27a0ecSDave Kleikamp return err; 5838ac27a0ecSDave Kleikamp } 5839ac27a0ecSDave Kleikamp 5840689c958cSLi Xi #ifdef CONFIG_QUOTA 5841689c958cSLi Xi static int ext4_statfs_project(struct super_block *sb, 5842689c958cSLi Xi kprojid_t projid, struct kstatfs *buf) 5843689c958cSLi Xi { 5844689c958cSLi Xi struct kqid qid; 5845689c958cSLi Xi struct dquot *dquot; 5846689c958cSLi Xi u64 limit; 5847689c958cSLi Xi u64 curblock; 5848689c958cSLi Xi 5849689c958cSLi Xi qid = make_kqid_projid(projid); 5850689c958cSLi Xi dquot = dqget(sb, qid); 5851689c958cSLi Xi if (IS_ERR(dquot)) 5852689c958cSLi Xi return PTR_ERR(dquot); 58537b9ca4c6SJan Kara spin_lock(&dquot->dq_dqb_lock); 5854689c958cSLi Xi 5855a08fe66eSChengguang Xu limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 5856a08fe66eSChengguang Xu dquot->dq_dqb.dqb_bhardlimit); 585757c32ea4SChengguang Xu limit >>= sb->s_blocksize_bits; 585857c32ea4SChengguang Xu 5859689c958cSLi Xi if (limit && buf->f_blocks > limit) { 5860f06925c7SKonstantin Khlebnikov curblock = (dquot->dq_dqb.dqb_curspace + 5861f06925c7SKonstantin Khlebnikov dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 5862689c958cSLi Xi buf->f_blocks = limit; 5863689c958cSLi Xi buf->f_bfree = buf->f_bavail = 5864689c958cSLi Xi (buf->f_blocks > curblock) ? 5865689c958cSLi Xi (buf->f_blocks - curblock) : 0; 5866689c958cSLi Xi } 5867689c958cSLi Xi 5868a08fe66eSChengguang Xu limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 5869a08fe66eSChengguang Xu dquot->dq_dqb.dqb_ihardlimit); 5870689c958cSLi Xi if (limit && buf->f_files > limit) { 5871689c958cSLi Xi buf->f_files = limit; 5872689c958cSLi Xi buf->f_ffree = 5873689c958cSLi Xi (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 5874689c958cSLi Xi (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 5875689c958cSLi Xi } 5876689c958cSLi Xi 58777b9ca4c6SJan Kara spin_unlock(&dquot->dq_dqb_lock); 5878689c958cSLi Xi dqput(dquot); 5879689c958cSLi Xi return 0; 5880689c958cSLi Xi } 5881689c958cSLi Xi #endif 5882689c958cSLi Xi 5883617ba13bSMingming Cao static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 5884ac27a0ecSDave Kleikamp { 5885ac27a0ecSDave Kleikamp struct super_block *sb = dentry->d_sb; 5886617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 5887617ba13bSMingming Cao struct ext4_super_block *es = sbi->s_es; 588827dd4385SLukas Czerner ext4_fsblk_t overhead = 0, resv_blocks; 5889960cc398SPekka Enberg u64 fsid; 5890d02a9391SKazuya Mio s64 bfree; 589127dd4385SLukas Czerner resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 5892ac27a0ecSDave Kleikamp 5893952fc18eSTheodore Ts'o if (!test_opt(sb, MINIX_DF)) 5894952fc18eSTheodore Ts'o overhead = sbi->s_overhead; 5895ac27a0ecSDave Kleikamp 5896617ba13bSMingming Cao buf->f_type = EXT4_SUPER_MAGIC; 5897ac27a0ecSDave Kleikamp buf->f_bsize = sb->s_blocksize; 5898b72f78cbSEric Sandeen buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 589957042651STheodore Ts'o bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 590057042651STheodore Ts'o percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 5901d02a9391SKazuya Mio /* prevent underflow in case that few free space is available */ 590257042651STheodore Ts'o buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 590327dd4385SLukas Czerner buf->f_bavail = buf->f_bfree - 590427dd4385SLukas Czerner (ext4_r_blocks_count(es) + resv_blocks); 590527dd4385SLukas Czerner if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 5906ac27a0ecSDave Kleikamp buf->f_bavail = 0; 5907ac27a0ecSDave Kleikamp buf->f_files = le32_to_cpu(es->s_inodes_count); 590852d9f3b4SPeter Zijlstra buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 5909617ba13bSMingming Cao buf->f_namelen = EXT4_NAME_LEN; 5910960cc398SPekka Enberg fsid = le64_to_cpup((void *)es->s_uuid) ^ 5911960cc398SPekka Enberg le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 5912960cc398SPekka Enberg buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 5913960cc398SPekka Enberg buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 59140b8e58a1SAndreas Dilger 5915689c958cSLi Xi #ifdef CONFIG_QUOTA 5916689c958cSLi Xi if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 