]> git.karo-electronics.de Git - mv-sheeva.git/blobdiff - fs/ext4/super.c
Merge git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild
[mv-sheeva.git] / fs / ext4 / super.c
index 64067de70c6f34683cf751c04a168e06783dc6f6..055a0cd0168e5029ee87bb3ce8c55153fae3b171 100644 (file)
@@ -503,6 +503,7 @@ static void ext4_put_super (struct super_block * sb)
        struct ext4_super_block *es = sbi->s_es;
        int i;
 
+       ext4_mb_release(sb);
        ext4_ext_release(sb);
        ext4_xattr_put_super(sb);
        jbd2_journal_destroy(sbi->s_journal);
@@ -569,6 +570,8 @@ static struct inode *ext4_alloc_inode(struct super_block *sb)
        ei->i_block_alloc_info = NULL;
        ei->vfs_inode.i_version = 1;
        memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
+       INIT_LIST_HEAD(&ei->i_prealloc_list);
+       spin_lock_init(&ei->i_prealloc_lock);
        return &ei->vfs_inode;
 }
 
@@ -593,7 +596,7 @@ static void init_once(struct kmem_cache *cachep, void *foo)
 #ifdef CONFIG_EXT4DEV_FS_XATTR
        init_rwsem(&ei->xattr_sem);
 #endif
-       mutex_init(&ei->truncate_mutex);
+       init_rwsem(&ei->i_data_sem);
        inode_init_once(&ei->vfs_inode);
 }
 
@@ -665,18 +668,20 @@ static inline void ext4_show_quota_options(struct seq_file *seq, struct super_bl
  */
 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
 {
+       int def_errors;
+       unsigned long def_mount_opts;
        struct super_block *sb = vfs->mnt_sb;
        struct ext4_sb_info *sbi = EXT4_SB(sb);
        struct ext4_super_block *es = sbi->s_es;
-       unsigned long def_mount_opts;
 
        def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
+       def_errors     = le16_to_cpu(es->s_errors);
 
        if (sbi->s_sb_block != 1)
                seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
        if (test_opt(sb, MINIX_DF))
                seq_puts(seq, ",minixdf");
-       if (test_opt(sb, GRPID))
+       if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
                seq_puts(seq, ",grpid");
        if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
                seq_puts(seq, ",nogrpid");
@@ -688,34 +693,33 @@ static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
            le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
                seq_printf(seq, ",resgid=%u", sbi->s_resgid);
        }
-       if (test_opt(sb, ERRORS_CONT)) {
-               int def_errors = le16_to_cpu(es->s_errors);
-
+       if (test_opt(sb, ERRORS_RO)) {
                if (def_errors == EXT4_ERRORS_PANIC ||
-                   def_errors == EXT4_ERRORS_RO) {
-                       seq_puts(seq, ",errors=continue");
+                   def_errors == EXT4_ERRORS_CONTINUE) {
+                       seq_puts(seq, ",errors=remount-ro");
                }
        }
-       if (test_opt(sb, ERRORS_RO))
-               seq_puts(seq, ",errors=remount-ro");
-       if (test_opt(sb, ERRORS_PANIC))
+       if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
+               seq_puts(seq, ",errors=continue");
+       if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
                seq_puts(seq, ",errors=panic");
-       if (test_opt(sb, NO_UID32))
+       if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
                seq_puts(seq, ",nouid32");
-       if (test_opt(sb, DEBUG))
+       if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
                seq_puts(seq, ",debug");
        if (test_opt(sb, OLDALLOC))
                seq_puts(seq, ",oldalloc");
-#ifdef CONFIG_EXT4_FS_XATTR
-       if (test_opt(sb, XATTR_USER))
+#ifdef CONFIG_EXT4DEV_FS_XATTR
+       if (test_opt(sb, XATTR_USER) &&
+               !(def_mount_opts & EXT4_DEFM_XATTR_USER))
                seq_puts(seq, ",user_xattr");
        if (!test_opt(sb, XATTR_USER) &&
            (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
                seq_puts(seq, ",nouser_xattr");
        }
 #endif
-#ifdef CONFIG_EXT4_FS_POSIX_ACL
-       if (test_opt(sb, POSIX_ACL))
+#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
+       if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
                seq_puts(seq, ",acl");
        if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
                seq_puts(seq, ",noacl");
@@ -732,7 +736,17 @@ static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
                seq_puts(seq, ",nobh");
        if (!test_opt(sb, EXTENTS))
                seq_puts(seq, ",noextents");
+       if (!test_opt(sb, MBALLOC))
+               seq_puts(seq, ",nomballoc");
+       if (test_opt(sb, I_VERSION))
+               seq_puts(seq, ",i_version");
 
