Ocfs2: Fix hole punching to correctly do CoW during cluster zeroing.
[safe/jmp/linux-2.6] / fs / ocfs2 / stack_user.c
index a5e58e2..5ae8812 100644 (file)
 #include <linux/fs.h>
 #include <linux/miscdevice.h>
 #include <linux/mutex.h>
+#include <linux/smp_lock.h>
 #include <linux/reboot.h>
 #include <asm/uaccess.h>
 
 #include "stackglue.h"
 
+#include <linux/dlm_plock.h>
 
 /*
  * The control protocol starts with a handshake.  Until the handshake
  * unknown, -EINVAL is returned.  Once the negotiation is complete, the
  * client can start sending messages.
  *
- * The T01 protocol only has one message, "DOWN".  It has the following
- * syntax:
+ * The T01 protocol has three messages.  First is the "SETN" message.
+ * It has the following syntax:
+ *
+ *  SETN<space><8-char-hex-nodenum><newline>
+ *
+ * This is 14 characters.
+ *
+ * The "SETN" message must be the first message following the protocol.
+ * It tells ocfs2_control the local node number.
+ *
+ * Next comes the "SETV" message.  It has the following syntax:
+ *
+ *  SETV<space><2-char-hex-major><space><2-char-hex-minor><newline>
+ *
+ * This is 11 characters.
+ *
+ * The "SETV" message sets the filesystem locking protocol version as
+ * negotiated by the client.  The client negotiates based on the maximum
+ * version advertised in /sys/fs/ocfs2/max_locking_protocol.  The major
+ * number from the "SETV" message must match
+ * ocfs2_user_plugin.sp_max_proto.pv_major, and the minor number
+ * must be less than or equal to ...sp_max_version.pv_minor.
+ *
+ * Once this information has been set, mounts will be allowed.  From this
+ * point on, the "DOWN" message can be sent for node down notification.
+ * It has the following syntax:
  *
  *  DOWN<space><32-char-cap-hex-uuid><space><8-char-hex-nodenum><newline>
  *
  */
 #define OCFS2_CONTROL_PROTO                    "T01\n"
 #define OCFS2_CONTROL_PROTO_LEN                        4
+
+/* Handshake states */
 #define OCFS2_CONTROL_HANDSHAKE_INVALID                (0)
 #define OCFS2_CONTROL_HANDSHAKE_READ           (1)
-#define OCFS2_CONTROL_HANDSHAKE_VALID          (2)
-#define OCFS2_CONTROL_MESSAGE_DOWN             "DOWN"
-#define OCFS2_CONTROL_MESSAGE_DOWN_LEN         4
+#define OCFS2_CONTROL_HANDSHAKE_PROTOCOL       (2)
+#define OCFS2_CONTROL_HANDSHAKE_VALID          (3)
+
+/* Messages */
+#define OCFS2_CONTROL_MESSAGE_OP_LEN           4
+#define OCFS2_CONTROL_MESSAGE_SETNODE_OP       "SETN"
+#define OCFS2_CONTROL_MESSAGE_SETNODE_TOTAL_LEN        14
+#define OCFS2_CONTROL_MESSAGE_SETVERSION_OP    "SETV"
+#define OCFS2_CONTROL_MESSAGE_SETVERSION_TOTAL_LEN     11
+#define OCFS2_CONTROL_MESSAGE_DOWN_OP          "DOWN"
 #define OCFS2_CONTROL_MESSAGE_DOWN_TOTAL_LEN   47
 #define OCFS2_TEXT_UUID_LEN                    32
+#define OCFS2_CONTROL_MESSAGE_VERNUM_LEN       2
 #define OCFS2_CONTROL_MESSAGE_NODENUM_LEN      8
 
 /*
@@ -79,9 +115,50 @@ struct ocfs2_live_connection {
 struct ocfs2_control_private {
        struct list_head op_list;
        int op_state;
+       int op_this_node;
+       struct ocfs2_protocol_version op_proto;
+};
+
+/* SETN<space><8-char-hex-nodenum><newline> */
+struct ocfs2_control_message_setn {
+       char    tag[OCFS2_CONTROL_MESSAGE_OP_LEN];
+       char    space;
+       char    nodestr[OCFS2_CONTROL_MESSAGE_NODENUM_LEN];
+       char    newline;
+};
+
+/* SETV<space><2-char-hex-major><space><2-char-hex-minor><newline> */
+struct ocfs2_control_message_setv {
+       char    tag[OCFS2_CONTROL_MESSAGE_OP_LEN];
+       char    space1;
+       char    major[OCFS2_CONTROL_MESSAGE_VERNUM_LEN];
+       char    space2;
+       char    minor[OCFS2_CONTROL_MESSAGE_VERNUM_LEN];
+       char    newline;
 };
 
