tunnels: fix netns vs proto registration ordering
[safe/jmp/linux-2.6] / fs / ocfs2 / ocfs2_fs.h
index 19ac421..1a1a679 100644 (file)
 #define OCFS2_INODE_SIGNATURE          "INODE01"
 #define OCFS2_EXTENT_BLOCK_SIGNATURE   "EXBLK01"
 #define OCFS2_GROUP_DESC_SIGNATURE      "GROUP01"
+#define OCFS2_XATTR_BLOCK_SIGNATURE    "XATTR01"
+#define OCFS2_DIR_TRAILER_SIGNATURE    "DIRTRL1"
+#define OCFS2_DX_ROOT_SIGNATURE                "DXDIR01"
+#define OCFS2_DX_LEAF_SIGNATURE                "DXLEAF1"
+#define OCFS2_REFCOUNT_BLOCK_SIGNATURE "REFCNT1"
 
 /* Compatibility flags */
 #define OCFS2_HAS_COMPAT_FEATURE(sb,mask)                      \
 #define OCFS2_CLEAR_INCOMPAT_FEATURE(sb,mask)                  \
        OCFS2_SB(sb)->s_feature_incompat &= ~(mask)
 
-#define OCFS2_FEATURE_COMPAT_SUPP      OCFS2_FEATURE_COMPAT_BACKUP_SB
+#define OCFS2_FEATURE_COMPAT_SUPP      (OCFS2_FEATURE_COMPAT_BACKUP_SB \
+                                        | OCFS2_FEATURE_COMPAT_JBD2_SB)
 #define OCFS2_FEATURE_INCOMPAT_SUPP    (OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT \
                                         | OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC \
-                                        | OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
-#define OCFS2_FEATURE_RO_COMPAT_SUPP   OCFS2_FEATURE_RO_COMPAT_UNWRITTEN
+                                        | OCFS2_FEATURE_INCOMPAT_INLINE_DATA \
+                                        | OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP \
+                                        | OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK \
+                                        | OCFS2_FEATURE_INCOMPAT_XATTR \
+                                        | OCFS2_FEATURE_INCOMPAT_META_ECC \
+                                        | OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS \
+                                        | OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE)
+#define OCFS2_FEATURE_RO_COMPAT_SUPP   (OCFS2_FEATURE_RO_COMPAT_UNWRITTEN \
+                                        | OCFS2_FEATURE_RO_COMPAT_USRQUOTA \
+                                        | OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 
 /*
  * Heartbeat-only devices are missing journals and other files.  The
 #define OCFS2_FEATURE_INCOMPAT_INLINE_DATA     0x0040
 
 /*
+ * Support for alternate, userspace cluster stacks.  If set, the superblock
+ * field s_cluster_info contains a tag for the alternate stack in use as
+ * well as the name of the cluster being joined.
+ * mount.ocfs2 must pass in a matching stack name.
+ *
+ * If not set, the classic stack will be used.  This is compatbile with
+ * all older versions.
+ */
+#define OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK 0x0080
+
+/* Support for the extended slot map */
+#define OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP 0x100
+
+/* Support for extended attributes */
+#define OCFS2_FEATURE_INCOMPAT_XATTR           0x0200
+
+/* Support for indexed directores */
+#define OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS    0x0400
+
+/* Metadata checksum and error correction */
+#define OCFS2_FEATURE_INCOMPAT_META_ECC                0x0800
+
+/* Refcount tree support */
+#define OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE   0x1000
+
+/*
  * backup superblock flag is used to indicate that this volume
  * has backup superblocks.
  */
 #define OCFS2_FEATURE_COMPAT_BACKUP_SB         0x0001
 
 /*
+ * The filesystem will correctly handle journal feature bits.
+ */
+#define OCFS2_FEATURE_COMPAT_JBD2_SB           0x0002
+
+/*
  * Unwritten extents support.
  */
 #define OCFS2_FEATURE_RO_COMPAT_UNWRITTEN      0x0001
 
+/*
+ * Maintain quota information for this filesystem
+ */
+#define OCFS2_FEATURE_RO_COMPAT_USRQUOTA       0x0002
+#define OCFS2_FEATURE_RO_COMPAT_GRPQUOTA       0x0004
+
 /* The byte offset of the first backup block will be 1G.
  * The following will be 4G, 16G, 64G, 256G and 1T.
  */
 #define OCFS2_HEARTBEAT_FL     (0x00000200)    /* Heartbeat area */
 #define OCFS2_CHAIN_FL         (0x00000400)    /* Chain allocator */
 #define OCFS2_DEALLOC_FL       (0x00000800)    /* Truncate log */
+#define OCFS2_QUOTA_FL         (0x00001000)    /* Quota file */
 
 /*
  * Flags on ocfs2_dinode.i_dyn_features
 #define OCFS2_HAS_XATTR_FL     (0x0002)
 #define OCFS2_INLINE_XATTR_FL  (0x0004)
 #define OCFS2_INDEXED_DIR_FL   (0x0008)
+#define OCFS2_HAS_REFCOUNT_FL   (0x0010)
 
 /* Inode attributes, keep in sync with EXT2 */
 #define OCFS2_SECRM_FL         (0x00000001)    /* Secure deletion */
 /*
  * Extent record flags (e_node.leaf.flags)
  */
-#define OCFS2_EXT_UNWRITTEN    (0x01)  /* Extent is allocated but
-                                        * unwritten */
+#define OCFS2_EXT_UNWRITTEN            (0x01)  /* Extent is allocated but
+                                                * unwritten */
+#define OCFS2_EXT_REFCOUNTED           (0x02)  /* Extent is reference
+                                                * counted in an associated
+                                                * refcount tree */
 