5917689c958cSLi Xi sb_has_quota_limits_enabled(sb, PRJQUOTA)) 5918689c958cSLi Xi ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 5919689c958cSLi Xi #endif 5920ac27a0ecSDave Kleikamp return 0; 5921ac27a0ecSDave Kleikamp } 5922ac27a0ecSDave Kleikamp 5923ac27a0ecSDave Kleikamp 5924ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 5925ac27a0ecSDave Kleikamp 5926bc8230eeSJan Kara /* 5927bc8230eeSJan Kara * Helper functions so that transaction is started before we acquire dqio_sem 5928bc8230eeSJan Kara * to keep correct lock ordering of transaction > dqio_sem 5929bc8230eeSJan Kara */ 5930ac27a0ecSDave Kleikamp static inline struct inode *dquot_to_inode(struct dquot *dquot) 5931ac27a0ecSDave Kleikamp { 59324c376dcaSEric W. Biederman return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 5933ac27a0ecSDave Kleikamp } 5934ac27a0ecSDave Kleikamp 5935617ba13bSMingming Cao static int ext4_write_dquot(struct dquot *dquot) 5936ac27a0ecSDave Kleikamp { 5937ac27a0ecSDave Kleikamp int ret, err; 5938ac27a0ecSDave Kleikamp handle_t *handle; 5939ac27a0ecSDave Kleikamp struct inode *inode; 5940ac27a0ecSDave Kleikamp 5941ac27a0ecSDave Kleikamp inode = dquot_to_inode(dquot); 59429924a92aSTheodore Ts'o handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 5943617ba13bSMingming Cao EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 5944ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 5945ac27a0ecSDave Kleikamp return PTR_ERR(handle); 5946ac27a0ecSDave Kleikamp ret = dquot_commit(dquot); 5947617ba13bSMingming Cao err = ext4_journal_stop(handle); 5948ac27a0ecSDave Kleikamp if (!ret) 5949ac27a0ecSDave Kleikamp ret = err; 5950ac27a0ecSDave Kleikamp return ret; 5951ac27a0ecSDave Kleikamp } 5952ac27a0ecSDave Kleikamp 5953617ba13bSMingming Cao static int ext4_acquire_dquot(struct dquot *dquot) 5954ac27a0ecSDave Kleikamp { 5955ac27a0ecSDave Kleikamp int ret, err; 5956ac27a0ecSDave Kleikamp handle_t *handle; 5957ac27a0ecSDave Kleikamp 59589924a92aSTheodore Ts'o handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5959617ba13bSMingming Cao EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 5960ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 5961ac27a0ecSDave Kleikamp return PTR_ERR(handle); 5962ac27a0ecSDave Kleikamp ret = dquot_acquire(dquot); 5963617ba13bSMingming Cao err = ext4_journal_stop(handle); 5964ac27a0ecSDave Kleikamp if (!ret) 5965ac27a0ecSDave Kleikamp ret = err; 5966ac27a0ecSDave Kleikamp return ret; 5967ac27a0ecSDave Kleikamp } 5968ac27a0ecSDave Kleikamp 5969617ba13bSMingming Cao static int ext4_release_dquot(struct dquot *dquot) 5970ac27a0ecSDave Kleikamp { 5971ac27a0ecSDave Kleikamp int ret, err; 5972ac27a0ecSDave Kleikamp handle_t *handle; 5973ac27a0ecSDave Kleikamp 59749924a92aSTheodore Ts'o handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 5975617ba13bSMingming Cao EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 59769c3013e9SJan Kara if (IS_ERR(handle)) { 59779c3013e9SJan Kara /* Release dquot anyway to avoid endless cycle in dqput() */ 59789c3013e9SJan Kara dquot_release(dquot); 5979ac27a0ecSDave Kleikamp return PTR_ERR(handle); 59809c3013e9SJan Kara } 5981ac27a0ecSDave Kleikamp ret = dquot_release(dquot); 5982617ba13bSMingming Cao err = ext4_journal_stop(handle); 5983ac27a0ecSDave Kleikamp if (!ret) 5984ac27a0ecSDave Kleikamp ret = err; 5985ac27a0ecSDave Kleikamp return ret; 5986ac27a0ecSDave Kleikamp } 5987ac27a0ecSDave Kleikamp 5988617ba13bSMingming Cao static int ext4_mark_dquot_dirty(struct dquot *dquot) 5989ac27a0ecSDave Kleikamp { 5990262b4662SJan Kara struct super_block *sb = dquot->dq_sb; 5991262b4662SJan Kara struct ext4_sb_info *sbi = EXT4_SB(sb); 5992262b4662SJan Kara 59932c8be6b2SJan Kara /* Are we journaling quotas? */ 5994e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb) || 5995262b4662SJan Kara sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 5996ac27a0ecSDave Kleikamp dquot_mark_dquot_dirty(dquot); 5997617ba13bSMingming Cao return ext4_write_dquot(dquot); 5998ac27a0ecSDave Kleikamp } else { 5999ac27a0ecSDave Kleikamp return dquot_mark_dquot_dirty(dquot); 6000ac27a0ecSDave Kleikamp } 6001ac27a0ecSDave Kleikamp } 