+/* DOWN<space><32-char-cap-hex-uuid><space><8-char-hex-nodenum><newline> */
+struct ocfs2_control_message_down {
+       char    tag[OCFS2_CONTROL_MESSAGE_OP_LEN];
+       char    space1;
+       char    uuid[OCFS2_TEXT_UUID_LEN];
+       char    space2;
+       char    nodestr[OCFS2_CONTROL_MESSAGE_NODENUM_LEN];
+       char    newline;
+};
+
+union ocfs2_control_message {
+       char                                    tag[OCFS2_CONTROL_MESSAGE_OP_LEN];
+       struct ocfs2_control_message_setn       u_setn;
+       struct ocfs2_control_message_setv       u_setv;
+       struct ocfs2_control_message_down       u_down;
+};
+
+static struct ocfs2_stack_plugin ocfs2_user_plugin;
+
 static atomic_t ocfs2_control_opened;
+static int ocfs2_control_this_node = -1;
+static struct ocfs2_protocol_version running_proto;
 
 static LIST_HEAD(ocfs2_live_connection_list);
 static LIST_HEAD(ocfs2_control_private_list);
@@ -166,38 +243,37 @@ static void ocfs2_live_connection_drop(struct ocfs2_live_connection *c)
        kfree(c);
 }
 
-static ssize_t ocfs2_control_cfu(void *target, size_t target_len,
-                                const char __user *buf, size_t count)
+static int ocfs2_control_cfu(void *target, size_t target_len,
+                            const char __user *buf, size_t count)
 {
        /* The T01 expects write(2) calls to have exactly one command */
-       if (count != target_len)
+       if ((count != target_len) ||
+           (count > sizeof(union ocfs2_control_message)))
                return -EINVAL;
 
        if (copy_from_user(target, buf, target_len))
                return -EFAULT;
 
-       return count;
+       return 0;
 }
 
-static ssize_t ocfs2_control_validate_handshake(struct file *file,
-                                               const char __user *buf,
-                                               size_t count)
+static ssize_t ocfs2_control_validate_protocol(struct file *file,
+                                              const char __user *buf,
+                                              size_t count)
 {
        ssize_t ret;
        char kbuf[OCFS2_CONTROL_PROTO_LEN];
 
        ret = ocfs2_control_cfu(kbuf, OCFS2_CONTROL_PROTO_LEN,
                                buf, count);
-       if (ret != count)
+       if (ret)
                return ret;
 
        if (strncmp(kbuf, OCFS2_CONTROL_PROTO, OCFS2_CONTROL_PROTO_LEN))
                return -EINVAL;
 
-       atomic_inc(&ocfs2_control_opened);
        ocfs2_control_set_handshake_state(file,
-                                         OCFS2_CONTROL_HANDSHAKE_VALID);
-
+                                         OCFS2_CONTROL_HANDSHAKE_PROTOCOL);
 
        return count;
 }
@@ -219,45 +295,174 @@ static void ocfs2_control_send_down(const char *uuid,
        mutex_unlock(&ocfs2_control_lock);
 }
 
-/* DOWN<space><32-char-cap-hex-uuid><space><8-char-hex-nodenum><newline> */
-struct ocfs2_control_message_down {
-       char    tag[OCFS2_CONTROL_MESSAGE_DOWN_LEN];
-       char    space1;
-       char    uuid[OCFS2_TEXT_UUID_LEN];
-       char    space2;
-       char    nodestr[OCFS2_CONTROL_MESSAGE_NODENUM_LEN];
-       char    newline;
-};
+/*
+ * Called whenever configuration elements are sent to /dev/ocfs2_control.
+ * If all configuration elements are present, try to set the global
+ * values.  If there is a problem, return an error.  Skip any missing
+ * elements, and only bump ocfs2_control_opened when we have all elements
+ * and are successful.
+ */
+static int ocfs2_control_install_private(struct file *file)
+{
+       int rc = 0;
+       int set_p = 1;
+       struct ocfs2_control_private *p = file->private_data;
 