 /*
  * ioctl commands
@@ -231,7 +287,28 @@ struct ocfs2_space_resv {
 #define OCFS2_IOC_RESVSP64     _IOW ('X', 42, struct ocfs2_space_resv)
 #define OCFS2_IOC_UNRESVSP64   _IOW ('X', 43, struct ocfs2_space_resv)
 
+/* Used to pass group descriptor data when online resize is done */
+struct ocfs2_new_group_input {
+       __u64 group;            /* Group descriptor's blkno. */
+       __u32 clusters;         /* Total number of clusters in this group */
+       __u32 frees;            /* Total free clusters in this group */
+       __u16 chain;            /* Chain for this group */
+       __u16 reserved1;
+       __u32 reserved2;
+};
+
 #define OCFS2_IOC_GROUP_EXTEND _IOW('o', 1, int)
+#define OCFS2_IOC_GROUP_ADD    _IOW('o', 2,struct ocfs2_new_group_input)
+#define OCFS2_IOC_GROUP_ADD64  _IOW('o', 3,struct ocfs2_new_group_input)
+
+/* Used to pass 2 file names to reflink. */
+struct reflink_arguments {
+       __u64 old_path;
+       __u64 new_path;
+       __u64 preserve;
+};
+#define OCFS2_IOC_REFLINK      _IOW('o', 4, struct reflink_arguments)
+
 
 /*
  * Journal Flags (ocfs2_dinode.id1.journal1.i_flags)
@@ -255,9 +332,27 @@ struct ocfs2_space_resv {
 #define OCFS2_VOL_UUID_LEN             16
 #define OCFS2_MAX_VOL_LABEL_LEN                64
 
+/* The alternate, userspace stack fields */
+#define OCFS2_STACK_LABEL_LEN          4
+#define OCFS2_CLUSTER_NAME_LEN         16
+
 /* Journal limits (in bytes) */
 #define OCFS2_MIN_JOURNAL_SIZE         (4 * 1024 * 1024)
 
+/*
+ * Default local alloc size (in megabytes)
+ *
+ * The value chosen should be such that most allocations, including new
+ * block groups, use local alloc.
+ */
+#define OCFS2_DEFAULT_LOCAL_ALLOC_SIZE 8
+
+/*
+ * Inline extended attribute size (in bytes)
+ * The value chosen should be aligned to 16 byte boundaries.
+ */
+#define OCFS2_MIN_XATTR_INLINE_SIZE     256
+
 struct ocfs2_system_inode_info {
        char    *si_name;
        int     si_iflags;
@@ -272,13 +367,17 @@ enum {
 #define OCFS2_FIRST_ONLINE_SYSTEM_INODE SLOT_MAP_SYSTEM_INODE
        HEARTBEAT_SYSTEM_INODE,
        GLOBAL_BITMAP_SYSTEM_INODE,
-#define OCFS2_LAST_GLOBAL_SYSTEM_INODE GLOBAL_BITMAP_SYSTEM_INODE
+       USER_QUOTA_SYSTEM_INODE,
+       GROUP_QUOTA_SYSTEM_INODE,
+#define OCFS2_LAST_GLOBAL_SYSTEM_INODE GROUP_QUOTA_SYSTEM_INODE
        ORPHAN_DIR_SYSTEM_INODE,
        EXTENT_ALLOC_SYSTEM_INODE,
        INODE_ALLOC_SYSTEM_INODE,
        JOURNAL_SYSTEM_INODE,
        LOCAL_ALLOC_SYSTEM_INODE,
        TRUNCATE_LOG_SYSTEM_INODE,
+       LOCAL_USER_QUOTA_SYSTEM_INODE,
+       LOCAL_GROUP_QUOTA_SYSTEM_INODE,
        NUM_SYSTEM_INODES
 };
 
@@ -292,6 +391,8 @@ static struct ocfs2_system_inode_info ocfs2_system_inodes[NUM_SYSTEM_INODES] = {
        [SLOT_MAP_SYSTEM_INODE]                 = { "slot_map", 0, S_IFREG | 0644 },
        [HEARTBEAT_SYSTEM_INODE]                = { "heartbeat", OCFS2_HEARTBEAT_FL, S_IFREG | 0644 },
        [GLOBAL_BITMAP_SYSTEM_INODE]            = { "global_bitmap", 0, S_IFREG | 0644 },
+       [USER_QUOTA_SYSTEM_INODE]               = { "aquota.user", OCFS2_QUOTA_FL, S_IFREG | 0644 },
+       [GROUP_QUOTA_SYSTEM_INODE]              = { "aquota.group", OCFS2_QUOTA_FL, S_IFREG | 0644 },
 
        /* Slot-specific system inodes (one copy per slot) */
        [ORPHAN_DIR_SYSTEM_INODE]               = { "orphan_dir:%04d", 0, S_IFDIR | 0755 },
@@ -299,7 +400,9 @@ static struct ocfs2_system_inode_info ocfs2_system_inodes[NUM_SYSTEM_INODES] = {
        [INODE_ALLOC_SYSTEM_INODE]              = { "inode_alloc:%04d", OCFS2_BITMAP_FL | OCFS2_CHAIN_FL, S_IFREG | 0644 },
        [JOURNAL_SYSTEM_INODE]                  = { "journal:%04d", OCFS2_JOURNAL_FL, S_IFREG | 0644 },
        [LOCAL_ALLOC_SYSTEM_INODE]              = { "local_alloc:%04d", OCFS2_BITMAP_FL | OCFS2_LOCAL_ALLOC_FL, S_IFREG | 0644 },
-       [TRUNCATE_LOG_SYSTEM_INODE]             = { "truncate_log:%04d", OCFS2_DEALLOC_FL, S_IFREG | 0644 }
+       [TRUNCATE_LOG_SYSTEM_INODE]             = { "truncate_log:%04d", OCFS2_DEALLOC_FL, S_IFREG | 0644 },
+       [LOCAL_USER_QUOTA_SYSTEM_INODE]         = { "aquota.user:%04d", OCFS2_QUOTA_FL, S_IFREG | 0644 },
+       [LOCAL_GROUP_QUOTA_SYSTEM_INODE]        = { "aquota.group:%04d", OCFS2_QUOTA_FL, S_IFREG | 0644 },
 };
 