+       if (sbi->s_stripe)
+               seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
+       /*
+        * journal mode get enabled in different ways
+        * So just print the value even if we didn't specify it
+        */
        if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
                seq_puts(seq, ",data=journal");
        else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
@@ -741,7 +755,6 @@ static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
                seq_puts(seq, ",data=writeback");
 
        ext4_show_quota_options(seq, sb);
-
        return 0;
 }
 
@@ -869,11 +882,13 @@ enum {
        Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
        Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
        Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
+       Opt_journal_checksum, Opt_journal_async_commit,
        Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
        Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
        Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
        Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
-       Opt_grpquota, Opt_extents, Opt_noextents,
+       Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
+       Opt_mballoc, Opt_nomballoc, Opt_stripe,
 };
 
 static match_table_t tokens = {
@@ -908,6 +923,8 @@ static match_table_t tokens = {
        {Opt_journal_update, "journal=update"},
        {Opt_journal_inum, "journal=%u"},
        {Opt_journal_dev, "journal_dev=%u"},
+       {Opt_journal_checksum, "journal_checksum"},
+       {Opt_journal_async_commit, "journal_async_commit"},
        {Opt_abort, "abort"},
        {Opt_data_journal, "data=journal"},
        {Opt_data_ordered, "data=ordered"},
@@ -925,6 +942,10 @@ static match_table_t tokens = {
        {Opt_barrier, "barrier=%u"},
        {Opt_extents, "extents"},
        {Opt_noextents, "noextents"},
+       {Opt_i_version, "i_version"},
+       {Opt_mballoc, "mballoc"},
+       {Opt_nomballoc, "nomballoc"},
+       {Opt_stripe, "stripe=%u"},
        {Opt_err, NULL},
        {Opt_resize, "resize"},
 };
@@ -1095,6 +1116,13 @@ static int parse_options (char *options, struct super_block *sb,
                                return 0;
                        *journal_devnum = option;
                        break;
+               case Opt_journal_checksum:
+                       set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
+                       break;
+               case Opt_journal_async_commit:
+                       set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
+                       set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
+                       break;
                case Opt_noload:
                        set_opt (sbi->s_mount_opt, NOLOAD);
                        break;
@@ -1263,6 +1291,23 @@ clear_qf_name:
                case Opt_noextents:
                        clear_opt (sbi->s_mount_opt, EXTENTS);
                        break;
+               case Opt_i_version:
+                       set_opt(sbi->s_mount_opt, I_VERSION);
+                       sb->s_flags |= MS_I_VERSION;
+                       break;
+               case Opt_mballoc:
+                       set_opt(sbi->s_mount_opt, MBALLOC);
+                       break;
+               case Opt_nomballoc:
+                       clear_opt(sbi->s_mount_opt, MBALLOC);
+                       break;
+               case Opt_stripe:
+                       if (match_int(&args[0], &option))
+                               return 0;
+                       if (option < 0)
+                               return 0;
+                       sbi->s_stripe = option;
+                       break;
                default:
                        printk (KERN_ERR
                                "EXT4-fs: Unrecognized mount option \"%s\" "
@@ -1600,19 +1645,58 @@ static void ext4_orphan_cleanup (struct super_block * sb,
 #endif
        sb->s_flags = s_flags; /* Restore MS_RDONLY status */
 }
+/*
+ * Maximal extent format file size.
+ * Resulting logical blkno at s_maxbytes must fit in our on-disk
+ * extent format containers, within a sector_t, and within i_blocks
+ * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
+ * so that won't be a limiting factor.
+ *
+ * Note, this does *not* consider any metadata overhead for vfs i_blocks.
+ */
+static loff_t ext4_max_size(int blkbits)
+{
+       loff_t res;
+       loff_t upper_limit = MAX_LFS_FILESIZE;
+
+       /* small i_blocks in vfs inode? */
+       if (sizeof(blkcnt_t) < sizeof(u64)) {
+               /*
+                * CONFIG_LSF is not enabled implies the inode
+                * i_block represent total blocks in 512 bytes
+                * 32 == size of vfs inode i_blocks * 8
+                */
+               upper_limit = (1LL << 32) - 1;
+
+               /* total blocks in file system block size */
+               upper_limit >>= (blkbits - 9);
+               upper_limit <<= blkbits;
+       }
+
+       /* 32-bit extent-start container, ee_block */
+       res = 1LL << 32;
+       res <<= blkbits;
+       res -= 1;
+
+       /* Sanity check against vm- & vfs- imposed limits */
+       if (res > upper_limit)
+               res = upper_limit;
+
+       return res;
+}
 
