Back to home page

OSCL-LXR

 
 

    


0001 // SPDX-License-Identifier: GPL-2.0-only
0002 /*
0003  *  linux/fs/ext2/super.c
0004  *
0005  * Copyright (C) 1992, 1993, 1994, 1995
0006  * Remy Card (card@masi.ibp.fr)
0007  * Laboratoire MASI - Institut Blaise Pascal
0008  * Universite Pierre et Marie Curie (Paris VI)
0009  *
0010  *  from
0011  *
0012  *  linux/fs/minix/inode.c
0013  *
0014  *  Copyright (C) 1991, 1992  Linus Torvalds
0015  *
0016  *  Big-endian to little-endian byte-swapping/bitmaps by
0017  *        David S. Miller (davem@caip.rutgers.edu), 1995
0018  */
0019 
0020 #include <linux/module.h>
0021 #include <linux/string.h>
0022 #include <linux/fs.h>
0023 #include <linux/slab.h>
0024 #include <linux/init.h>
0025 #include <linux/blkdev.h>
0026 #include <linux/parser.h>
0027 #include <linux/random.h>
0028 #include <linux/buffer_head.h>
0029 #include <linux/exportfs.h>
0030 #include <linux/vfs.h>
0031 #include <linux/seq_file.h>
0032 #include <linux/mount.h>
0033 #include <linux/log2.h>
0034 #include <linux/quotaops.h>
0035 #include <linux/uaccess.h>
0036 #include <linux/dax.h>
0037 #include <linux/iversion.h>
0038 #include "ext2.h"
0039 #include "xattr.h"
0040 #include "acl.h"
0041 
0042 static void ext2_write_super(struct super_block *sb);
0043 static int ext2_remount (struct super_block * sb, int * flags, char * data);
0044 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
0045 static int ext2_sync_fs(struct super_block *sb, int wait);
0046 static int ext2_freeze(struct super_block *sb);
0047 static int ext2_unfreeze(struct super_block *sb);
0048 
0049 void ext2_error(struct super_block *sb, const char *function,
0050         const char *fmt, ...)
0051 {
0052     struct va_format vaf;
0053     va_list args;
0054     struct ext2_sb_info *sbi = EXT2_SB(sb);
0055     struct ext2_super_block *es = sbi->s_es;
0056 
0057     if (!sb_rdonly(sb)) {
0058         spin_lock(&sbi->s_lock);
0059         sbi->s_mount_state |= EXT2_ERROR_FS;
0060         es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
0061         spin_unlock(&sbi->s_lock);
0062         ext2_sync_super(sb, es, 1);
0063     }
0064 
0065     va_start(args, fmt);
0066 
0067     vaf.fmt = fmt;
0068     vaf.va = &args;
0069 
0070     printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
0071            sb->s_id, function, &vaf);
0072 
0073     va_end(args);
0074 
0075     if (test_opt(sb, ERRORS_PANIC))
0076         panic("EXT2-fs: panic from previous error\n");
0077     if (!sb_rdonly(sb) && test_opt(sb, ERRORS_RO)) {
0078         ext2_msg(sb, KERN_CRIT,
0079                  "error: remounting filesystem read-only");
0080         sb->s_flags |= SB_RDONLY;
0081     }
0082 }
0083 
0084 void ext2_msg(struct super_block *sb, const char *prefix,
0085         const char *fmt, ...)
0086 {
0087     struct va_format vaf;
0088     va_list args;
0089 
0090     va_start(args, fmt);
0091 
0092     vaf.fmt = fmt;
0093     vaf.va = &args;
0094 
0095     printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
0096 
0097     va_end(args);
0098 }
0099 
0100 /*
0101  * This must be called with sbi->s_lock held.
0102  */
0103 void ext2_update_dynamic_rev(struct super_block *sb)
0104 {
0105     struct ext2_super_block *es = EXT2_SB(sb)->s_es;
0106 
0107     if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
0108         return;
0109 
0110     ext2_msg(sb, KERN_WARNING,
0111              "warning: updating to rev %d because of "
0112              "new feature flag, running e2fsck is recommended",
0113              EXT2_DYNAMIC_REV);
0114 
0115     es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
0116     es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
0117     es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
0118     /* leave es->s_feature_*compat flags alone */
0119     /* es->s_uuid will be set by e2fsck if empty */
0120 
0121     /*
0122      * The rest of the superblock fields should be zero, and if not it
0123      * means they are likely already in use, so leave them alone.  We
0124      * can leave it up to e2fsck to clean up any inconsistencies there.
0125      */
0126 }
0127 
0128 #ifdef CONFIG_QUOTA
0129 static int ext2_quota_off(struct super_block *sb, int type);
0130 
0131 static void ext2_quota_off_umount(struct super_block *sb)
0132 {
0133     int type;
0134 
0135     for (type = 0; type < MAXQUOTAS; type++)
0136         ext2_quota_off(sb, type);
0137 }
0138 #else
0139 static inline void ext2_quota_off_umount(struct super_block *sb)
0140 {
0141 }
0142 #endif
0143 
0144 static void ext2_put_super (struct super_block * sb)
0145 {
0146     int db_count;
0147     int i;
0148     struct ext2_sb_info *sbi = EXT2_SB(sb);
0149 
0150     ext2_quota_off_umount(sb);
0151 
0152     ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
0153     sbi->s_ea_block_cache = NULL;
0154 
0155     if (!sb_rdonly(sb)) {
0156         struct ext2_super_block *es = sbi->s_es;
0157 
0158         spin_lock(&sbi->s_lock);
0159         es->s_state = cpu_to_le16(sbi->s_mount_state);
0160         spin_unlock(&sbi->s_lock);
0161         ext2_sync_super(sb, es, 1);
0162     }
0163     db_count = sbi->s_gdb_count;
0164     for (i = 0; i < db_count; i++)
0165         brelse(sbi->s_group_desc[i]);
0166     kfree(sbi->s_group_desc);
0167     kfree(sbi->s_debts);
0168     percpu_counter_destroy(&sbi->s_freeblocks_counter);
0169     percpu_counter_destroy(&sbi->s_freeinodes_counter);
0170     percpu_counter_destroy(&sbi->s_dirs_counter);
0171     brelse (sbi->s_sbh);
0172     sb->s_fs_info = NULL;
0173     kfree(sbi->s_blockgroup_lock);
0174     fs_put_dax(sbi->s_daxdev, NULL);
0175     kfree(sbi);
0176 }
0177 
0178 static struct kmem_cache * ext2_inode_cachep;
0179 
0180 static struct inode *ext2_alloc_inode(struct super_block *sb)
0181 {
0182     struct ext2_inode_info *ei;
0183     ei = alloc_inode_sb(sb, ext2_inode_cachep, GFP_KERNEL);
0184     if (!ei)
0185         return NULL;
0186     ei->i_block_alloc_info = NULL;
0187     inode_set_iversion(&ei->vfs_inode, 1);
0188 #ifdef CONFIG_QUOTA
0189     memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
0190 #endif
0191 
0192     return &ei->vfs_inode;
0193 }
0194 
0195 static void ext2_free_in_core_inode(struct inode *inode)
0196 {
0197     kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
0198 }
0199 
0200 static void init_once(void *foo)
0201 {
0202     struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
0203 
0204     rwlock_init(&ei->i_meta_lock);
0205 #ifdef CONFIG_EXT2_FS_XATTR
0206     init_rwsem(&ei->xattr_sem);
0207 #endif
0208     mutex_init(&ei->truncate_mutex);
0209     inode_init_once(&ei->vfs_inode);
0210 }
0211 
0212 static int __init init_inodecache(void)
0213 {
0214     ext2_inode_cachep = kmem_cache_create_usercopy("ext2_inode_cache",
0215                 sizeof(struct ext2_inode_info), 0,
0216                 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
0217                     SLAB_ACCOUNT),
0218                 offsetof(struct ext2_inode_info, i_data),
0219                 sizeof_field(struct ext2_inode_info, i_data),
0220                 init_once);
0221     if (ext2_inode_cachep == NULL)
0222         return -ENOMEM;
0223     return 0;
0224 }
0225 
0226 static void destroy_inodecache(void)
0227 {
0228     /*
0229      * Make sure all delayed rcu free inodes are flushed before we
0230      * destroy cache.