 /* Parameter passed from mount.ocfs2 to module */
@@ -332,8 +435,12 @@ static struct ocfs2_system_inode_info ocfs2_system_inodes[NUM_SYSTEM_INODES] = {
 #define OCFS2_DIR_REC_LEN(name_len)    (((name_len) + OCFS2_DIR_MEMBER_LEN + \
                                           OCFS2_DIR_ROUND) & \
                                         ~OCFS2_DIR_ROUND)
+#define OCFS2_DIR_MIN_REC_LEN  OCFS2_DIR_REC_LEN(1)
 
 #define OCFS2_LINK_MAX         32000
+#define        OCFS2_DX_LINK_MAX       ((1U << 31) - 1U)
+#define        OCFS2_LINKS_HI_SHIFT    16
+#define        OCFS2_DX_ENTRIES_MAX    (0xffffffffU)
 
 #define S_SHIFT                        12
 static unsigned char ocfs2_type_by_mode[S_IFMT >> S_SHIFT] = {
@@ -353,6 +460,22 @@ static unsigned char ocfs2_type_by_mode[S_IFMT >> S_SHIFT] = {
 #define OCFS2_RAW_SB(dinode)           (&((dinode)->id2.i_super))
 
 /*
+ * Block checking structure.  This is used in metadata to validate the
+ * contents.  If OCFS2_FEATURE_INCOMPAT_META_ECC is not set, it is all
+ * zeros.
+ */
+struct ocfs2_block_check {
+/*00*/ __le32 bc_crc32e;       /* 802.3 Ethernet II CRC32 */
+       __le16 bc_ecc;          /* Single-error-correction parity vector.
+                                  This is a simple Hamming code dependant
+                                  on the blocksize.  OCFS2's maximum
+                                  blocksize, 4K, requires 16 parity bits,
+                                  so we fit in __le16. */
+       __le16 bc_reserved1;
+/*08*/
+};
+
+/*
  * On disk extent record for OCFS2
  * It describes a range of clusters on disk.
  *
@@ -439,7 +562,7 @@ struct ocfs2_truncate_log {
 struct ocfs2_extent_block
 {
 /*00*/ __u8 h_signature[8];            /* Signature for verification */
-       __le64 h_reserved1;
+       struct ocfs2_block_check h_check;       /* Error checking */
 /*10*/ __le16 h_suballoc_slot;         /* Slot suballocator this
                                           extent_header belongs to */
        __le16 h_suballoc_bit;          /* Bit offset in suballocator
@@ -455,6 +578,47 @@ struct ocfs2_extent_block
 };
 
 /*
+ * On disk slot map for OCFS2.  This defines the contents of the "slot_map"
+ * system file.  A slot is valid if it contains a node number >= 0.  The
+ * value -1 (0xFFFF) is OCFS2_INVALID_SLOT.  This marks a slot empty.
+ */
+struct ocfs2_slot_map {
+/*00*/ __le16 sm_slots[0];
+/*
+ * Actual on-disk size is one block.  OCFS2_MAX_SLOTS is 255,
+ * 255 * sizeof(__le16) == 512B, within the 512B block minimum blocksize.
+ */
+};
+
+struct ocfs2_extended_slot {
+/*00*/ __u8    es_valid;
+       __u8    es_reserved1[3];
+       __le32  es_node_num;
+/*10*/
+};
+
+/*
+ * The extended slot map, used when OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP
+ * is set.  It separates out the valid marker from the node number, and
+ * has room to grow.  Unlike the old slot map, this format is defined by
+ * i_size.
+ */
+struct ocfs2_slot_map_extended {
+/*00*/ struct ocfs2_extended_slot se_slots[0];
+/*
+ * Actual size is i_size of the slot_map system file.  It should
+ * match s_max_slots * sizeof(struct ocfs2_extended_slot)
+ */
+};
+
+struct ocfs2_cluster_info {
+/*00*/ __u8   ci_stack[OCFS2_STACK_LABEL_LEN];
+       __le32 ci_reserved;
+/*08*/ __u8   ci_cluster[OCFS2_CLUSTER_NAME_LEN];
+/*18*/
+};
+
+/*
  * On disk superblock for OCFS2
  * Note that it is contained inside an ocfs2_dinode, so all offsets
  * are relative to the start of ocfs2_dinode.id2.
@@ -481,12 +645,30 @@ struct ocfs2_super_block {
 /*40*/ __le16 s_max_slots;             /* Max number of simultaneous mounts
                                           before tunefs required */
        __le16 s_tunefs_flag;
-       __le32 s_reserved1;
+       __le32 s_uuid_hash;             /* hash value of uuid */
        __le64 s_first_cluster_group;   /* Block offset of 1st cluster
                                         * group header */
 /*50*/ __u8  s_label[OCFS2_MAX_VOL_LABEL_LEN]; /* Label for mounting, etc. */
 /*90*/ __u8  s_uuid[OCFS2_VOL_UUID_LEN];       /* 128-bit uuid */
-/*A0*/
+/*A0*/  struct ocfs2_cluster_info s_cluster_info; /* Selected userspace
+                                                    stack.  Only valid
+                                                    with INCOMPAT flag. */
+/*B8*/ __le16 s_xattr_inline_size;     /* extended attribute inline size
+                                          for this fs*/
+       __le16 s_reserved0;
+       __le32 s_dx_seed[3];            /* seed[0-2] for dx dir hash.
+                                        * s_uuid_hash serves as seed[3]. */
+/*C0*/  __le64 s_reserved2[15];                /* Fill out superblock */
+/*140*/
+
+       /*
+        * NOTE: As stated above, all offsets are relative to
+        * ocfs2_dinode.id2, which is at 0xC0 in the inode.
+        * 0xC0 + 0x140 = 0x200 or 512 bytes.  A superblock must fit within
+        * our smallest blocksize, which is 512 bytes.  To ensure this,
+        * we reserve the space in s_reserved2.  Anything past s_reserved2
+        * will not be available on the smallest blocksize.
+        */
 };
 