 /*
- * Maximal file size.  There is a direct, and {,double-,triple-}indirect
+ * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
  * We need to be 1 filesystem block less than the 2^48 sector limit.
  */
-static loff_t ext4_max_size(int bits)
+static loff_t ext4_max_bitmap_size(int bits)
 {
        loff_t res = EXT4_NDIR_BLOCKS;
        int meta_blocks;
        loff_t upper_limit;
        /* This is calculated to be the largest file size for a
-        * dense, file such that the total number of
+        * dense, bitmapped file such that the total number of
         * sectors in the file, including data and all indirect blocks,
         * does not exceed 2^48 -1
         * __u32 i_blocks_lo and _u16 i_blocks_high representing the
@@ -1682,6 +1766,34 @@ static ext4_fsblk_t descriptor_loc(struct super_block *sb,
        return (has_super + ext4_group_first_block_no(sb, bg));
 }
 
+/**
+ * ext4_get_stripe_size: Get the stripe size.
+ * @sbi: In memory super block info
+ *
+ * If we have specified it via mount option, then
+ * use the mount option value. If the value specified at mount time is
+ * greater than the blocks per group use the super block value.
+ * If the super block value is greater than blocks per group return 0.
+ * Allocator needs it be less than blocks per group.
+ *
+ */
+static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
+{
+       unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
+       unsigned long stripe_width =
+                       le32_to_cpu(sbi->s_es->s_raid_stripe_width);
+
+       if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
+               return sbi->s_stripe;
+
+       if (stripe_width <= sbi->s_blocks_per_group)
+               return stripe_width;
+
+       if (stride <= sbi->s_blocks_per_group)
+               return stride;
+
+       return 0;
+}
 
 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
                                __releases(kernel_sem)
@@ -1700,7 +1812,6 @@ static int ext4_fill_super (struct super_block *sb, void *data, int silent)
        unsigned long def_mount_opts;
        struct inode *root;
        int blocksize;
-       int hblock;
        int db_count;
        int i;
        int needs_recovery;
@@ -1780,10 +1891,10 @@ static int ext4_fill_super (struct super_block *sb, void *data, int silent)
 
        if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
                set_opt(sbi->s_mount_opt, ERRORS_PANIC);
-       else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_RO)
-               set_opt(sbi->s_mount_opt, ERRORS_RO);
-       else
+       else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
                set_opt(sbi->s_mount_opt, ERRORS_CONT);
+       else
+               set_opt(sbi->s_mount_opt, ERRORS_RO);
 
        sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
        sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
@@ -1795,6 +1906,11 @@ static int ext4_fill_super (struct super_block *sb, void *data, int silent)
         * User -o noextents to turn it off
         */
        set_opt(sbi->s_mount_opt, EXTENTS);
+       /*
+        * turn on mballoc feature by default in ext4 filesystem
+        * User -o nomballoc to turn it off
+        */
+       set_opt(sbi->s_mount_opt, MBALLOC);
 
        if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
                            NULL, 0))
@@ -1852,20 +1968,16 @@ static int ext4_fill_super (struct super_block *sb, void *data, int silent)
                goto failed_mount;
        }
 