0231      */
0232     rcu_barrier();
0233     kmem_cache_destroy(ext2_inode_cachep);
0234 }
0235 
0236 static int ext2_show_options(struct seq_file *seq, struct dentry *root)
0237 {
0238     struct super_block *sb = root->d_sb;
0239     struct ext2_sb_info *sbi = EXT2_SB(sb);
0240     struct ext2_super_block *es = sbi->s_es;
0241     unsigned long def_mount_opts;
0242 
0243     spin_lock(&sbi->s_lock);
0244     def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
0245 
0246     if (sbi->s_sb_block != 1)
0247         seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
0248     if (test_opt(sb, MINIX_DF))
0249         seq_puts(seq, ",minixdf");
0250     if (test_opt(sb, GRPID))
0251         seq_puts(seq, ",grpid");
0252     if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
0253         seq_puts(seq, ",nogrpid");
0254     if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
0255         le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
0256         seq_printf(seq, ",resuid=%u",
0257                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
0258     }
0259     if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
0260         le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
0261         seq_printf(seq, ",resgid=%u",
0262                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
0263     }
0264     if (test_opt(sb, ERRORS_RO)) {
0265         int def_errors = le16_to_cpu(es->s_errors);
0266 
0267         if (def_errors == EXT2_ERRORS_PANIC ||
0268             def_errors == EXT2_ERRORS_CONTINUE) {
0269             seq_puts(seq, ",errors=remount-ro");
0270         }
0271     }
0272     if (test_opt(sb, ERRORS_CONT))
0273         seq_puts(seq, ",errors=continue");
0274     if (test_opt(sb, ERRORS_PANIC))
0275         seq_puts(seq, ",errors=panic");
0276     if (test_opt(sb, NO_UID32))
0277         seq_puts(seq, ",nouid32");
0278     if (test_opt(sb, DEBUG))
0279         seq_puts(seq, ",debug");
0280     if (test_opt(sb, OLDALLOC))
0281         seq_puts(seq, ",oldalloc");
0282 
0283 #ifdef CONFIG_EXT2_FS_XATTR
0284     if (test_opt(sb, XATTR_USER))
0285         seq_puts(seq, ",user_xattr");
0286     if (!test_opt(sb, XATTR_USER) &&
0287         (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
0288         seq_puts(seq, ",nouser_xattr");
0289     }
0290 #endif
0291 
0292 #ifdef CONFIG_EXT2_FS_POSIX_ACL
0293     if (test_opt(sb, POSIX_ACL))
0294         seq_puts(seq, ",acl");
0295     if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
0296         seq_puts(seq, ",noacl");
0297 #endif
0298 
0299     if (test_opt(sb, USRQUOTA))
0300         seq_puts(seq, ",usrquota");
0301 
0302     if (test_opt(sb, GRPQUOTA))
0303         seq_puts(seq, ",grpquota");
0304 
0305     if (test_opt(sb, XIP))
0306         seq_puts(seq, ",xip");
0307 
0308     if (test_opt(sb, DAX))
0309         seq_puts(seq, ",dax");
0310 
0311     if (!test_opt(sb, RESERVATION))
0312         seq_puts(seq, ",noreservation");
0313 
0314     spin_unlock(&sbi->s_lock);
0315     return 0;
0316 }
0317 
0318 #ifdef CONFIG_QUOTA
0319 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
0320 static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
0321 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
0322              const struct path *path);
0323 static struct dquot **ext2_get_dquots(struct inode *inode)
0324 {
0325     return EXT2_I(inode)->i_dquot;
0326 }
0327 
0328 static const struct quotactl_ops ext2_quotactl_ops = {
0329     .quota_on   = ext2_quota_on,
0330     .quota_off  = ext2_quota_off,
0331     .quota_sync = dquot_quota_sync,
0332     .get_state  = dquot_get_state,
0333     .set_info   = dquot_set_dqinfo,
0334     .get_dqblk  = dquot_get_dqblk,
0335     .set_dqblk  = dquot_set_dqblk,
0336     .get_nextdqblk  = dquot_get_next_dqblk,
0337 };
0338 #endif
0339 
0340 static const struct super_operations ext2_sops = {
0341     .alloc_inode    = ext2_alloc_inode,
0342     .free_inode = ext2_free_in_core_inode,
0343     .write_inode    = ext2_write_inode,
0344     .evict_inode    = ext2_evict_inode,
0345     .put_super  = ext2_put_super,
0346     .sync_fs    = ext2_sync_fs,
0347     .freeze_fs  = ext2_freeze,
0348     .unfreeze_fs    = ext2_unfreeze,
0349     .statfs     = ext2_statfs,
0350     .remount_fs = ext2_remount,
0351     .show_options   = ext2_show_options,
0352 #ifdef CONFIG_QUOTA
0353     .quota_read = ext2_quota_read,
0354     .quota_write    = ext2_quota_write,
0355     .get_dquots = ext2_get_dquots,
0356 #endif
0357 };
0358 
0359 static struct inode *ext2_nfs_get_inode(struct super_block *sb,
0360         u64 ino, u32 generation)
0361 {
0362     struct inode *inode;
0363 
0364     if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
0365         return ERR_PTR(-ESTALE);
0366     if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
0367         return ERR_PTR(-ESTALE);
0368 
0369     /*
0370      * ext2_iget isn't quite right if the inode is currently unallocated!
0371      * However ext2_iget currently does appropriate checks to handle stale
0372      * inodes so everything is OK.