 /*
@@ -526,7 +708,8 @@ struct ocfs2_dinode {
                                           belongs to */
        __le16 i_suballoc_bit;          /* Bit offset in suballocator
                                           block group */
-/*10*/ __le32 i_reserved0;
+/*10*/ __le16 i_links_count_hi;        /* High 16 bits of links count */
+       __le16 i_xattr_inline_size;
        __le32 i_clusters;              /* Cluster count */
        __le32 i_uid;                   /* Owner UID */
        __le32 i_gid;                   /* Owning GID */
@@ -545,11 +728,15 @@ struct ocfs2_dinode {
        __le32 i_atime_nsec;
        __le32 i_ctime_nsec;
        __le32 i_mtime_nsec;
-       __le32 i_attr;
+/*70*/ __le32 i_attr;
        __le16 i_orphaned_slot;         /* Only valid when OCFS2_ORPHANED_FL
                                           was set in i_flags */
        __le16 i_dyn_features;
-/*70*/ __le64 i_reserved2[8];
+       __le64 i_xattr_loc;
+/*80*/ struct ocfs2_block_check i_check;       /* Error checking */
+/*88*/ __le64 i_dx_root;               /* Pointer to dir index root block */
+/*90*/ __le64 i_refcount_loc;
+       __le64 i_reserved2[4];
 /*B8*/ union {
                __le64 i_pad1;          /* Generic way to refer to this
                                           64bit union */
@@ -565,7 +752,10 @@ struct ocfs2_dinode {
                struct {                /* Info for journal system
                                           inodes */
                        __le32 ij_flags;        /* Mounted, version, etc. */
-                       __le32 ij_pad;
+                       __le32 ij_recovery_generation; /* Incremented when the
+                                                         journal is recovered
+                                                         after an unclean
+                                                         shutdown */
                } journal1;
        } id1;                          /* Inode type dependant 1 */
 /*C0*/ union {
@@ -595,6 +785,118 @@ struct ocfs2_dir_entry {
 } __attribute__ ((packed));
 