-static ssize_t ocfs2_control_message(struct file *file,
-                                    const char __user *buf,
-                                    size_t count)
+       BUG_ON(p->op_state != OCFS2_CONTROL_HANDSHAKE_PROTOCOL);
+
+       mutex_lock(&ocfs2_control_lock);
+
+       if (p->op_this_node < 0) {
+               set_p = 0;
+       } else if ((ocfs2_control_this_node >= 0) &&
+                  (ocfs2_control_this_node != p->op_this_node)) {
+               rc = -EINVAL;
+               goto out_unlock;
+       }
+
+       if (!p->op_proto.pv_major) {
+               set_p = 0;
+       } else if (!list_empty(&ocfs2_live_connection_list) &&
+                  ((running_proto.pv_major != p->op_proto.pv_major) ||
+                   (running_proto.pv_minor != p->op_proto.pv_minor))) {
+               rc = -EINVAL;
+               goto out_unlock;
+       }
+
+       if (set_p) {
+               ocfs2_control_this_node = p->op_this_node;
+               running_proto.pv_major = p->op_proto.pv_major;
+               running_proto.pv_minor = p->op_proto.pv_minor;
+       }
+
+out_unlock:
+       mutex_unlock(&ocfs2_control_lock);
+
+       if (!rc && set_p) {
+               /* We set the global values successfully */
+               atomic_inc(&ocfs2_control_opened);
+               ocfs2_control_set_handshake_state(file,
+                                       OCFS2_CONTROL_HANDSHAKE_VALID);
+       }
+
+       return rc;
+}
+
+static int ocfs2_control_get_this_node(void)
+{
+       int rc;
+
+       mutex_lock(&ocfs2_control_lock);
+       if (ocfs2_control_this_node < 0)
+               rc = -EINVAL;
+       else
+               rc = ocfs2_control_this_node;
+       mutex_unlock(&ocfs2_control_lock);
+
+       return rc;
+}
+
+static int ocfs2_control_do_setnode_msg(struct file *file,
+                                       struct ocfs2_control_message_setn *msg)
 {
-       ssize_t ret;
-       char *p = NULL;
        long nodenum;
-       struct ocfs2_control_message_down msg;
+       char *ptr = NULL;
+       struct ocfs2_control_private *p = file->private_data;
 
-       /* Try to catch padding issues */
-       WARN_ON(offsetof(struct ocfs2_control_message_down, uuid) !=
-               (sizeof(msg.tag) + sizeof(msg.space1)));
+       if (ocfs2_control_get_handshake_state(file) !=
+           OCFS2_CONTROL_HANDSHAKE_PROTOCOL)
+               return -EINVAL;
 
-       memset(&msg, 0, sizeof(struct ocfs2_control_message_down));
-       ret = ocfs2_control_cfu(&msg, OCFS2_CONTROL_MESSAGE_DOWN_TOTAL_LEN,
-                               buf, count);
-       if (ret != count)
-               return ret;
+       if (strncmp(msg->tag, OCFS2_CONTROL_MESSAGE_SETNODE_OP,
+                   OCFS2_CONTROL_MESSAGE_OP_LEN))
+               return -EINVAL;
+
+       if ((msg->space != ' ') || (msg->newline != '\n'))
+               return -EINVAL;
+       msg->space = msg->newline = '\0';
 
-       if (strncmp(msg.tag, OCFS2_CONTROL_MESSAGE_DOWN,
-                   strlen(OCFS2_CONTROL_MESSAGE_DOWN)))
+       nodenum = simple_strtol(msg->nodestr, &ptr, 16);
+       if (!ptr || *ptr)
                return -EINVAL;
 
-       if ((msg.space1 != ' ') || (msg.space2 != ' ') ||
-           (msg.newline != '\n'))
+       if ((nodenum == LONG_MIN) || (nodenum == LONG_MAX) ||
+           (nodenum > INT_MAX) || (nodenum < 0))
+               return -ERANGE;
+       p->op_this_node = nodenum;
+
+       return ocfs2_control_install_private(file);
+}
+
+static int ocfs2_control_do_setversion_msg(struct file *file,
+                                          struct ocfs2_control_message_setv *msg)
+ {
+       long major, minor;
+       char *ptr = NULL;
+       struct ocfs2_control_private *p = file->private_data;
+       struct ocfs2_protocol_version *max =
+               &ocfs2_user_plugin.sp_max_proto;
+
+       if (ocfs2_control_get_handshake_state(file) !=
+           OCFS2_CONTROL_HANDSHAKE_PROTOCOL)
+               return -EINVAL;
+
+       if (strncmp(msg->tag, OCFS2_CONTROL_MESSAGE_SETVERSION_OP,
+                   OCFS2_CONTROL_MESSAGE_OP_LEN))
                return -EINVAL;
-       msg.space1 = msg.space2 = msg.newline = '\0';
 