6002ac27a0ecSDave Kleikamp 6003617ba13bSMingming Cao static int ext4_write_info(struct super_block *sb, int type) 6004ac27a0ecSDave Kleikamp { 6005ac27a0ecSDave Kleikamp int ret, err; 6006ac27a0ecSDave Kleikamp handle_t *handle; 6007ac27a0ecSDave Kleikamp 6008ac27a0ecSDave Kleikamp /* Data block + inode block */ 60092b0143b5SDavid Howells handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); 6010ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 6011ac27a0ecSDave Kleikamp return PTR_ERR(handle); 6012ac27a0ecSDave Kleikamp ret = dquot_commit_info(sb, type); 6013617ba13bSMingming Cao err = ext4_journal_stop(handle); 6014ac27a0ecSDave Kleikamp if (!ret) 6015ac27a0ecSDave Kleikamp ret = err; 6016ac27a0ecSDave Kleikamp return ret; 6017ac27a0ecSDave Kleikamp } 6018ac27a0ecSDave Kleikamp 6019ac27a0ecSDave Kleikamp /* 6020ac27a0ecSDave Kleikamp * Turn on quotas during mount time - we need to find 6021ac27a0ecSDave Kleikamp * the quota file and such... 6022ac27a0ecSDave Kleikamp */ 6023617ba13bSMingming Cao static int ext4_quota_on_mount(struct super_block *sb, int type) 6024ac27a0ecSDave Kleikamp { 602533458eabSTheodore Ts'o return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type), 6026617ba13bSMingming Cao EXT4_SB(sb)->s_jquota_fmt, type); 6027ac27a0ecSDave Kleikamp } 6028ac27a0ecSDave Kleikamp 6029daf647d2STheodore Ts'o static void lockdep_set_quota_inode(struct inode *inode, int subclass) 6030daf647d2STheodore Ts'o { 6031daf647d2STheodore Ts'o struct ext4_inode_info *ei = EXT4_I(inode); 6032daf647d2STheodore Ts'o 6033daf647d2STheodore Ts'o /* The first argument of lockdep_set_subclass has to be 6034daf647d2STheodore Ts'o * *exactly* the same as the argument to init_rwsem() --- in 6035daf647d2STheodore Ts'o * this case, in init_once() --- or lockdep gets unhappy 6036daf647d2STheodore Ts'o * because the name of the lock is set using the 6037daf647d2STheodore Ts'o * stringification of the argument to init_rwsem(). 6038daf647d2STheodore Ts'o */ 6039daf647d2STheodore Ts'o (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 6040daf647d2STheodore Ts'o lockdep_set_subclass(&ei->i_data_sem, subclass); 6041daf647d2STheodore Ts'o } 6042daf647d2STheodore Ts'o 6043ac27a0ecSDave Kleikamp /* 6044ac27a0ecSDave Kleikamp * Standard function to be called on quota_on 6045ac27a0ecSDave Kleikamp */ 6046617ba13bSMingming Cao static int ext4_quota_on(struct super_block *sb, int type, int format_id, 60478c54ca9cSAl Viro const struct path *path) 6048ac27a0ecSDave Kleikamp { 6049ac27a0ecSDave Kleikamp int err; 6050ac27a0ecSDave Kleikamp 6051ac27a0ecSDave Kleikamp if (!test_opt(sb, QUOTA)) 6052ac27a0ecSDave Kleikamp return -EINVAL; 60530623543bSJan Kara 6054ac27a0ecSDave Kleikamp /* Quotafile not on the same filesystem? */ 6055d8c9584eSAl Viro if (path->dentry->d_sb != sb) 6056ac27a0ecSDave Kleikamp return -EXDEV; 60570623543bSJan Kara /* Journaling quota? */ 60580623543bSJan Kara if (EXT4_SB(sb)->s_qf_names[type]) { 60592b2d6d01STheodore Ts'o /* Quotafile not in fs root? */ 6060f00c9e44SJan Kara if (path->dentry->d_parent != sb->s_root) 6061b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 6062b31e1552SEric Sandeen "Quota file not on filesystem root. " 6063b31e1552SEric Sandeen "Journaled quota will not work"); 606491389240SJan Kara sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 606591389240SJan Kara } else { 606691389240SJan Kara /* 606791389240SJan Kara * Clear the flag just in case mount options changed since 606891389240SJan Kara * last time. 606991389240SJan Kara */ 607091389240SJan Kara sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 60710623543bSJan Kara } 60720623543bSJan Kara 60730623543bSJan Kara /* 60740623543bSJan Kara * When we journal data on quota file, we have to flush journal to see 60750623543bSJan Kara * all updates to the file when we bypass pagecache... 60760623543bSJan Kara */ 60770390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal && 60782b0143b5SDavid Howells ext4_should_journal_data(d_inode(path->dentry))) { 60790623543bSJan Kara /* 60800623543bSJan Kara * We don't need to lock updates but journal_flush() could 60810623543bSJan Kara * otherwise be livelocked... 