 /*
+ * Per-block record for the unindexed directory btree. This is carefully
+ * crafted so that the rec_len and name_len records of an ocfs2_dir_entry are
+ * mirrored. That way, the directory manipulation code needs a minimal amount
+ * of update.
+ *
+ * NOTE: Keep this structure aligned to a multiple of 4 bytes.
+ */
+struct ocfs2_dir_block_trailer {
+/*00*/ __le64          db_compat_inode;        /* Always zero. Was inode */
+
+       __le16          db_compat_rec_len;      /* Backwards compatible with
+                                                * ocfs2_dir_entry. */
+       __u8            db_compat_name_len;     /* Always zero. Was name_len */
+       __u8            db_reserved0;
+       __le16          db_reserved1;
+       __le16          db_free_rec_len;        /* Size of largest empty hole
+                                                * in this block. (unused) */
+/*10*/ __u8            db_signature[8];        /* Signature for verification */
+       __le64          db_reserved2;
+       __le64          db_free_next;           /* Next block in list (unused) */
+/*20*/ __le64          db_blkno;               /* Offset on disk, in blocks */
+       __le64          db_parent_dinode;       /* dinode which owns me, in
+                                                  blocks */
+/*30*/ struct ocfs2_block_check db_check;      /* Error checking */
+/*40*/
+};
+
+ /*
+ * A directory entry in the indexed tree. We don't store the full name here,
+ * but instead provide a pointer to the full dirent in the unindexed tree.
+ *
+ * We also store name_len here so as to reduce the number of leaf blocks we
+ * need to search in case of collisions.
+ */
+struct ocfs2_dx_entry {
+       __le32          dx_major_hash;  /* Used to find logical
+                                        * cluster in index */
+       __le32          dx_minor_hash;  /* Lower bits used to find
+                                        * block in cluster */
+       __le64          dx_dirent_blk;  /* Physical block in unindexed
+                                        * tree holding this dirent. */
+};
+
+struct ocfs2_dx_entry_list {
+       __le32          de_reserved;
+       __le16          de_count;       /* Maximum number of entries
+                                        * possible in de_entries */
+       __le16          de_num_used;    /* Current number of
+                                        * de_entries entries */
+       struct  ocfs2_dx_entry          de_entries[0];  /* Indexed dir entries
+                                                        * in a packed array of
+                                                        * length de_num_used */
+};
+
+#define OCFS2_DX_FLAG_INLINE   0x01
+
+/*
+ * A directory indexing block. Each indexed directory has one of these,
+ * pointed to by ocfs2_dinode.
+ *
+ * This block stores an indexed btree root, and a set of free space
+ * start-of-list pointers.
+ */
+struct ocfs2_dx_root_block {
+       __u8            dr_signature[8];        /* Signature for verification */
+       struct ocfs2_block_check dr_check;      /* Error checking */
+       __le16          dr_suballoc_slot;       /* Slot suballocator this
+                                                * block belongs to. */
+       __le16          dr_suballoc_bit;        /* Bit offset in suballocator
+                                                * block group */
+       __le32          dr_fs_generation;       /* Must match super block */
+       __le64          dr_blkno;               /* Offset on disk, in blocks */
+       __le64          dr_last_eb_blk;         /* Pointer to last
+                                                * extent block */
+       __le32          dr_clusters;            /* Clusters allocated
+                                                * to the indexed tree. */
+       __u8            dr_flags;               /* OCFS2_DX_FLAG_* flags */
+       __u8            dr_reserved0;
+       __le16          dr_reserved1;
+       __le64          dr_dir_blkno;           /* Pointer to parent inode */
+       __le32          dr_num_entries;         /* Total number of
+                                                * names stored in
+                                                * this directory.*/
+       __le32          dr_reserved2;
+       __le64          dr_free_blk;            /* Pointer to head of free
+                                                * unindexed block list. */
+       __le64          dr_reserved3[15];
+       union {
+               struct ocfs2_extent_list dr_list; /* Keep this aligned to 128
+                                                  * bits for maximum space
+                                                  * efficiency. */
+               struct ocfs2_dx_entry_list dr_entries; /* In-root-block list of
+                                                       * entries. We grow out
+                                                       * to extents if this
+                                                       * gets too big. */
+       };
+};
+
+/*
+ * The header of a leaf block in the indexed tree.
+ */
+struct ocfs2_dx_leaf {
+       __u8            dl_signature[8];/* Signature for verification */
+       struct ocfs2_block_check dl_check;      /* Error checking */
+       __le64          dl_blkno;       /* Offset on disk, in blocks */
+       __le32          dl_fs_generation;/* Must match super block */
+       __le32          dl_reserved0;
+       __le64          dl_reserved1;
+       struct ocfs2_dx_entry_list      dl_list;
+};
+
+/*
  * On disk allocator group structure for OCFS2
  */
 struct ocfs2_group_desc
@@ -613,10 +915,323 @@ struct ocfs2_group_desc
 /*20*/ __le64   bg_parent_dinode;       /* dinode which owns me, in
                                           blocks */
        __le64   bg_blkno;               /* Offset on disk, in blocks */
-/*30*/ __le64   bg_reserved2[2];
+/*30*/ struct ocfs2_block_check bg_check;      /* Error checking */
+       __le64   bg_reserved2;
 /*40*/ __u8    bg_bitmap[0];
 };
 