-       nodenum = simple_strtol(msg.nodestr, &p, 16);
+       if ((msg->space1 != ' ') || (msg->space2 != ' ') ||
+           (msg->newline != '\n'))
+               return -EINVAL;
+       msg->space1 = msg->space2 = msg->newline = '\0';
+
+       major = simple_strtol(msg->major, &ptr, 16);
+       if (!ptr || *ptr)
+               return -EINVAL;
+       minor = simple_strtol(msg->minor, &ptr, 16);
+       if (!ptr || *ptr)
+               return -EINVAL;
+
+       /*
+        * The major must be between 1 and 255, inclusive.  The minor
+        * must be between 0 and 255, inclusive.  The version passed in
+        * must be within the maximum version supported by the filesystem.
+        */
+       if ((major == LONG_MIN) || (major == LONG_MAX) ||
+           (major > (u8)-1) || (major < 1))
+               return -ERANGE;
+       if ((minor == LONG_MIN) || (minor == LONG_MAX) ||
+           (minor > (u8)-1) || (minor < 0))
+               return -ERANGE;
+       if ((major != max->pv_major) ||
+           (minor > max->pv_minor))
+               return -EINVAL;
+
+       p->op_proto.pv_major = major;
+       p->op_proto.pv_minor = minor;
+
+       return ocfs2_control_install_private(file);
+}
+
+static int ocfs2_control_do_down_msg(struct file *file,
+                                    struct ocfs2_control_message_down *msg)
+{
+       long nodenum;
+       char *p = NULL;
+
+       if (ocfs2_control_get_handshake_state(file) !=
+           OCFS2_CONTROL_HANDSHAKE_VALID)
+               return -EINVAL;
+
+       if (strncmp(msg->tag, OCFS2_CONTROL_MESSAGE_DOWN_OP,
+                   OCFS2_CONTROL_MESSAGE_OP_LEN))
+               return -EINVAL;
+
+       if ((msg->space1 != ' ') || (msg->space2 != ' ') ||
+           (msg->newline != '\n'))
+               return -EINVAL;
+       msg->space1 = msg->space2 = msg->newline = '\0';
+
+       nodenum = simple_strtol(msg->nodestr, &p, 16);
        if (!p || *p)
                return -EINVAL;
 
@@ -265,9 +470,44 @@ static ssize_t ocfs2_control_message(struct file *file,
            (nodenum > INT_MAX) || (nodenum < 0))
                return -ERANGE;
 
-       ocfs2_control_send_down(msg.uuid, nodenum);
+       ocfs2_control_send_down(msg->uuid, nodenum);
 
-       return count;
+       return 0;
+}
+
+static ssize_t ocfs2_control_message(struct file *file,
+                                    const char __user *buf,
+                                    size_t count)
+{
+       ssize_t ret;
+       union ocfs2_control_message msg;
+
+       /* Try to catch padding issues */
+       WARN_ON(offsetof(struct ocfs2_control_message_down, uuid) !=
+               (sizeof(msg.u_down.tag) + sizeof(msg.u_down.space1)));
+
+       memset(&msg, 0, sizeof(union ocfs2_control_message));
+       ret = ocfs2_control_cfu(&msg, count, buf, count);
+       if (ret)
+               goto out;
+
+       if ((count == OCFS2_CONTROL_MESSAGE_SETNODE_TOTAL_LEN) &&
+           !strncmp(msg.tag, OCFS2_CONTROL_MESSAGE_SETNODE_OP,
+                    OCFS2_CONTROL_MESSAGE_OP_LEN))
+               ret = ocfs2_control_do_setnode_msg(file, &msg.u_setn);
+       else if ((count == OCFS2_CONTROL_MESSAGE_SETVERSION_TOTAL_LEN) &&
+                !strncmp(msg.tag, OCFS2_CONTROL_MESSAGE_SETVERSION_OP,
+                         OCFS2_CONTROL_MESSAGE_OP_LEN))
+               ret = ocfs2_control_do_setversion_msg(file, &msg.u_setv);
+       else if ((count == OCFS2_CONTROL_MESSAGE_DOWN_TOTAL_LEN) &&
+                !strncmp(msg.tag, OCFS2_CONTROL_MESSAGE_DOWN_OP,
+                         OCFS2_CONTROL_MESSAGE_OP_LEN))
+               ret = ocfs2_control_do_down_msg(file, &msg.u_down);
+       else
+               ret = -EINVAL;
+
+out:
+       return ret ? ret : count;
 }
 