60820623543bSJan Kara */ 60830623543bSJan Kara jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 60847ffe1ea8SHidehiro Kawai err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 60850623543bSJan Kara jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 6086f00c9e44SJan Kara if (err) 60877ffe1ea8SHidehiro Kawai return err; 60887ffe1ea8SHidehiro Kawai } 6089957153fcSJan Kara 6090daf647d2STheodore Ts'o lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 6091daf647d2STheodore Ts'o err = dquot_quota_on(sb, type, format_id, path); 6092957153fcSJan Kara if (err) { 6093daf647d2STheodore Ts'o lockdep_set_quota_inode(path->dentry->d_inode, 6094daf647d2STheodore Ts'o I_DATA_SEM_NORMAL); 6095957153fcSJan Kara } else { 6096957153fcSJan Kara struct inode *inode = d_inode(path->dentry); 6097957153fcSJan Kara handle_t *handle; 6098957153fcSJan Kara 609961a92987SJan Kara /* 610061a92987SJan Kara * Set inode flags to prevent userspace from messing with quota 610161a92987SJan Kara * files. If this fails, we return success anyway since quotas 610261a92987SJan Kara * are already enabled and this is not a hard failure. 610361a92987SJan Kara */ 6104957153fcSJan Kara inode_lock(inode); 6105957153fcSJan Kara handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 6106957153fcSJan Kara if (IS_ERR(handle)) 6107957153fcSJan Kara goto unlock_inode; 6108957153fcSJan Kara EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 6109957153fcSJan Kara inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 6110957153fcSJan Kara S_NOATIME | S_IMMUTABLE); 61114209ae12SHarshad Shirwadkar err = ext4_mark_inode_dirty(handle, inode); 6112957153fcSJan Kara ext4_journal_stop(handle); 6113957153fcSJan Kara unlock_inode: 6114957153fcSJan Kara inode_unlock(inode); 6115957153fcSJan Kara } 6116daf647d2STheodore Ts'o return err; 6117ac27a0ecSDave Kleikamp } 6118ac27a0ecSDave Kleikamp 61197c319d32SAditya Kali static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 61207c319d32SAditya Kali unsigned int flags) 61217c319d32SAditya Kali { 61227c319d32SAditya Kali int err; 61237c319d32SAditya Kali struct inode *qf_inode; 6124a2d4a646SJan Kara unsigned long qf_inums[EXT4_MAXQUOTAS] = { 61257c319d32SAditya Kali le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 6126689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 6127689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 61287c319d32SAditya Kali }; 61297c319d32SAditya Kali 6130e2b911c5SDarrick J. Wong BUG_ON(!ext4_has_feature_quota(sb)); 61317c319d32SAditya Kali 61327c319d32SAditya Kali if (!qf_inums[type]) 61337c319d32SAditya Kali return -EPERM; 61347c319d32SAditya Kali 61358a363970STheodore Ts'o qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 61367c319d32SAditya Kali if (IS_ERR(qf_inode)) { 61377c319d32SAditya Kali ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 61387c319d32SAditya Kali return PTR_ERR(qf_inode); 61397c319d32SAditya Kali } 61407c319d32SAditya Kali 6141bcb13850SJan Kara /* Don't account quota for quota files to avoid recursion */ 6142bcb13850SJan Kara qf_inode->i_flags |= S_NOQUOTA; 6143daf647d2STheodore Ts'o lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 61447212b95eSJan Kara err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 6145daf647d2STheodore Ts'o if (err) 6146daf647d2STheodore Ts'o lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 614761157b24SPan Bian iput(qf_inode); 61487c319d32SAditya Kali 61497c319d32SAditya Kali return err; 61507c319d32SAditya Kali } 61517c319d32SAditya Kali 61527c319d32SAditya Kali /* Enable usage tracking for all quota types. */ 61537c319d32SAditya Kali static int ext4_enable_quotas(struct super_block *sb) 61547c319d32SAditya Kali { 61557c319d32SAditya Kali int type, err = 0; 6156a2d4a646SJan Kara unsigned long qf_inums[EXT4_MAXQUOTAS] = { 61577c319d32SAditya Kali le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 6158689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 6159689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 61607c319d32SAditya Kali }; 616149da9392SJan Kara bool quota_mopt[EXT4_MAXQUOTAS] = { 616249da9392SJan Kara test_opt(sb, USRQUOTA), 616349da9392SJan Kara