+struct ocfs2_refcount_rec {
+/*00*/ __le64 r_cpos;          /* Physical offset, in clusters */
+       __le32 r_clusters;      /* Clusters covered by this extent */
+       __le32 r_refcount;      /* Reference count of this extent */
+/*10*/
+};
+#define OCFS2_32BIT_POS_MASK           (0xffffffffULL)
+
+#define OCFS2_REFCOUNT_LEAF_FL          (0x00000001)
+#define OCFS2_REFCOUNT_TREE_FL          (0x00000002)
+
+struct ocfs2_refcount_list {
+/*00*/ __le16 rl_count;        /* Maximum number of entries possible
+                                  in rl_records */
+       __le16 rl_used;         /* Current number of used records */
+       __le32 rl_reserved2;
+       __le64 rl_reserved1;    /* Pad to sizeof(ocfs2_refcount_record) */
+/*10*/ struct ocfs2_refcount_rec rl_recs[0];   /* Refcount records */
+};
+
+
+struct ocfs2_refcount_block {
+/*00*/ __u8 rf_signature[8];           /* Signature for verification */
+       __le16 rf_suballoc_slot;        /* Slot suballocator this block
+                                          belongs to */
+       __le16 rf_suballoc_bit;         /* Bit offset in suballocator
+                                          block group */
+       __le32 rf_fs_generation;        /* Must match superblock */
+/*10*/ __le64 rf_blkno;                /* Offset on disk, in blocks */
+       __le64 rf_parent;               /* Parent block, only valid if
+                                          OCFS2_REFCOUNT_LEAF_FL is set in
+                                          rf_flags */
+/*20*/ struct ocfs2_block_check rf_check;      /* Error checking */
+       __le64 rf_last_eb_blk;          /* Pointer to last extent block */
+/*30*/ __le32 rf_count;                /* Number of inodes sharing this
+                                          refcount tree */
+       __le32 rf_flags;                /* See the flags above */
+       __le32 rf_clusters;             /* clusters covered by refcount tree. */
+       __le32 rf_cpos;                 /* cluster offset in refcount tree.*/
+/*40*/ __le32 rf_generation;           /* generation number. all be the same
+                                        * for the same refcount tree. */
+       __le32 rf_reserved0;
+       __le64 rf_reserved1[7];
+/*80*/ union {
+               struct ocfs2_refcount_list rf_records;  /* List of refcount
+                                                         records */
+               struct ocfs2_extent_list rf_list;       /* Extent record list,
+                                                       only valid if
+                                                       OCFS2_REFCOUNT_TREE_FL
+                                                       is set in rf_flags */
+       };
+/* Actual on-disk size is one block */
+};
+
+/*
+ * On disk extended attribute structure for OCFS2.
+ */
+
+/*
+ * ocfs2_xattr_entry indicates one extend attribute.
+ *
+ * Note that it can be stored in inode, one block or one xattr bucket.
+ */
+struct ocfs2_xattr_entry {
+       __le32  xe_name_hash;    /* hash value of xattr prefix+suffix. */
+       __le16  xe_name_offset;  /* byte offset from the 1st entry in the
+                                   local xattr storage(inode, xattr block or
+                                   xattr bucket). */
+       __u8    xe_name_len;     /* xattr name len, does't include prefix. */
+       __u8    xe_type;         /* the low 7 bits indicate the name prefix
+                                 * type and the highest bit indicates whether
+                                 * the EA is stored in the local storage. */
+       __le64  xe_value_size;   /* real xattr value length. */
+};
+
+/*
+ * On disk structure for xattr header.
+ *
+ * One ocfs2_xattr_header describes how many ocfs2_xattr_entry records in
+ * the local xattr storage.
+ */
+struct ocfs2_xattr_header {
+       __le16  xh_count;                       /* contains the count of how
+                                                  many records are in the
+                                                  local xattr storage. */
+       __le16  xh_free_start;                  /* current offset for storing
+                                                  xattr. */
+       __le16  xh_name_value_len;              /* total length of name/value
+                                                  length in this bucket. */
+       __le16  xh_num_buckets;                 /* Number of xattr buckets
+                                                  in this extent record,
+                                                  only valid in the first
+                                                  bucket. */
+       struct ocfs2_block_check xh_check;      /* Error checking
+                                                  (Note, this is only
+                                                   used for xattr
+                                                   buckets.  A block uses
+                                                   xb_check and sets
+                                                   this field to zero.) */
+       struct ocfs2_xattr_entry xh_entries[0]; /* xattr entry list. */
+};
+
+/*
+ * On disk structure for xattr value root.
+ *
+ * When an xattr's value is large enough, it is stored in an external
+ * b-tree like file data.  The xattr value root points to this structure.
+ */
+struct ocfs2_xattr_value_root {
+/*00*/ __le32  xr_clusters;              /* clusters covered by xattr value. */
+       __le32  xr_reserved0;
+       __le64  xr_last_eb_blk;           /* Pointer to last extent block */
+/*10*/ struct ocfs2_extent_list xr_list; /* Extent record list */
+};
+
+/*
+ * On disk structure for xattr tree root.
+ *
+ * It is used when there are too many extended attributes for one file. These
+ * attributes will be organized and stored in an indexed-btree.
+ */
+struct ocfs2_xattr_tree_root {
+/*00*/ __le32  xt_clusters;              /* clusters covered by xattr. */
+       __le32  xt_reserved0;
+       __le64  xt_last_eb_blk;           /* Pointer to last extent block */
+/*10*/ struct ocfs2_extent_list xt_list; /* Extent record list */
+};
+
+#define OCFS2_XATTR_INDEXED    0x1
+#define OCFS2_HASH_SHIFT       5
+#define OCFS2_XATTR_ROUND      3
+#define OCFS2_XATTR_SIZE(size) (((size) + OCFS2_XATTR_ROUND) & \
+                               ~(OCFS2_XATTR_ROUND))
+
+#define OCFS2_XATTR_BUCKET_SIZE                        4096
+#define OCFS2_XATTR_MAX_BLOCKS_PER_BUCKET      (OCFS2_XATTR_BUCKET_SIZE \
+                                                / OCFS2_MIN_BLOCKSIZE)
+
+/*
+ * On disk structure for xattr block.
+ */