0373      */
0374     inode = ext2_iget(sb, ino);
0375     if (IS_ERR(inode))
0376         return ERR_CAST(inode);
0377     if (generation && inode->i_generation != generation) {
0378         /* we didn't find the right inode.. */
0379         iput(inode);
0380         return ERR_PTR(-ESTALE);
0381     }
0382     return inode;
0383 }
0384 
0385 static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
0386         int fh_len, int fh_type)
0387 {
0388     return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
0389                     ext2_nfs_get_inode);
0390 }
0391 
0392 static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
0393         int fh_len, int fh_type)
0394 {
0395     return generic_fh_to_parent(sb, fid, fh_len, fh_type,
0396                     ext2_nfs_get_inode);
0397 }
0398 
0399 static const struct export_operations ext2_export_ops = {
0400     .fh_to_dentry = ext2_fh_to_dentry,
0401     .fh_to_parent = ext2_fh_to_parent,
0402     .get_parent = ext2_get_parent,
0403 };
0404 
0405 static unsigned long get_sb_block(void **data)
0406 {
0407     unsigned long   sb_block;
0408     char        *options = (char *) *data;
0409 
0410     if (!options || strncmp(options, "sb=", 3) != 0)
0411         return 1;   /* Default location */
0412     options += 3;
0413     sb_block = simple_strtoul(options, &options, 0);
0414     if (*options && *options != ',') {
0415         printk("EXT2-fs: Invalid sb specification: %s\n",
0416                (char *) *data);
0417         return 1;
0418     }
0419     if (*options == ',')
0420         options++;
0421     *data = (void *) options;
0422     return sb_block;
0423 }
0424 
0425 enum {
0426     Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
0427     Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
0428     Opt_err_ro, Opt_nouid32, Opt_debug,
0429     Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
0430     Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
0431     Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
0432 };
0433 
0434 static const match_table_t tokens = {
0435     {Opt_bsd_df, "bsddf"},
0436     {Opt_minix_df, "minixdf"},
0437     {Opt_grpid, "grpid"},
0438     {Opt_grpid, "bsdgroups"},
0439     {Opt_nogrpid, "nogrpid"},
0440     {Opt_nogrpid, "sysvgroups"},
0441     {Opt_resgid, "resgid=%u"},
0442     {Opt_resuid, "resuid=%u"},
0443     {Opt_sb, "sb=%u"},
0444     {Opt_err_cont, "errors=continue"},
0445     {Opt_err_panic, "errors=panic"},
0446     {Opt_err_ro, "errors=remount-ro"},
0447     {Opt_nouid32, "nouid32"},
0448     {Opt_debug, "debug"},
0449     {Opt_oldalloc, "oldalloc"},
0450     {Opt_orlov, "orlov"},
0451     {Opt_nobh, "nobh"},
0452     {Opt_user_xattr, "user_xattr"},
0453     {Opt_nouser_xattr, "nouser_xattr"},
0454     {Opt_acl, "acl"},
0455     {Opt_noacl, "noacl"},
0456     {Opt_xip, "xip"},
0457     {Opt_dax, "dax"},
0458     {Opt_grpquota, "grpquota"},
0459     {Opt_ignore, "noquota"},
0460     {Opt_quota, "quota"},
0461     {Opt_usrquota, "usrquota"},
0462     {Opt_reservation, "reservation"},
0463     {Opt_noreservation, "noreservation"},
0464     {Opt_err, NULL}
0465 };
0466 
0467 static int parse_options(char *options, struct super_block *sb,
0468              struct ext2_mount_options *opts)
0469 {
0470     char *p;
0471     substring_t args[MAX_OPT_ARGS];
0472     int option;
0473     kuid_t uid;
0474     kgid_t gid;
0475 
0476     if (!options)
0477         return 1;
0478 
0479     while ((p = strsep (&options, ",")) != NULL) {
0480         int token;
0481         if (!*p)
0482             continue;
0483 
0484         token = match_token(p, tokens, args);
0485         switch (token) {
0486         case Opt_bsd_df:
0487             clear_opt (opts->s_mount_opt, MINIX_DF);
0488             break;
0489         case Opt_minix_df:
0490             set_opt (opts->s_mount_opt, MINIX_DF);
0491             break;
0492         case Opt_grpid:
0493             set_opt (opts->s_mount_opt, GRPID);
0494             break;
0495         case Opt_nogrpid:
0496             clear_opt (opts->s_mount_opt, GRPID);
0497             break;
0498         case Opt_resuid:
0499             if (match_int(&args[0], &option))
0500                 return 0;
0501             uid = make_kuid(current_user_ns(), option);
0502             if (!uid_valid(uid)) {
0503                 ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
0504                 return 0;
0505 
0506             }
0507             opts->s_resuid = uid;
0508             break;
0509         case Opt_resgid:
0510             if (match_int(&args[0], &option))
0511                 return 0;
0512             gid = make_kgid(current_user_ns(), option);
0513             if (!gid_valid(gid)) {
0514                 ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
0515                 return 0;
0516             }
0517             opts->s_resgid = gid;
0518             break;
0519         case Opt_sb:
0520             /* handled by get_sb_block() instead of here */
0521             /* *sb_block = match_int(&args[0]); */
0522             break;
0523         case Opt_err_panic:
0524             clear_opt (opts->s_mount_opt, ERRORS_CONT);
0525             clear_opt (opts->s_mount_opt, ERRORS_RO);
0526             set_opt (opts->s_mount_opt, ERRORS_PANIC);
0527             break;
0528         case Opt_err_ro:
0529             clear_opt (opts->s_mount_opt, ERRORS_CONT);
0530             clear_opt (opts->s_mount_opt, ERRORS_PANIC);
0531             set_opt (opts->s_mount_opt, ERRORS_RO);
0532             break;
0533         case Opt_err_cont:
0534             clear_opt (opts->s_mount_opt, ERRORS_RO);
0535             clear_opt (opts->s_mount_opt, ERRORS_PANIC);
0536             set_opt (opts->s_mount_opt, ERRORS_CONT);
0537             break;
0538         case Opt_nouid32:
0539             set_opt (opts->s_mount_opt, NO_UID32);
0540             break;
0541         case Opt_debug:
0542             set_opt (opts->s_mount_opt, DEBUG);
0543             break;
0544         case Opt_oldalloc:
0545             set_opt (opts->s_mount_opt, OLDALLOC);
0546             break;
0547         case Opt_orlov:
0548             clear_opt (opts->s_mount_opt, OLDALLOC);
0549             break;
0550         case Opt_nobh:
0551             ext2_msg(sb, KERN_INFO,
0552                 "nobh option not supported");
0553             break;
0554 #ifdef CONFIG_EXT2_FS_XATTR
0555         case Opt_user_xattr:
0556             set_opt (opts->s_mount_opt, XATTR_USER);
0557             break;
0558         case Opt_nouser_xattr:
0559             clear_opt (opts->s_mount_opt, XATTR_USER);
0560             break;
0561 #else
0562         case Opt_user_xattr:
0563         case Opt_nouser_xattr:
0564             ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
0565                 "not supported");
0566             break;
0567 #endif
0568 #ifdef CONFIG_EXT2_FS_POSIX_ACL
0569         case Opt_acl:
0570             set_opt(opts->s_mount_opt, POSIX_ACL);
0571             break;
0572         case Opt_noacl:
0573             clear_opt(opts->s_mount_opt, POSIX_ACL);
0574             break;
0575 #else
0576         case Opt_acl:
0577         case Opt_noacl:
0578             ext2_msg(sb, KERN_INFO,
0579                 "(no)acl options not supported");
0580             break;
0581 #endif
0582         case Opt_xip:
0583             ext2_msg(sb, KERN_INFO, "use dax instead of xip");
0584             set_opt(opts->s_mount_opt, XIP);
0585             fallthrough;
0586         case Opt_dax:
0587 #ifdef CONFIG_FS_DAX
0588             ext2_msg(sb, KERN_WARNING,
0589         "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
0590             set_opt(opts->s_mount_opt, DAX);
0591 #else
0592             ext2_msg(sb, KERN_INFO, "dax option not supported");
0593 #endif
0594             break;
0595 
0596 #if defined(CONFIG_QUOTA)
0597         case Opt_quota:
0598         case Opt_usrquota:
0599             set_opt(opts->s_mount_opt, USRQUOTA);
0600             break;
0601 
0602         case Opt_grpquota:
0603             set_opt(opts->s_mount_opt, GRPQUOTA);
0604             break;
0605 #else
0606         case Opt_quota:
0607         case Opt_usrquota:
0608         case Opt_grpquota:
0609             ext2_msg(sb, KERN_INFO,
0610                 "quota operations not supported");
0611             break;
0612 #endif
0613 
0614         case Opt_reservation:
0615             set_opt(opts->s_mount_opt, RESERVATION);
0616             ext2_msg(sb, KERN_INFO, "reservations ON");
0617             break;
0618         case Opt_noreservation:
0619             clear_opt(opts->s_mount_opt, RESERVATION);
0620             ext2_msg(sb, KERN_INFO, "reservations OFF");
0621             break;
0622         case Opt_ignore:
0623             break;
0624         default:
0625             return 0;
0626         }
0627     }
0628     return 1;
0629 }
0630 
0631 static int ext2_setup_super (struct super_block * sb,
0632                   struct ext2_super_block * es,
0633                   int read_only)
0634 {
0635     int res = 0;
0636     struct ext2_sb_info *sbi = EXT2_SB(sb);
0637 
0638     if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
0639         ext2_msg(sb, KERN_ERR,
0640             "error: revision level too high, "
0641             "forcing read-only mode");
0642         res = SB_RDONLY;
0643     }
0644     if (read_only)
0645         return res;
0646     if (!(sbi->s_mount_state & EXT2_VALID_FS))
0647         ext2_msg(sb, KERN_WARNING,
0648             "warning: mounting unchecked fs, "
0649             "running e2fsck is recommended");
0650     else if ((sbi->s_mount_state & EXT2_ERROR_FS))
0651         ext2_msg(sb, KERN_WARNING,
0652             "warning: mounting fs with errors, "
0653             "running e2fsck is recommended");
0654     else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
0655          le16_to_cpu(es->s_mnt_count) >=
0656          (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
0657         ext2_msg(sb, KERN_WARNING,
0658             "warning: maximal mount count reached, "
0659             "running e2fsck is recommended");
0660     else if (le32_to_cpu(es->s_checkinterval) &&
0661         (le32_to_cpu(es->s_lastcheck) +
0662             le32_to_cpu(es->s_checkinterval) <=
0663             ktime_get_real_seconds()))
0664         ext2_msg(sb, KERN_WARNING,
0665             "warning: checktime reached, "
0666             "running e2fsck is recommended");
0667     if (!le16_to_cpu(es->s_max_mnt_count))
0668         es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
0669     le16_add_cpu(&es->s_mnt_count, 1);
0670     if (test_opt (sb, DEBUG))
0671         ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
0672             "bpg=%lu, ipg=%lu, mo=%04lx]",
0673             EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
0674             sbi->s_frag_size,
0675             sbi->s_groups_count,
0676             EXT2_BLOCKS_PER_GROUP(sb),
0677             EXT2_INODES_PER_GROUP(sb),
0678             sbi->s_mount_opt);
0679     return res;
0680 }
0681 
0682 static int ext2_check_descriptors(struct super_block *sb)
0683 {
0684     int i;
0685     struct ext2_sb_info *sbi = EXT2_SB(sb);
0686 
0687     ext2_debug ("Checking group descriptors");
0688 
0689     for (i = 0; i < sbi->s_groups_count; i++) {
0690         struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
0691         ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
0692         ext2_fsblk_t last_block = ext2_group_last_block_no(sb, i);
0693 
0694         if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
0695             le32_to_cpu(gdp->bg_block_bitmap) > last_block)
0696         {
0697             ext2_error (sb, "ext2_check_descriptors",
0698                     "Block bitmap for group %d"
0699                     " not in group (block %lu)!",
0700                     i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
0701             return 0;
0702         }
0703         if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
0704             le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
0705         {
0706             ext2_error (sb, "ext2_check_descriptors",
0707                     "Inode bitmap for group %d"
0708                     " not in group (block %lu)!",
0709                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
0710             return 0;
0711         }
0712         if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
0713             le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
0714             last_block)
0715         {
0716             ext2_error (sb, "ext2_check_descriptors",
0717                     "Inode table for group %d"
0718                     " not in group (block %lu)!",
0719                     i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
0720             return 0;
0721         }
0722     }
0723     return 1;
0724 }
0725 
0726 /*
0727  * Maximal file size.  There is a direct, and {,double-,triple-}indirect
0728  * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
0729  * We need to be 1 filesystem block less than the 2^32 sector limit.
0730  */
0731 static loff_t ext2_max_size(int bits)
0732 {
0733     loff_t res = EXT2_NDIR_BLOCKS;
0734     int meta_blocks;
0735     unsigned int upper_limit;
0736     unsigned int ppb = 1 << (bits-2);
0737 
0738     /* This is calculated to be the largest file size for a
0739      * dense, file such that the total number of
0740      * sectors in the file, including data and all indirect blocks,
0741      * does not exceed 2^32 -1
0742      * __u32 i_blocks representing the total number of
0743      * 512 bytes blocks of the file
0744      */
0745     upper_limit = (1LL << 32) - 1;
0746 
0747     /* total blocks in file system block size */
0748     upper_limit >>= (bits - 9);
0749 
0750     /* Compute how many blocks we can address by block tree */
0751     res += 1LL << (bits-2);
0752     res += 1LL << (2*(bits-2));
0753     res += 1LL << (3*(bits-2));
0754     /* Compute how many metadata blocks are needed */
0755     meta_blocks = 1;
0756     meta_blocks += 1 + ppb;
0757     meta_blocks += 1 + ppb + ppb * ppb;
0758     /* Does block tree limit file size? */
0759     if (res + meta_blocks <= upper_limit)
0760         goto check_lfs;
0761 
0762     res = upper_limit;
0763     /* How many metadata blocks are needed for addressing upper_limit? */
0764     upper_limit -= EXT2_NDIR_BLOCKS;
0765     /* indirect blocks */
0766     meta_blocks = 1;
0767     upper_limit -= ppb;
0768     /* double indirect blocks */
0769     if (upper_limit < ppb * ppb) {
0770         meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb);
0771         res -= meta_blocks;
0772         goto check_lfs;
0773     }
0774     meta_blocks += 1 + ppb;
0775     upper_limit -= ppb * ppb;
0776     /* tripple indirect blocks for the rest */
0777     meta_blocks += 1 + DIV_ROUND_UP(upper_limit, ppb) +
0778         DIV_ROUND_UP(upper_limit, ppb*ppb);
0779     res -= meta_blocks;
0780 check_lfs:
0781     res <<= bits;
0782     if (res > MAX_LFS_FILESIZE)
0783         res = MAX_LFS_FILESIZE;
0784 
0785     return res;
0786 }
0787 
0788 static unsigned long descriptor_loc(struct super_block *sb,
0789                     unsigned long logic_sb_block,
0790                     int nr)
0791 {
0792     struct ext2_sb_info *sbi = EXT2_SB(sb);
0793     unsigned long bg, first_meta_bg;
0794     
0795     first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
0796 
0797     if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
0798         nr < first_meta_bg)
0799         return (logic_sb_block + nr + 1);
0800     bg = sbi->s_desc_per_block * nr;
0801 
0802     return ext2_group_first_block_no(sb, bg) + ext2_bg_has_super(sb, bg);
0803 }
0804 
0805 static int ext2_fill_super(struct super_block *sb, void *data, int silent)
0806 {
0807     struct buffer_head * bh;
0808     struct ext2_sb_info * sbi;
0809     struct ext2_super_block * es;
0810     struct inode *root;
0811     unsigned long block;
0812     unsigned long sb_block = get_sb_block(&data);
0813     unsigned long logic_sb_block;
0814     unsigned long offset = 0;
0815     unsigned long def_mount_opts;
0816     long ret = -ENOMEM;
0817     int blocksize = BLOCK_SIZE;
0818     int db_count;
0819     int i, j;
0820     __le32 features;
0821     int err;
0822     struct ext2_mount_options opts;
0823 
0824     sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
0825     if (!sbi)
0826         return -ENOMEM;
0827 
0828     sbi->s_blockgroup_lock =
0829         kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
0830     if (!sbi->s_blockgroup_lock) {
0831         kfree(sbi);
0832         return -ENOMEM;
0833     }
0834     sb->s_fs_info = sbi;
0835     sbi->s_sb_block = sb_block;
0836     sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
0837                        NULL, NULL);
0838 
0839     spin_lock_init(&sbi->s_lock);
0840     ret = -EINVAL;
0841 
0842     /*
0843      * See what the current blocksize for the device is, and
0844      * use that as the blocksize.  Otherwise (or if the blocksize
0845      * is smaller than the default) use the default.