test_opt(sb, GRPQUOTA), 616449da9392SJan Kara test_opt(sb, PRJQUOTA), 616549da9392SJan Kara }; 61667c319d32SAditya Kali 616791389240SJan Kara sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 6168a2d4a646SJan Kara for (type = 0; type < EXT4_MAXQUOTAS; type++) { 61697c319d32SAditya Kali if (qf_inums[type]) { 61707c319d32SAditya Kali err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 617149da9392SJan Kara DQUOT_USAGE_ENABLED | 617249da9392SJan Kara (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 61737c319d32SAditya Kali if (err) { 61747c319d32SAditya Kali ext4_warning(sb, 617572ba7450STheodore Ts'o "Failed to enable quota tracking " 617672ba7450STheodore Ts'o "(type=%d, err=%d). Please run " 617772ba7450STheodore Ts'o "e2fsck to fix.", type, err); 61787f144fd0SJunichi Uekawa for (type--; type >= 0; type--) 61797f144fd0SJunichi Uekawa dquot_quota_off(sb, type); 61807f144fd0SJunichi Uekawa 61817c319d32SAditya Kali return err; 61827c319d32SAditya Kali } 61837c319d32SAditya Kali } 61847c319d32SAditya Kali } 61857c319d32SAditya Kali return 0; 61867c319d32SAditya Kali } 61877c319d32SAditya Kali 6188ca0e05e4SDmitry Monakhov static int ext4_quota_off(struct super_block *sb, int type) 6189ca0e05e4SDmitry Monakhov { 619021f97697SJan Kara struct inode *inode = sb_dqopt(sb)->files[type]; 619121f97697SJan Kara handle_t *handle; 6192957153fcSJan Kara int err; 619321f97697SJan Kara 619487009d86SDmitry Monakhov /* Force all delayed allocation blocks to be allocated. 619587009d86SDmitry Monakhov * Caller already holds s_umount sem */ 619687009d86SDmitry Monakhov if (test_opt(sb, DELALLOC)) 6197ca0e05e4SDmitry Monakhov sync_filesystem(sb); 6198ca0e05e4SDmitry Monakhov 6199957153fcSJan Kara if (!inode || !igrab(inode)) 62000b268590SAmir Goldstein goto out; 62010b268590SAmir Goldstein 6202957153fcSJan Kara err = dquot_quota_off(sb, type); 6203964edf66SJan Kara if (err || ext4_has_feature_quota(sb)) 6204957153fcSJan Kara goto out_put; 6205957153fcSJan Kara 6206957153fcSJan Kara inode_lock(inode); 620761a92987SJan Kara /* 620861a92987SJan Kara * Update modification times of quota files when userspace can 620961a92987SJan Kara * start looking at them. If we fail, we return success anyway since 621061a92987SJan Kara * this is not a hard failure and quotas are already disabled. 621161a92987SJan Kara */ 62129924a92aSTheodore Ts'o handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 62134209ae12SHarshad Shirwadkar if (IS_ERR(handle)) { 62144209ae12SHarshad Shirwadkar err = PTR_ERR(handle); 6215957153fcSJan Kara goto out_unlock; 62164209ae12SHarshad Shirwadkar } 6217957153fcSJan Kara EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 6218957153fcSJan Kara inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 6219eeca7ea1SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 62204209ae12SHarshad Shirwadkar err = ext4_mark_inode_dirty(handle, inode); 622121f97697SJan Kara ext4_journal_stop(handle); 6222957153fcSJan Kara out_unlock: 6223957153fcSJan Kara inode_unlock(inode); 6224957153fcSJan Kara out_put: 6225964edf66SJan Kara lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 6226957153fcSJan Kara iput(inode); 6227957153fcSJan Kara return err; 622821f97697SJan Kara out: 6229ca0e05e4SDmitry Monakhov return dquot_quota_off(sb, type); 6230ca0e05e4SDmitry Monakhov } 6231ca0e05e4SDmitry Monakhov 6232ac27a0ecSDave Kleikamp /* Read data from quotafile - avoid pagecache and such because we cannot afford 6233ac27a0ecSDave Kleikamp * acquiring the locks... As quota files are never truncated and quota code 6234ac27a0ecSDave Kleikamp * itself serializes the operations (and no one else should touch the files) 6235ac27a0ecSDave Kleikamp * we don't have to be afraid of races */ 6236617ba13bSMingming Cao static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 6237ac27a0ecSDave Kleikamp size_t len, loff_t off) 6238ac27a0ecSDave Kleikamp { 6239ac27a0ecSDave Kleikamp struct inode *inode = sb_dqopt(sb)->files[type]; 6240725d26d3SAneesh Kumar K.V ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 6241ac27a0ecSDave Kleikamp int offset = off & (sb->s_blocksize - 1); 6242ac27a0ecSDave Kleikamp int tocopy; 6243ac27a0ecSDave Kleikamp size_t toread; 6244ac27a0ecSDave Kleikamp struct buffer_head *bh; 6245ac27a0ecSDave Kleikamp loff_t i_size = i_size_read(inode); 6246ac27a0ecSDave Kleikamp 6247ac27a0ecSDave Kleikamp if (off > i_size) 6248ac27a0ecSDave Kleikamp return 0; 6249ac27a0ecSDave Kleikamp if (off+len > i_size) 6250ac27a0ecSDave Kleikamp len = i_size-off; 6251ac27a0ecSDave Kleikamp toread = len; 6252ac27a0ecSDave Kleikamp while (toread > 0) { 6253ac27a0ecSDave Kleikamp tocopy = sb->s_blocksize - offset < toread ? 