+struct ocfs2_xattr_block {
+/*00*/ __u8    xb_signature[8];     /* Signature for verification */
+       __le16  xb_suballoc_slot;    /* Slot suballocator this
+                                       block belongs to. */
+       __le16  xb_suballoc_bit;     /* Bit offset in suballocator
+                                       block group */
+       __le32  xb_fs_generation;    /* Must match super block */
+/*10*/ __le64  xb_blkno;            /* Offset on disk, in blocks */
+       struct ocfs2_block_check xb_check;      /* Error checking */
+/*20*/ __le16  xb_flags;            /* Indicates whether this block contains
+                                       real xattr or a xattr tree. */
+       __le16  xb_reserved0;
+       __le32  xb_reserved1;
+       __le64  xb_reserved2;
+/*30*/ union {
+               struct ocfs2_xattr_header xb_header; /* xattr header if this
+                                                       block contains xattr */
+               struct ocfs2_xattr_tree_root xb_root;/* xattr tree root if this
+                                                       block cotains xattr
+                                                       tree. */
+       } xb_attrs;
+};
+
+#define OCFS2_XATTR_ENTRY_LOCAL                0x80
+#define OCFS2_XATTR_TYPE_MASK          0x7F
+static inline void ocfs2_xattr_set_local(struct ocfs2_xattr_entry *xe,
+                                        int local)
+{
+       if (local)
+               xe->xe_type |= OCFS2_XATTR_ENTRY_LOCAL;
+       else
+               xe->xe_type &= ~OCFS2_XATTR_ENTRY_LOCAL;
+}
+
+static inline int ocfs2_xattr_is_local(struct ocfs2_xattr_entry *xe)
+{
+       return xe->xe_type & OCFS2_XATTR_ENTRY_LOCAL;
+}
+
+static inline void ocfs2_xattr_set_type(struct ocfs2_xattr_entry *xe, int type)
+{
+       xe->xe_type |= type & OCFS2_XATTR_TYPE_MASK;
+}
+
+static inline int ocfs2_xattr_get_type(struct ocfs2_xattr_entry *xe)
+{
+       return xe->xe_type & OCFS2_XATTR_TYPE_MASK;
+}
+
+/*
+ *  On disk structures for global quota file
+ */
+
+/* Magic numbers and known versions for global quota files */
+#define OCFS2_GLOBAL_QMAGICS {\
+       0x0cf52470, /* USRQUOTA */ \
+       0x0cf52471  /* GRPQUOTA */ \
+}
+
+#define OCFS2_GLOBAL_QVERSIONS {\
+       0, \
+       0, \
+}
+
+
+/* Each block of each quota file has a certain fixed number of bytes reserved
+ * for OCFS2 internal use at its end. OCFS2 can use it for things like
+ * checksums, etc. */
+#define OCFS2_QBLK_RESERVED_SPACE 8
+
+/* Generic header of all quota files */
+struct ocfs2_disk_dqheader {
+       __le32 dqh_magic;       /* Magic number identifying file */
+       __le32 dqh_version;     /* Quota format version */
+};
+
+#define OCFS2_GLOBAL_INFO_OFF (sizeof(struct ocfs2_disk_dqheader))
+
+/* Information header of global quota file (immediately follows the generic
+ * header) */
+struct ocfs2_global_disk_dqinfo {
+/*00*/ __le32 dqi_bgrace;      /* Grace time for space softlimit excess */
+       __le32 dqi_igrace;      /* Grace time for inode softlimit excess */
+       __le32 dqi_syncms;      /* Time after which we sync local changes to
+                                * global quota file */
+       __le32 dqi_blocks;      /* Number of blocks in quota file */
+/*10*/ __le32 dqi_free_blk;    /* First free block in quota file */
+       __le32 dqi_free_entry;  /* First block with free dquot entry in quota
+                                * file */
+};
+
+/* Structure with global user / group information. We reserve some space
+ * for future use. */
+struct ocfs2_global_disk_dqblk {
+/*00*/ __le32 dqb_id;          /* ID the structure belongs to */
+       __le32 dqb_use_count;   /* Number of nodes having reference to this structure */
+       __le64 dqb_ihardlimit;  /* absolute limit on allocated inodes */
+/*10*/ __le64 dqb_isoftlimit;  /* preferred inode limit */
+       __le64 dqb_curinodes;   /* current # allocated inodes */
+/*20*/ __le64 dqb_bhardlimit;  /* absolute limit on disk space */
+       __le64 dqb_bsoftlimit;  /* preferred limit on disk space */
+/*30*/ __le64 dqb_curspace;    /* current space occupied */
+       __le64 dqb_btime;       /* time limit for excessive disk use */
+/*40*/ __le64 dqb_itime;       /* time limit for excessive inode use */
+       __le64 dqb_pad1;
+/*50*/ __le64 dqb_pad2;
+};
+
+/*
+ *  On-disk structures for local quota file
+ */
+
+/* Magic numbers and known versions for local quota files */
+#define OCFS2_LOCAL_QMAGICS {\
+       0x0cf524c0, /* USRQUOTA */ \
+       0x0cf524c1  /* GRPQUOTA */ \
+}
+
+#define OCFS2_LOCAL_QVERSIONS {\
+       0, \
+       0, \
+}
+
+/* Quota flags in dqinfo header */
+#define OLQF_CLEAN     0x0001  /* Quota file is empty (this should be after\
+                                * quota has been cleanly turned off) */
+
+#define OCFS2_LOCAL_INFO_OFF (sizeof(struct ocfs2_disk_dqheader))
+
+/* Information header of local quota file (immediately follows the generic
+ * header) */
+struct ocfs2_local_disk_dqinfo {
+       __le32 dqi_flags;       /* Flags for quota file */
+       __le32 dqi_chunks;      /* Number of chunks of quota structures
+                                * with a bitmap */
+       __le32 dqi_blocks;      /* Number of blocks allocated for quota file */
+};
+
+/* Header of one chunk of a quota file */
+struct ocfs2_local_disk_chunk {
+       __le32 dqc_free;        /* Number of free entries in the bitmap */
+       __u8 dqc_bitmap[0];     /* Bitmap of entries in the corresponding
+                                * chunk of quota file */
+};
+
+/* One entry in local quota file */
+struct ocfs2_local_disk_dqblk {
+/*00*/ __le64 dqb_id;          /* id this quota applies to */
+       __le64 dqb_spacemod;    /* Change in the amount of used space */
+/*10*/ __le64 dqb_inodemod;    /* Change in the amount of used inodes */
+};
+
+
+/*
+ * The quota trailer lives at the end of each quota block.
+ */
+
+struct ocfs2_disk_dqtrailer {
+/*00*/ struct ocfs2_block_check dq_check;      /* Error checking */
+/*08*/ /* Cannot be larger than OCFS2_QBLK_RESERVED_SPACE */
+};
+
+static inline struct ocfs2_disk_dqtrailer *ocfs2_block_dqtrailer(int blocksize,
+                                                                void *buf)
+{
+       char *ptr = buf;
+       ptr += blocksize - OCFS2_QBLK_RESERVED_SPACE;
+
+       return (struct ocfs2_disk_dqtrailer *)ptr;
+}
+
 #ifdef __KERNEL__
 static inline int ocfs2_fast_symlink_chars(struct super_block *sb)
 {
@@ -624,10 +1239,18 @@ static inline int ocfs2_fast_symlink_chars(struct super_block *sb)
                 offsetof(struct ocfs2_dinode, id2.i_symlink);
 }
 