0846      * This is important for devices that have a hardware
0847      * sectorsize that is larger than the default.
0848      */
0849     blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
0850     if (!blocksize) {
0851         ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
0852         goto failed_sbi;
0853     }
0854 
0855     /*
0856      * If the superblock doesn't start on a hardware sector boundary,
0857      * calculate the offset.  
0858      */
0859     if (blocksize != BLOCK_SIZE) {
0860         logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
0861         offset = (sb_block*BLOCK_SIZE) % blocksize;
0862     } else {
0863         logic_sb_block = sb_block;
0864     }
0865 
0866     if (!(bh = sb_bread(sb, logic_sb_block))) {
0867         ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
0868         goto failed_sbi;
0869     }
0870     /*
0871      * Note: s_es must be initialized as soon as possible because
0872      *       some ext2 macro-instructions depend on its value
0873      */
0874     es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
0875     sbi->s_es = es;
0876     sb->s_magic = le16_to_cpu(es->s_magic);
0877 
0878     if (sb->s_magic != EXT2_SUPER_MAGIC)
0879         goto cantfind_ext2;
0880 
0881     opts.s_mount_opt = 0;
0882     /* Set defaults before we parse the mount options */
0883     def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
0884     if (def_mount_opts & EXT2_DEFM_DEBUG)
0885         set_opt(opts.s_mount_opt, DEBUG);
0886     if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
0887         set_opt(opts.s_mount_opt, GRPID);
0888     if (def_mount_opts & EXT2_DEFM_UID16)
0889         set_opt(opts.s_mount_opt, NO_UID32);
0890 #ifdef CONFIG_EXT2_FS_XATTR
0891     if (def_mount_opts & EXT2_DEFM_XATTR_USER)
0892         set_opt(opts.s_mount_opt, XATTR_USER);
0893 #endif
0894 #ifdef CONFIG_EXT2_FS_POSIX_ACL
0895     if (def_mount_opts & EXT2_DEFM_ACL)
0896         set_opt(opts.s_mount_opt, POSIX_ACL);
0897 #endif
0898     
0899     if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
0900         set_opt(opts.s_mount_opt, ERRORS_PANIC);
0901     else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
0902         set_opt(opts.s_mount_opt, ERRORS_CONT);
0903     else
0904         set_opt(opts.s_mount_opt, ERRORS_RO);
0905 
0906     opts.s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
0907     opts.s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
0908     
0909     set_opt(opts.s_mount_opt, RESERVATION);
0910 
0911     if (!parse_options((char *) data, sb, &opts))
0912         goto failed_mount;
0913 
0914     sbi->s_mount_opt = opts.s_mount_opt;
0915     sbi->s_resuid = opts.s_resuid;
0916     sbi->s_resgid = opts.s_resgid;
0917 
0918     sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
0919         (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
0920     sb->s_iflags |= SB_I_CGROUPWB;
0921 
0922     if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
0923         (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
0924          EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
0925          EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
0926         ext2_msg(sb, KERN_WARNING,
0927             "warning: feature flags set on rev 0 fs, "
0928             "running e2fsck is recommended");
0929     /*
0930      * Check feature flags regardless of the revision level, since we
0931      * previously didn't change the revision level when setting the flags,
0932      * so there is a chance incompat flags are set on a rev 0 filesystem.
0933      */
0934     features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
0935     if (features) {
0936         ext2_msg(sb, KERN_ERR,  "error: couldn't mount because of "
0937                "unsupported optional features (%x)",
0938             le32_to_cpu(features));
0939         goto failed_mount;
0940     }
0941     if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
0942         ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
0943                "unsupported optional features (%x)",
0944                le32_to_cpu(features));
0945         goto failed_mount;
0946     }
0947 
0948     blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
0949 
0950     if (test_opt(sb, DAX)) {
0951         if (!sbi->s_daxdev) {
0952             ext2_msg(sb, KERN_ERR,
0953                 "DAX unsupported by block device. Turning off DAX.");
0954             clear_opt(sbi->s_mount_opt, DAX);
0955         } else if (blocksize != PAGE_SIZE) {
0956             ext2_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
0957             clear_opt(sbi->s_mount_opt, DAX);
0958         }
0959     }
0960 
0961     /* If the blocksize doesn't match, re-read the thing.. */
0962     if (sb->s_blocksize != blocksize) {
0963         brelse(bh);
0964 
0965         if (!sb_set_blocksize(sb, blocksize)) {
0966             ext2_msg(sb, KERN_ERR,
0967                 "error: bad blocksize %d", blocksize);
0968             goto failed_sbi;
0969         }
0970 
0971         logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
0972         offset = (sb_block*BLOCK_SIZE) % blocksize;
0973         bh = sb_bread(sb, logic_sb_block);
0974         if(!bh) {
0975             ext2_msg(sb, KERN_ERR, "error: couldn't read"
0976                 "superblock on 2nd try");
0977             goto failed_sbi;
0978         }
0979         es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
0980         sbi->s_es = es;
0981         if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
0982             ext2_msg(sb, KERN_ERR, "error: magic mismatch");
0983             goto failed_mount;
0984         }
0985     }
0986 
0987     sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
0988     sb->s_max_links = EXT2_LINK_MAX;
0989     sb->s_time_min = S32_MIN;
0990     sb->s_time_max = S32_MAX;
0991 
0992     if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
0993         sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
0994         sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
0995     } else {
0996         sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
0997         sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
0998         if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
0999             !is_power_of_2(sbi->s_inode_size) ||
1000             (sbi->s_inode_size > blocksize)) {
1001             ext2_msg(sb, KERN_ERR,
1002                 "error: unsupported inode size: %d",
1003                 sbi->s_inode_size);
1004             goto failed_mount;
1005         }
1006     }
1007 
1008     sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
1009                    le32_to_cpu(es->s_log_frag_size);
1010     if (sbi->s_frag_size == 0)
1011         goto cantfind_ext2;
1012     sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
1013 
1014     sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1015     sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1016     sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1017 
1018     sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
1019     if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
1020         goto cantfind_ext2;
1021     sbi->s_itb_per_group = sbi->s_inodes_per_group /
1022                     sbi->s_inodes_per_block;
1023     sbi->s_desc_per_block = sb->s_blocksize /
1024                     sizeof (struct ext2_group_desc);
1025     sbi->s_sbh = bh;
1026     sbi->s_mount_state = le16_to_cpu(es->s_state);
1027     sbi->s_addr_per_block_bits =
1028         ilog2 (EXT2_ADDR_PER_BLOCK(sb));
1029     sbi->s_desc_per_block_bits =
1030         ilog2 (EXT2_DESC_PER_BLOCK(sb));
1031 
1032     if (sb->s_magic != EXT2_SUPER_MAGIC)
1033         goto cantfind_ext2;