-       hblock = bdev_hardsect_size(sb->s_bdev);
        if (sb->s_blocksize != blocksize) {
-               /*
-                * Make sure the blocksize for the filesystem is larger
-                * than the hardware sectorsize for the machine.
-                */
-               if (blocksize < hblock) {
-                       printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
-                              "device blocksize %d.\n", blocksize, hblock);
+
+               /* Validate the filesystem blocksize */
+               if (!sb_set_blocksize(sb, blocksize)) {
+                       printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
+                                       blocksize);
                        goto failed_mount;
                }
 
                brelse (bh);
-               sb_set_blocksize(sb, blocksize);
                logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
                offset = do_div(logical_sb_block, blocksize);
                bh = sb_bread(sb, logical_sb_block);
@@ -1883,6 +1995,7 @@ static int ext4_fill_super (struct super_block *sb, void *data, int silent)
                }
        }
 
+       sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
        sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
 
        if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
@@ -1957,6 +2070,17 @@ static int ext4_fill_super (struct super_block *sb, void *data, int silent)
 
        if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
                goto cantfind_ext4;
+
+       /* ensure blocks_count calculation below doesn't sign-extend */
+       if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
+           le32_to_cpu(es->s_first_data_block) + 1) {
+               printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
+                      "first data block %u, blocks per group %lu\n",
+                       ext4_blocks_count(es),
+                       le32_to_cpu(es->s_first_data_block),
+                       EXT4_BLOCKS_PER_GROUP(sb));
+               goto failed_mount;
+       }
        blocks_count = (ext4_blocks_count(es) -
                        le32_to_cpu(es->s_first_data_block) +
                        EXT4_BLOCKS_PER_GROUP(sb) - 1);
@@ -2019,6 +2143,8 @@ static int ext4_fill_super (struct super_block *sb, void *data, int silent)
        sbi->s_rsv_window_head.rsv_goal_size = 0;
        ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
 
+       sbi->s_stripe = ext4_get_stripe_size(sbi);
+
        /*
         * set up enough so that it can read an inode
         */
@@ -2063,6 +2189,21 @@ static int ext4_fill_super (struct super_block *sb, void *data, int silent)
                goto failed_mount4;
        }
 
+       if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
+               jbd2_journal_set_features(sbi->s_journal,
+                               JBD2_FEATURE_COMPAT_CHECKSUM, 0,
+                               JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
+       } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
+               jbd2_journal_set_features(sbi->s_journal,
+                               JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
+               jbd2_journal_clear_features(sbi->s_journal, 0, 0,
+                               JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
+       } else {
+               jbd2_journal_clear_features(sbi->s_journal,
+                               JBD2_FEATURE_COMPAT_CHECKSUM, 0,
+                               JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
+       }
+
        /* We have now updated the journal if required, so we can
         * validate the data journaling mode. */
        switch (test_opt(sb, DATA_FLAGS)) {
@@ -2163,6 +2304,7 @@ static int ext4_fill_super (struct super_block *sb, void *data, int silent)
                "writeback");
 
        ext4_ext_init(sb);
+       ext4_mb_init(sb, needs_recovery);
 
        lock_kernel();
        return 0;
@@ -3121,7 +3263,6 @@ out:
                i_size_write(inode, off+len-towrite);
                EXT4_I(inode)->i_disksize = inode->i_size;
        }
-       inode->i_version++;
        inode->i_mtime = inode->i_ctime = CURRENT_TIME;
        ext4_mark_inode_dirty(handle, inode);
        mutex_unlock(&inode->i_mutex);
@@ -3146,9 +3287,15 @@ static struct file_system_type ext4dev_fs_type = {
 
 static int __init init_ext4_fs(void)
 {
-       int err = init_ext4_xattr();
+       int err;
+
+       err = init_ext4_mballoc();
        if (err)
                return err;
+
+       err = init_ext4_xattr();
+       if (err)
+               goto out2;
        err = init_inodecache();
        if (err)
                goto out1;
@@ -3160,6 +3307,8 @@ out:
        destroy_inodecache();
 out1:
        exit_ext4_xattr();
+out2:
+       exit_ext4_mballoc();
        return err;
 }
 
@@ -3168,6 +3317,7 @@ static void __exit exit_ext4_fs(void)
        unregister_filesystem(&ext4dev_fs_type);
        destroy_inodecache();
        exit_ext4_xattr();
+       exit_ext4_mballoc();
 }
 
 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");