-static inline int ocfs2_max_inline_data(struct super_block *sb)
+static inline int ocfs2_max_inline_data_with_xattr(struct super_block *sb,
+                                                  struct ocfs2_dinode *di)
 {
-       return sb->s_blocksize -
-               offsetof(struct ocfs2_dinode, id2.i_data.id_data);
+       unsigned int xattrsize = le16_to_cpu(di->i_xattr_inline_size);
+
+       if (le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_XATTR_FL)
+               return sb->s_blocksize -
+                       offsetof(struct ocfs2_dinode, id2.i_data.id_data) -
+                       xattrsize;
+       else
+               return sb->s_blocksize -
+                       offsetof(struct ocfs2_dinode, id2.i_data.id_data);
 }
 
 static inline int ocfs2_extent_recs_per_inode(struct super_block *sb)
@@ -640,6 +1263,34 @@ static inline int ocfs2_extent_recs_per_inode(struct super_block *sb)
        return size / sizeof(struct ocfs2_extent_rec);
 }
 
+static inline int ocfs2_extent_recs_per_inode_with_xattr(
+                                               struct super_block *sb,
+                                               struct ocfs2_dinode *di)
+{
+       int size;
+       unsigned int xattrsize = le16_to_cpu(di->i_xattr_inline_size);
+
+       if (le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_XATTR_FL)
+               size = sb->s_blocksize -
+                       offsetof(struct ocfs2_dinode, id2.i_list.l_recs) -
+                       xattrsize;
+       else
+               size = sb->s_blocksize -
+                       offsetof(struct ocfs2_dinode, id2.i_list.l_recs);
+
+       return size / sizeof(struct ocfs2_extent_rec);
+}
+
+static inline int ocfs2_extent_recs_per_dx_root(struct super_block *sb)
+{
+       int size;
+
+       size = sb->s_blocksize -
+               offsetof(struct ocfs2_dx_root_block, dr_list.l_recs);
+
+       return size / sizeof(struct ocfs2_extent_rec);
+}
+
 static inline int ocfs2_chain_recs_per_inode(struct super_block *sb)
 {
        int size;
@@ -660,6 +1311,26 @@ static inline u16 ocfs2_extent_recs_per_eb(struct super_block *sb)
        return size / sizeof(struct ocfs2_extent_rec);
 }
 
+static inline int ocfs2_dx_entries_per_leaf(struct super_block *sb)
+{
+       int size;
+
+       size = sb->s_blocksize -
+               offsetof(struct ocfs2_dx_leaf, dl_list.de_entries);
+
+       return size / sizeof(struct ocfs2_dx_entry);
+}
+
+static inline int ocfs2_dx_entries_per_root(struct super_block *sb)
+{
+       int size;
+
+       size = sb->s_blocksize -
+               offsetof(struct ocfs2_dx_root_block, dr_entries.de_entries);
+
+       return size / sizeof(struct ocfs2_dx_entry);
+}
+
 static inline u16 ocfs2_local_alloc_size(struct super_block *sb)
 {
        u16 size;
@@ -703,6 +1374,43 @@ static inline u64 ocfs2_backup_super_blkno(struct super_block *sb, int index)
        return 0;
 
 }
+
+static inline u16 ocfs2_xattr_recs_per_xb(struct super_block *sb)
+{
+       int size;
+
+       size = sb->s_blocksize -
+               offsetof(struct ocfs2_xattr_block,
+                        xb_attrs.xb_root.xt_list.l_recs);
+
+       return size / sizeof(struct ocfs2_extent_rec);
+}
+
+static inline u16 ocfs2_extent_recs_per_rb(struct super_block *sb)
+{
+       int size;
+
+       size = sb->s_blocksize -
+               offsetof(struct ocfs2_refcount_block, rf_list.l_recs);
+
+       return size / sizeof(struct ocfs2_extent_rec);
+}
+
+static inline u16 ocfs2_refcount_recs_per_rb(struct super_block *sb)
+{
+       int size;
+
+       size = sb->s_blocksize -
+               offsetof(struct ocfs2_refcount_block, rf_records.rl_recs);
+
+       return size / sizeof(struct ocfs2_refcount_rec);
+}
+
+static inline u32
+ocfs2_get_ref_rec_low_cpos(const struct ocfs2_refcount_rec *rec)
+{
+       return le64_to_cpu(rec->r_cpos) & OCFS2_32BIT_POS_MASK;
+}
 #else
 static inline int ocfs2_fast_symlink_chars(int blocksize)
 {
@@ -786,6 +1494,17 @@ static inline uint64_t ocfs2_backup_super_blkno(int blocksize, int index)
 
        return 0;
 }
+
+static inline int ocfs2_xattr_recs_per_xb(int blocksize)
+{
+       int size;
+
+       size = blocksize -
+               offsetof(struct ocfs2_xattr_block,
+                        xb_attrs.xb_root.xt_list.l_recs);
+
+       return size / sizeof(struct ocfs2_extent_rec);
+}
 #endif  /* __KERNEL__ */
 
 
@@ -806,7 +1525,7 @@ static inline int ocfs2_sprintf_system_inode_name(char *buf, int len,
          * list has a copy per slot.
          */
        if (type <= OCFS2_LAST_GLOBAL_SYSTEM_INODE)
-               chars = snprintf(buf, len,
+               chars = snprintf(buf, len, "%s",
                                 ocfs2_system_inodes[type].si_name);
        else
                chars = snprintf(buf, len,