6254ac27a0ecSDave Kleikamp sb->s_blocksize - offset : toread; 62551c215028STheodore Ts'o bh = ext4_bread(NULL, inode, blk, 0); 62561c215028STheodore Ts'o if (IS_ERR(bh)) 62571c215028STheodore Ts'o return PTR_ERR(bh); 6258ac27a0ecSDave Kleikamp if (!bh) /* A hole? */ 6259ac27a0ecSDave Kleikamp memset(data, 0, tocopy); 6260ac27a0ecSDave Kleikamp else 6261ac27a0ecSDave Kleikamp memcpy(data, bh->b_data+offset, tocopy); 6262ac27a0ecSDave Kleikamp brelse(bh); 6263ac27a0ecSDave Kleikamp offset = 0; 6264ac27a0ecSDave Kleikamp toread -= tocopy; 6265ac27a0ecSDave Kleikamp data += tocopy; 6266ac27a0ecSDave Kleikamp blk++; 6267ac27a0ecSDave Kleikamp } 6268ac27a0ecSDave Kleikamp return len; 6269ac27a0ecSDave Kleikamp } 6270ac27a0ecSDave Kleikamp 6271ac27a0ecSDave Kleikamp /* Write to quotafile (we know the transaction is already started and has 6272ac27a0ecSDave Kleikamp * enough credits) */ 6273617ba13bSMingming Cao static ssize_t ext4_quota_write(struct super_block *sb, int type, 6274ac27a0ecSDave Kleikamp const char *data, size_t len, loff_t off) 6275ac27a0ecSDave Kleikamp { 6276ac27a0ecSDave Kleikamp struct inode *inode = sb_dqopt(sb)->files[type]; 6277725d26d3SAneesh Kumar K.V ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 62784209ae12SHarshad Shirwadkar int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1); 6279c5e298aeSTheodore Ts'o int retries = 0; 6280ac27a0ecSDave Kleikamp struct buffer_head *bh; 6281ac27a0ecSDave Kleikamp handle_t *handle = journal_current_handle(); 6282ac27a0ecSDave Kleikamp 62830390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal && !handle) { 6284b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 6285b31e1552SEric Sandeen " cancelled because transaction is not started", 62869c3013e9SJan Kara (unsigned long long)off, (unsigned long long)len); 62879c3013e9SJan Kara return -EIO; 62889c3013e9SJan Kara } 628967eeb568SDmitry Monakhov /* 629067eeb568SDmitry Monakhov * Since we account only one data block in transaction credits, 629167eeb568SDmitry Monakhov * then it is impossible to cross a block boundary. 629267eeb568SDmitry Monakhov */ 629367eeb568SDmitry Monakhov if (sb->s_blocksize - offset < len) { 629467eeb568SDmitry Monakhov ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 629567eeb568SDmitry Monakhov " cancelled because not block aligned", 629667eeb568SDmitry Monakhov (unsigned long long)off, (unsigned long long)len); 629767eeb568SDmitry Monakhov return -EIO; 629867eeb568SDmitry Monakhov } 629967eeb568SDmitry Monakhov 6300c5e298aeSTheodore Ts'o do { 6301c5e298aeSTheodore Ts'o bh = ext4_bread(handle, inode, blk, 6302c5e298aeSTheodore Ts'o EXT4_GET_BLOCKS_CREATE | 6303c5e298aeSTheodore Ts'o EXT4_GET_BLOCKS_METADATA_NOFAIL); 630445586c70SMasahiro Yamada } while (PTR_ERR(bh) == -ENOSPC && 6305c5e298aeSTheodore Ts'o ext4_should_retry_alloc(inode->i_sb, &retries)); 63061c215028STheodore Ts'o if (IS_ERR(bh)) 63071c215028STheodore Ts'o return PTR_ERR(bh); 6308ac27a0ecSDave Kleikamp if (!bh) 6309ac27a0ecSDave Kleikamp goto out; 63105d601255Sliang xie BUFFER_TRACE(bh, "get write access"); 6311617ba13bSMingming Cao err = ext4_journal_get_write_access(handle, bh); 6312ac27a0ecSDave Kleikamp if (err) { 6313ac27a0ecSDave Kleikamp brelse(bh); 63141c215028STheodore Ts'o return err; 6315ac27a0ecSDave Kleikamp } 6316ac27a0ecSDave Kleikamp lock_buffer(bh); 631767eeb568SDmitry Monakhov memcpy(bh->b_data+offset, data, len); 6318ac27a0ecSDave Kleikamp flush_dcache_page(bh->b_page); 6319ac27a0ecSDave Kleikamp unlock_buffer(bh); 63200390131bSFrank Mayhar err = ext4_handle_dirty_metadata(handle, NULL, bh); 6321ac27a0ecSDave Kleikamp brelse(bh); 6322ac27a0ecSDave