1034 
1035     if (sb->s_blocksize != bh->b_size) {
1036         if (!silent)
1037             ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
1038         goto failed_mount;
1039     }
1040 
1041     if (sb->s_blocksize != sbi->s_frag_size) {
1042         ext2_msg(sb, KERN_ERR,
1043             "error: fragsize %lu != blocksize %lu"
1044             "(not supported yet)",
1045             sbi->s_frag_size, sb->s_blocksize);
1046         goto failed_mount;
1047     }
1048 
1049     if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
1050         ext2_msg(sb, KERN_ERR,
1051             "error: #blocks per group too big: %lu",
1052             sbi->s_blocks_per_group);
1053         goto failed_mount;
1054     }
1055     if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
1056         ext2_msg(sb, KERN_ERR,
1057             "error: #fragments per group too big: %lu",
1058             sbi->s_frags_per_group);
1059         goto failed_mount;
1060     }
1061     if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
1062         sbi->s_inodes_per_group > sb->s_blocksize * 8) {
1063         ext2_msg(sb, KERN_ERR,
1064             "error: invalid #inodes per group: %lu",
1065             sbi->s_inodes_per_group);
1066         goto failed_mount;
1067     }
1068 
1069     if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
1070         goto cantfind_ext2;
1071     sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1072                 le32_to_cpu(es->s_first_data_block) - 1)
1073                     / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
1074     if ((u64)sbi->s_groups_count * sbi->s_inodes_per_group !=
1075         le32_to_cpu(es->s_inodes_count)) {
1076         ext2_msg(sb, KERN_ERR, "error: invalid #inodes: %u vs computed %llu",
1077              le32_to_cpu(es->s_inodes_count),
1078              (u64)sbi->s_groups_count * sbi->s_inodes_per_group);
1079         goto failed_mount;
1080     }
1081     db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
1082            EXT2_DESC_PER_BLOCK(sb);
1083     sbi->s_group_desc = kmalloc_array(db_count,
1084                        sizeof(struct buffer_head *),
1085                        GFP_KERNEL);
1086     if (sbi->s_group_desc == NULL) {
1087         ret = -ENOMEM;
1088         ext2_msg(sb, KERN_ERR, "error: not enough memory");
1089         goto failed_mount;
1090     }
1091     bgl_lock_init(sbi->s_blockgroup_lock);
1092     sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
1093     if (!sbi->s_debts) {
1094         ret = -ENOMEM;
1095         ext2_msg(sb, KERN_ERR, "error: not enough memory");
1096         goto failed_mount_group_desc;
1097     }
1098     for (i = 0; i < db_count; i++) {
1099         block = descriptor_loc(sb, logic_sb_block, i);
1100         sbi->s_group_desc[i] = sb_bread(sb, block);
1101         if (!sbi->s_group_desc[i]) {
1102             for (j = 0; j < i; j++)
1103                 brelse (sbi->s_group_desc[j]);
1104             ext2_msg(sb, KERN_ERR,
1105                 "error: unable to read group descriptors");
1106             goto failed_mount_group_desc;
1107         }
1108     }
1109     if (!ext2_check_descriptors (sb)) {
1110         ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
1111         goto failed_mount2;
1112     }
1113     sbi->s_gdb_count = db_count;
1114     get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1115     spin_lock_init(&sbi->s_next_gen_lock);
1116 
1117     /* per filesystem reservation list head & lock */
1118     spin_lock_init(&sbi->s_rsv_window_lock);
1119     sbi->s_rsv_window_root = RB_ROOT;
1120     /*
1121      * Add a single, static dummy reservation to the start of the
1122      * reservation window list --- it gives us a placeholder for
1123      * append-at-start-of-list which makes the allocation logic
1124      * _much_ simpler.
1125      */
1126     sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1127     sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
1128     sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1129     sbi->s_rsv_window_head.rsv_goal_size = 0;
1130     ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
1131 
1132     err = percpu_counter_init(&sbi->s_freeblocks_counter,
1133                 ext2_count_free_blocks(sb), GFP_KERNEL);
1134     if (!err) {
1135         err = percpu_counter_init(&sbi->s_freeinodes_counter,
1136                 ext2_count_free_inodes(sb), GFP_KERNEL);
1137     }
1138     if (!err) {
1139         err = percpu_counter_init(&sbi->s_dirs_counter,
1140                 ext2_count_dirs(sb), GFP_KERNEL);
1141     }
1142     if (err) {
1143         ret = err;
1144         ext2_msg(sb, KERN_ERR, "error: insufficient memory");
1145         goto failed_mount3;
1146     }
1147 
1148 #ifdef CONFIG_EXT2_FS_XATTR
1149     sbi->s_ea_block_cache = ext2_xattr_create_cache();
1150     if (!sbi->s_ea_block_cache) {
1151         ret = -ENOMEM;
1152         ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
1153         goto failed_mount3;
1154     }
1155 #endif
1156     /*
1157      * set up enough so that it can read an inode
1158      */
1159     sb->s_op = &ext2_sops;
1160     sb->s_export_op = &ext2_export_ops;
1161     sb->s_xattr = ext2_xattr_handlers;
1162 
1163 #ifdef CONFIG_QUOTA
1164     sb->dq_op = &dquot_operations;
1165     sb->s_qcop = &ext2_quotactl_ops;
1166     sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
1167 #endif
1168 
1169     root = ext2_iget(sb, EXT2_ROOT_INO);
1170     if (IS_ERR(root)) {
1171         ret = PTR_ERR(root);
1172         goto failed_mount3;
1173     }
1174     if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1175         iput(root);
1176         ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
1177         goto failed_mount3;
1178     }
1179 
1180     sb->s_root = d_make_root(root);
1181     if (!sb->s_root) {
1182         ext2_msg(sb, KERN_ERR, "error: get root inode failed");
1183         ret = -ENOMEM;
1184         goto failed_mount3;
1185     }
1186     if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
1187         ext2_msg(sb, KERN_WARNING,
1188             "warning: mounting ext3 filesystem as ext2");
1189     if (ext2_setup_super (sb, es, sb_rdonly(sb)))
1190         sb->s_flags |= SB_RDONLY;
1191     ext2_write_super(sb);
1192     return 0;
1193 
1194 cantfind_ext2:
1195     if (!silent)
1196         ext2_msg(sb, KERN_ERR,
1197             "error: can't find an ext2 filesystem on dev %s.",
1198             sb->s_id);
1199     goto failed_mount;
1200 failed_mount3:
1201     ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
1202     percpu_counter_destroy(&sbi->s_freeblocks_counter);
1203     percpu_counter_destroy(&sbi->s_freeinodes_counter);
1204     percpu_counter_destroy(&sbi->s_dirs_counter);
1205 failed_mount2:
1206     for (i = 0; i < db_count; i++)
1207         brelse(sbi->s_group_desc[i]);
1208 failed_mount_group_desc:
1209     kfree(sbi->s_group_desc);
1210     kfree(sbi->s_debts);
1211 failed_mount:
1212     brelse(bh);
1213 failed_sbi:
1214     fs_put_dax(sbi->s_daxdev, NULL);
1215     sb->s_fs_info = NULL;
1216     kfree(sbi->s_blockgroup_lock);
1217     kfree(sbi);
1218     return ret;
1219 }
1220 
1221 static void ext2_clear_super_error(struct super_block *sb)
1222 {
1223     struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
1224 
1225     if (buffer_write_io_error(sbh)) {
1226         /*
1227          * Oh, dear.  A previous attempt to write the
1228          * superblock failed.  This could happen because the
1229          * USB device was yanked out.  Or it could happen to
1230          * be a transient write error and maybe the block will
1231          * be remapped.  Nothing we can do but to retry the
1232          * write and hope for the best.
1233          */
1234         ext2_msg(sb, KERN_ERR,
1235                "previous I/O error to superblock detected");
1236         clear_buffer_write_io_error(sbh);
1237         set_buffer_uptodate(sbh);
1238     }
1239 }
1240 
1241 void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
1242              int wait)
1243 {
1244     ext2_clear_super_error(sb);
1245     spin_lock(&EXT2_SB(sb)->s_lock);
1246     es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
1247     es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
1248     es->s_wtime = cpu_to_le32(ktime_get_real_seconds());
1249     /* unlock before we do IO */
1250     spin_unlock(&EXT2_SB(sb)->s_lock);
1251     mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
1252     if (wait)
1253         sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
1254 }
1255 
1256 /*
1257  * In the second extended file system, it is not necessary to
1258  * write the super block since we use a mapping of the
1259  * disk super block in a buffer.
1260  *
1261  * However, this function is still used to set the fs valid
1262  * flags to 0.  We need to set this flag to 0 since the fs
1263  * may have been checked while mounted and e2fsck may have
1264  * set s_state to EXT2_VALID_FS after some corrections.
1265  */
1266 static int ext2_sync_fs(struct super_block *sb, int wait)
1267 {
1268     struct ext2_sb_info *sbi = EXT2_SB(sb);
1269     struct ext2_super_block *es = EXT2_SB(sb)->s_es;
1270 
1271     /*
1272      * Write quota structures to quota file, sync_blockdev() will write
1273      * them to disk later
1274      */
1275     dquot_writeback_dquots(sb, -1);
1276 
1277     spin_lock(&sbi->s_lock);
1278     if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
1279         ext2_debug("setting valid to 0\n");
1280         es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
1281     }
1282     spin_unlock(&sbi->s_lock);
1283     ext2_sync_super(sb, es, wait);
1284     return 0;
1285 }
1286 
1287 static int ext2_freeze(struct super_block *sb)
1288 {
1289     struct ext2_sb_info *sbi = EXT2_SB(sb);
1290 
1291     /*
1292      * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
1293      * because we have unattached inodes and thus filesystem is not fully
1294      * consistent.
1295      */
1296     if (atomic_long_read(&sb->s_remove_count)) {
1297         ext2_sync_fs(sb, 1);
1298         return 0;
1299     }
1300     /* Set EXT2_FS_VALID flag */
1301     spin_lock(&sbi->s_lock);
1302     sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
1303     spin_unlock(&sbi->s_lock);
1304     ext2_sync_super(sb, sbi->s_es, 1);
1305 
1306     return 0;
1307 }
1308 
1309 static int ext2_unfreeze(struct super_block *sb)
1310 {
1311     /* Just write sb to clear EXT2_VALID_FS flag */
1312     ext2_write_super(sb);
1313 
1314     return 0;
1315 }
1316 
1317 static void ext2_write_super(struct super_block *sb)
1318 {
1319     if (!sb_rdonly(sb))
1320         ext2_sync_fs(sb, 1);
1321 }
1322 
1323 static int ext2_remount (struct super_block * sb, int * flags, char * data)
1324 {
1325     struct ext2_sb_info * sbi = EXT2_SB(sb);
1326     struct ext2_super_block * es;
1327     struct ext2_mount_options new_opts;
1328     int err;
1329 
1330     sync_filesystem(sb);
1331 
1332     spin_lock(&sbi->s_lock);
1333     new_opts.s_mount_opt = sbi->s_mount_opt;
1334     new_opts.s_resuid = sbi->s_resuid;
1335     new_opts.s_resgid = sbi->s_resgid;
1336     spin_unlock(&sbi->s_lock);
1337 
1338     if (!parse_options(data, sb, &new_opts))
1339         return -EINVAL;
1340 
1341     spin_lock(&sbi->s_lock);
1342     es = sbi->s_es;
1343     if ((sbi->s_mount_opt ^ new_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
1344         ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
1345              "dax flag with busy inodes while remounting");
1346         new_opts.s_mount_opt ^= EXT2_MOUNT_DAX;
1347     }
1348     if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1349         goto out_set;
1350     if (*flags & SB_RDONLY) {
1351         if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
1352             !(sbi->s_mount_state & EXT2_VALID_FS))
1353             goto out_set;
1354 
1355         /*
1356          * OK, we are remounting a valid rw partition rdonly, so set
1357          * the rdonly flag and then mark the partition as valid again.
1358          */
1359         es->s_state = cpu_to_le16(sbi->s_mount_state);
1360         es->s_mtime = cpu_to_le32(ktime_get_real_seconds());
1361         spin_unlock(&sbi->s_lock);
1362 
1363         err = dquot_suspend(sb, -1);
1364         if (err < 0)
1365             return err;
1366 
1367         ext2_sync_super(sb, es, 1);
1368     } else {
1369         __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
1370                            ~EXT2_FEATURE_RO_COMPAT_SUPP);
1371         if (ret) {
1372             spin_unlock(&sbi->s_lock);
1373             ext2_msg(sb, KERN_WARNING,
1374                 "warning: couldn't remount RDWR because of "
1375                 "unsupported optional features (%x).",
1376                 le32_to_cpu(ret));
1377             return -EROFS;
1378         }
1379         /*
1380          * Mounting a RDONLY partition read-write, so reread and
1381          * store the current valid flag.  (It may have been changed
1382          * by e2fsck since we originally mounted the partition.)
1383          */
1384         sbi->s_mount_state = le16_to_cpu(es->s_state);
1385         if (!ext2_setup_super (sb, es, 0))
1386             sb->s_flags &= ~SB_RDONLY;
1387         spin_unlock(&sbi->s_lock);
1388 
1389         ext2_write_super(sb);
1390 
1391         dquot_resume(sb, -1);
1392     }
1393 
1394     spin_lock(&sbi->s_lock);
1395 out_set:
1396     sbi->s_mount_opt = new_opts.s_mount_opt;
1397     sbi->s_resuid = new_opts.s_resuid;
1398     sbi->s_resgid = new_opts.s_resgid;
1399     sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
1400         (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
1401     spin_unlock(&sbi->s_lock);
1402 
1403     return 0;
1404 }
1405 
1406 static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
1407 {
1408     struct super_block *sb = dentry->d_sb;
1409     struct ext2_sb_info *sbi = EXT2_SB(sb);
1410     struct ext2_super_block *es = sbi->s_es;
1411 
1412     spin_lock(&sbi->s_lock);
1413 
1414     if (test_opt (sb, MINIX_DF))
1415         sbi->s_overhead_last = 0;
1416     else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
1417         unsigned long i, overhead = 0;
1418         smp_rmb();
1419 
1420         /*
1421          * Compute the overhead (FS structures). This is constant
1422          * for a given filesystem unless the number of block groups
1423          * changes so we cache the previous value until it does.
1424          */
1425 
1426         /*
1427          * All of the blocks before first_data_block are
1428          * overhead
1429          */
1430         overhead = le32_to_cpu(es->s_first_data_block);
1431 
1432         /*
1433          * Add the overhead attributed to the superblock and
1434          * block group descriptors.  If the sparse superblocks
1435          * feature is turned on, then not all groups have this.