Kleikamp out: 632367eeb568SDmitry Monakhov if (inode->i_size < off + len) { 632467eeb568SDmitry Monakhov i_size_write(inode, off + len); 6325617ba13bSMingming Cao EXT4_I(inode)->i_disksize = inode->i_size; 63264209ae12SHarshad Shirwadkar err2 = ext4_mark_inode_dirty(handle, inode); 63274209ae12SHarshad Shirwadkar if (unlikely(err2 && !err)) 63284209ae12SHarshad Shirwadkar err = err2; 632921f97697SJan Kara } 63304209ae12SHarshad Shirwadkar return err ? err : len; 6331ac27a0ecSDave Kleikamp } 6332ac27a0ecSDave Kleikamp #endif 6333ac27a0ecSDave Kleikamp 6334152a0836SAl Viro static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 6335152a0836SAl Viro const char *dev_name, void *data) 6336ac27a0ecSDave Kleikamp { 6337152a0836SAl Viro return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 6338ac27a0ecSDave Kleikamp } 6339ac27a0ecSDave Kleikamp 6340c290ea01SJan Kara #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 634124b58424STheodore Ts'o static inline void register_as_ext2(void) 634224b58424STheodore Ts'o { 634324b58424STheodore Ts'o int err = register_filesystem(&ext2_fs_type); 634424b58424STheodore Ts'o if (err) 634524b58424STheodore Ts'o printk(KERN_WARNING 634624b58424STheodore Ts'o "EXT4-fs: Unable to register as ext2 (%d)\n", err); 634724b58424STheodore Ts'o } 634824b58424STheodore Ts'o 634924b58424STheodore Ts'o static inline void unregister_as_ext2(void) 635024b58424STheodore Ts'o { 635124b58424STheodore Ts'o unregister_filesystem(&ext2_fs_type); 635224b58424STheodore Ts'o } 63532035e776STheodore Ts'o 63542035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb) 63552035e776STheodore Ts'o { 6356e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext2_incompat_features(sb)) 63572035e776STheodore Ts'o return 0; 6358bc98a42cSDavid Howells if (sb_rdonly(sb)) 63592035e776STheodore Ts'o return 1; 6360e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext2_ro_compat_features(sb)) 63612035e776STheodore Ts'o return 0; 63622035e776STheodore Ts'o return 1; 63632035e776STheodore Ts'o } 636424b58424STheodore Ts'o #else 636524b58424STheodore Ts'o static inline void register_as_ext2(void) { } 636624b58424STheodore Ts'o static inline void unregister_as_ext2(void) { } 63672035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 636824b58424STheodore Ts'o #endif 636924b58424STheodore Ts'o 637024b58424STheodore Ts'o static inline void register_as_ext3(void) 637124b58424STheodore Ts'o { 637224b58424STheodore Ts'o int err = register_filesystem(&ext3_fs_type); 637324b58424STheodore Ts'o if (err) 637424b58424STheodore Ts'o printk(KERN_WARNING 637524b58424STheodore Ts'o "EXT4-fs: Unable to register as ext3 (%d)\n", err); 637624b58424STheodore Ts'o } 637724b58424STheodore Ts'o 637824b58424STheodore Ts'o static inline void unregister_as_ext3(void) 637924b58424STheodore Ts'o { 638024b58424STheodore Ts'o unregister_filesystem(&ext3_fs_type); 638124b58424STheodore Ts'o } 63822035e776STheodore Ts'o 63832035e776STheodore Ts'o static inline int ext3_feature_set_ok(struct super_block *sb) 63842035e776STheodore Ts'o { 6385e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext3_incompat_features(sb)) 63862035e776STheodore Ts'o return 0; 6387e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal(sb)) 63882035e776STheodore Ts'o return 0; 6389bc98a42cSDavid Howells if (sb_rdonly(sb)) 63902035e776STheodore Ts'o return 1; 6391e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext3_ro_compat_features(sb)) 63922035e776STheodore Ts'o return 0; 63932035e776STheodore Ts'o return 1; 63942035e776STheodore Ts'o } 639524b58424STheodore Ts'o 639603010a33STheodore Ts'o static struct file_system_type ext4_fs_type = { 6397ac27a0ecSDave Kleikamp .owner = THIS_MODULE, 639803010a33STheodore Ts'o .name = "ext4", 6399152a0836SAl Viro .mount = ext4_mount, 6400ac27a0ecSDave Kleikamp .kill_sb = kill_block_super, 6401ac27a0ecSDave Kleikamp .fs_flags = FS_REQUIRES_DEV, 6402ac27a0ecSDave Kleikamp }; 64037f78e035SEric W. Biederman MODULE_ALIAS_FS("ext4"); 6404ac27a0ecSDave Kleikamp 6405e9e3bcecSEric Sandeen /* Shared across all ext4 file systems */ 6406e9e3bcecSEric Sandeen wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 