1436          */
1437         for (i = 0; i < sbi->s_groups_count; i++)
1438             overhead += ext2_bg_has_super(sb, i) +
1439                 ext2_bg_num_gdb(sb, i);
1440 
1441         /*
1442          * Every block group has an inode bitmap, a block
1443          * bitmap, and an inode table.
1444          */
1445         overhead += (sbi->s_groups_count *
1446                  (2 + sbi->s_itb_per_group));
1447         sbi->s_overhead_last = overhead;
1448         smp_wmb();
1449         sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
1450     }
1451 
1452     buf->f_type = EXT2_SUPER_MAGIC;
1453     buf->f_bsize = sb->s_blocksize;
1454     buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
1455     buf->f_bfree = ext2_count_free_blocks(sb);
1456     es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
1457     buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
1458     if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
1459         buf->f_bavail = 0;
1460     buf->f_files = le32_to_cpu(es->s_inodes_count);
1461     buf->f_ffree = ext2_count_free_inodes(sb);
1462     es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
1463     buf->f_namelen = EXT2_NAME_LEN;
1464     buf->f_fsid = uuid_to_fsid(es->s_uuid);
1465     spin_unlock(&sbi->s_lock);
1466     return 0;
1467 }
1468 
1469 static struct dentry *ext2_mount(struct file_system_type *fs_type,
1470     int flags, const char *dev_name, void *data)
1471 {
1472     return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
1473 }
1474 
1475 #ifdef CONFIG_QUOTA
1476 
1477 /* Read data from quotafile - avoid pagecache and such because we cannot afford
1478  * acquiring the locks... As quota files are never truncated and quota code
1479  * itself serializes the operations (and no one else should touch the files)
1480  * we don't have to be afraid of races */
1481 static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
1482                    size_t len, loff_t off)
1483 {
1484     struct inode *inode = sb_dqopt(sb)->files[type];
1485     sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1486     int err = 0;
1487     int offset = off & (sb->s_blocksize - 1);
1488     int tocopy;
1489     size_t toread;
1490     struct buffer_head tmp_bh;
1491     struct buffer_head *bh;
1492     loff_t i_size = i_size_read(inode);
1493 
1494     if (off > i_size)
1495         return 0;
1496     if (off+len > i_size)
1497         len = i_size-off;
1498     toread = len;
1499     while (toread > 0) {
1500         tocopy = min_t(size_t, sb->s_blocksize - offset, toread);
1501 
1502         tmp_bh.b_state = 0;
1503         tmp_bh.b_size = sb->s_blocksize;
1504         err = ext2_get_block(inode, blk, &tmp_bh, 0);
1505         if (err < 0)
1506             return err;
1507         if (!buffer_mapped(&tmp_bh))    /* A hole? */
1508             memset(data, 0, tocopy);
1509         else {
1510             bh = sb_bread(sb, tmp_bh.b_blocknr);
1511             if (!bh)
1512                 return -EIO;
1513             memcpy(data, bh->b_data+offset, tocopy);
1514             brelse(bh);
1515         }
1516         offset = 0;
1517         toread -= tocopy;
1518         data += tocopy;
1519         blk++;
1520     }
1521     return len;
1522 }
1523 
1524 /* Write to quotafile */
1525 static ssize_t ext2_quota_write(struct super_block *sb, int type,
1526                 const char *data, size_t len, loff_t off)
1527 {
1528     struct inode *inode = sb_dqopt(sb)->files[type];
1529     sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
1530     int err = 0;
1531     int offset = off & (sb->s_blocksize - 1);
1532     int tocopy;
1533     size_t towrite = len;
1534     struct buffer_head tmp_bh;
1535     struct buffer_head *bh;
1536 
1537     while (towrite > 0) {
1538         tocopy = min_t(size_t, sb->s_blocksize - offset, towrite);
1539 
1540         tmp_bh.b_state = 0;
1541         tmp_bh.b_size = sb->s_blocksize;
1542         err = ext2_get_block(inode, blk, &tmp_bh, 1);
1543         if (err < 0)
1544             goto out;
1545         if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
1546             bh = sb_bread(sb, tmp_bh.b_blocknr);
1547         else
1548             bh = sb_getblk(sb, tmp_bh.b_blocknr);
1549         if (unlikely(!bh)) {
1550             err = -EIO;
1551             goto out;
1552         }
1553         lock_buffer(bh);
1554         memcpy(bh->b_data+offset, data, tocopy);
1555         flush_dcache_page(bh->b_page);
1556         set_buffer_uptodate(bh);
1557         mark_buffer_dirty(bh);
1558         unlock_buffer(bh);
1559         brelse(bh);
1560         offset = 0;
1561         towrite -= tocopy;
1562         data += tocopy;
1563         blk++;
1564     }
1565 out:
1566     if (len == towrite)
1567         return err;
1568     if (inode->i_size < off+len-towrite)
1569         i_size_write(inode, off+len-towrite);
1570     inode_inc_iversion(inode);
1571     inode->i_mtime = inode->i_ctime = current_time(inode);
1572     mark_inode_dirty(inode);
1573     return len - towrite;
1574 }
1575 
1576 static int ext2_quota_on(struct super_block *sb, int type, int format_id,
1577              const struct path *path)
1578 {
1579     int err;
1580     struct inode *inode;
1581 
1582     err = dquot_quota_on(sb, type, format_id, path);
1583     if (err)
1584         return err;
1585 
1586     inode = d_inode(path->dentry);
1587     inode_lock(inode);
1588     EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
1589     inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
1590             S_NOATIME | S_IMMUTABLE);
1591     inode_unlock(inode);
1592     mark_inode_dirty(inode);
1593 
1594     return 0;
1595 }
1596 
1597 static int ext2_quota_off(struct super_block *sb, int type)
1598 {
1599     struct inode *inode = sb_dqopt(sb)->files[type];
1600     int err;
1601 
1602     if (!inode || !igrab(inode))
1603         goto out;
1604 
1605     err = dquot_quota_off(sb, type);
1606     if (err)
1607         goto out_put;
1608 
1609     inode_lock(inode);
1610     EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
1611     inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
1612     inode_unlock(inode);
1613     mark_inode_dirty(inode);
1614 out_put:
1615     iput(inode);
1616     return err;
1617 out:
1618     return dquot_quota_off(sb, type);
1619 }
1620 
1621 #endif
1622 
1623 static struct file_system_type ext2_fs_type = {
1624     .owner      = THIS_MODULE,
1625     .name       = "ext2",
1626     .mount      = ext2_mount,
1627     .kill_sb    = kill_block_super,
1628     .fs_flags   = FS_REQUIRES_DEV,
1629 };
1630 MODULE_ALIAS_FS("ext2");
1631 
1632 static int __init init_ext2_fs(void)
1633 {
1634     int err;
1635 
1636     err = init_inodecache();
1637     if (err)
1638         return err;
1639         err = register_filesystem(&ext2_fs_type);
1640     if (err)
1641         goto out;
1642     return 0;
1643 out:
1644     destroy_inodecache();
1645     return err;
1646 }
1647 
1648 static void __exit exit_ext2_fs(void)
1649 {
1650     unregister_filesystem(&ext2_fs_type);
1651     destroy_inodecache();
1652 }
1653 
1654 MODULE_AUTHOR("Remy Card and others");
1655 MODULE_DESCRIPTION("Second Extended Filesystem");
1656 MODULE_LICENSE("GPL");
1657 module_init(init_ext2_fs)
1658 module_exit(exit_ext2_fs)