6407e9e3bcecSEric Sandeen 64085dabfc78STheodore Ts'o static int __init ext4_init_fs(void) 6409ac27a0ecSDave Kleikamp { 6410e9e3bcecSEric Sandeen int i, err; 6411c9de560dSAlex Tomas 6412e294a537STheodore Ts'o ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 641307c0c5d8SAl Viro ext4_li_info = NULL; 641407c0c5d8SAl Viro mutex_init(&ext4_li_mtx); 641507c0c5d8SAl Viro 64169a4c8019SCarlos Maiolino /* Build-time check for flags consistency */ 641712e9b892SDmitry Monakhov ext4_check_flag_values(); 6418e9e3bcecSEric Sandeen 6419e142d052SJan Kara for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 6420e9e3bcecSEric Sandeen init_waitqueue_head(&ext4__ioend_wq[i]); 6421e9e3bcecSEric Sandeen 642251865fdaSZheng Liu err = ext4_init_es(); 64236fd058f7STheodore Ts'o if (err) 64246fd058f7STheodore Ts'o return err; 642551865fdaSZheng Liu 64261dc0aa46SEric Whitney err = ext4_init_pending(); 64271dc0aa46SEric Whitney if (err) 642822cfe4b4SEric Biggers goto out7; 642922cfe4b4SEric Biggers 643022cfe4b4SEric Biggers err = ext4_init_post_read_processing(); 643122cfe4b4SEric Biggers if (err) 64321dc0aa46SEric Whitney goto out6; 64331dc0aa46SEric Whitney 643451865fdaSZheng Liu err = ext4_init_pageio(); 643551865fdaSZheng Liu if (err) 6436b5799018STheodore Ts'o goto out5; 643751865fdaSZheng Liu 64385dabfc78STheodore Ts'o err = ext4_init_system_zone(); 6439bd2d0210STheodore Ts'o if (err) 6440b5799018STheodore Ts'o goto out4; 6441857ac889SLukas Czerner 6442b5799018STheodore Ts'o err = ext4_init_sysfs(); 6443dd68314cSTheodore Ts'o if (err) 6444b5799018STheodore Ts'o goto out3; 6445857ac889SLukas Czerner 64465dabfc78STheodore Ts'o err = ext4_init_mballoc(); 6447ac27a0ecSDave Kleikamp if (err) 6448c9de560dSAlex Tomas goto out2; 6449ac27a0ecSDave Kleikamp err = init_inodecache(); 6450ac27a0ecSDave Kleikamp if (err) 6451ac27a0ecSDave Kleikamp goto out1; 645224b58424STheodore Ts'o register_as_ext3(); 64532035e776STheodore Ts'o register_as_ext2(); 645403010a33STheodore Ts'o err = register_filesystem(&ext4_fs_type); 6455ac27a0ecSDave Kleikamp if (err) 6456ac27a0ecSDave Kleikamp goto out; 6457bfff6873SLukas Czerner 6458ac27a0ecSDave Kleikamp return 0; 6459ac27a0ecSDave Kleikamp out: 646024b58424STheodore Ts'o unregister_as_ext2(); 646124b58424STheodore Ts'o unregister_as_ext3(); 6462ac27a0ecSDave Kleikamp destroy_inodecache(); 6463ac27a0ecSDave Kleikamp out1: 64645dabfc78STheodore Ts'o ext4_exit_mballoc(); 64659c191f70ST Makphaibulchoke out2: 6466b5799018STheodore Ts'o ext4_exit_sysfs(); 6467b5799018STheodore Ts'o out3: 6468dd68314cSTheodore Ts'o ext4_exit_system_zone(); 6469b5799018STheodore Ts'o out4: 64705dabfc78STheodore Ts'o ext4_exit_pageio(); 6471b5799018STheodore Ts'o out5: 647222cfe4b4SEric Biggers ext4_exit_post_read_processing(); 64731dc0aa46SEric Whitney out6: 647422cfe4b4SEric Biggers ext4_exit_pending(); 647522cfe4b4SEric Biggers out7: 647651865fdaSZheng Liu ext4_exit_es(); 647751865fdaSZheng Liu 6478ac27a0ecSDave Kleikamp return err; 6479ac27a0ecSDave Kleikamp } 6480ac27a0ecSDave Kleikamp 64815dabfc78STheodore Ts'o static void __exit ext4_exit_fs(void) 6482ac27a0ecSDave Kleikamp { 6483bfff6873SLukas Czerner ext4_destroy_lazyinit_thread(); 648424b58424STheodore Ts'o unregister_as_ext2(); 648524b58424STheodore Ts'o unregister_as_ext3(); 648603010a33STheodore Ts'o unregister_filesystem(&ext4_fs_type); 6487ac27a0ecSDave Kleikamp destroy_inodecache(); 64885dabfc78STheodore Ts'o ext4_exit_mballoc(); 6489b5799018STheodore Ts'o ext4_exit_sysfs(); 64905dabfc78STheodore Ts'o ext4_exit_system_zone(); 64915dabfc78STheodore Ts'o ext4_exit_pageio(); 649222cfe4b4SEric Biggers ext4_exit_post_read_processing(); 6493dd12ed14SEric Sandeen ext4_exit_es(); 64941dc0aa46SEric Whitney ext4_exit_pending(); 6495ac27a0ecSDave Kleikamp } 6496ac27a0ecSDave Kleikamp 6497ac27a0ecSDave Kleikamp MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 649883982b6fSTheodore Ts'o MODULE_DESCRIPTION("Fourth Extended Filesystem"); 6499ac27a0ecSDave Kleikamp MODULE_LICENSE("GPL"); 65007ef79ad5STheodore Ts'o MODULE_SOFTDEP("pre: crc32c"); 65015dabfc78STheodore Ts'o module_init(ext4_init_fs) 65025dabfc78STheodore Ts'o module_exit(ext4_exit_fs) 6503