 static ssize_t ocfs2_control_write(struct file *file,
@@ -283,10 +523,11 @@ static ssize_t ocfs2_control_write(struct file *file,
                        break;
 
                case OCFS2_CONTROL_HANDSHAKE_READ:
-                       ret = ocfs2_control_validate_handshake(file, buf,
-                                                              count);
+                       ret = ocfs2_control_validate_protocol(file, buf,
+                                                             count);
                        break;
 
+               case OCFS2_CONTROL_HANDSHAKE_PROTOCOL:
                case OCFS2_CONTROL_HANDSHAKE_VALID:
                        ret = ocfs2_control_message(file, buf, count);
                        break;
@@ -309,26 +550,17 @@ static ssize_t ocfs2_control_read(struct file *file,
                                  size_t count,
                                  loff_t *ppos)
 {
-       char *proto_string = OCFS2_CONTROL_PROTO;
-       size_t to_write = 0;
-
-       if (*ppos >= OCFS2_CONTROL_PROTO_LEN)
-               return 0;
-
-       to_write = OCFS2_CONTROL_PROTO_LEN - *ppos;
-       if (to_write > count)
-               to_write = count;
-       if (copy_to_user(buf, proto_string + *ppos, to_write))
-               return -EFAULT;
+       ssize_t ret;
 
-       *ppos += to_write;
+       ret = simple_read_from_buffer(buf, count, ppos,
+                       OCFS2_CONTROL_PROTO, OCFS2_CONTROL_PROTO_LEN);
 
        /* Have we read the whole protocol list? */
-       if (*ppos >= OCFS2_CONTROL_PROTO_LEN)
+       if (ret > 0 && *ppos >= OCFS2_CONTROL_PROTO_LEN)
                ocfs2_control_set_handshake_state(file,
                                                  OCFS2_CONTROL_HANDSHAKE_READ);
 
-       return to_write;
+       return ret;
 }
 
 static int ocfs2_control_release(struct inode *inode, struct file *file)
@@ -350,6 +582,13 @@ static int ocfs2_control_release(struct inode *inode, struct file *file)
                               "an emergency restart!\n");
                        emergency_restart();
                }
+               /*
+                * Last valid close clears the node number and resets
+                * the locking protocol version
+                */
+               ocfs2_control_this_node = -1;
+               running_proto.pv_major = 0;
+               running_proto.pv_major = 0;
        }
 
 out:
@@ -370,11 +609,14 @@ static int ocfs2_control_open(struct inode *inode, struct file *file)
        p = kzalloc(sizeof(struct ocfs2_control_private), GFP_KERNEL);
        if (!p)
                return -ENOMEM;
+       p->op_this_node = -1;
 
+       lock_kernel();
        mutex_lock(&ocfs2_control_lock);
        file->private_data = p;
        list_add(&p->op_list, &ocfs2_control_private_list);
        mutex_unlock(&ocfs2_control_lock);
+       unlock_kernel();
 
        return 0;
 }
@@ -387,7 +629,7 @@ static const struct file_operations ocfs2_control_fops = {
        .owner   = THIS_MODULE,
 };
 
-struct miscdevice ocfs2_control_device = {
+static struct miscdevice ocfs2_control_device = {
        .minor          = MISC_DYNAMIC_MINOR,
        .name           = "ocfs2_control",
        .fops           = &ocfs2_control_fops,
@@ -421,18 +663,247 @@ static void ocfs2_control_exit(void)
                       -rc);
 }
 
-static int __init user_stack_init(void)
+static void fsdlm_lock_ast_wrapper(void *astarg)
+{
+       struct ocfs2_dlm_lksb *lksb = astarg;
+       int status = lksb->lksb_fsdlm.sb_status;
+
+       /*
+        * For now we're punting on the issue of other non-standard errors
+        * where we can't tell if the unlock_ast or lock_ast should be called.
+        * The main "other error" that's possible is EINVAL which means the
+        * function was called with invalid args, which shouldn't be possible
+        * since the caller here is under our control.  Other non-standard
+        * errors probably fall into the same category, or otherwise are fatal
+        * which means we can't carry on anyway.
+        */
+
+       if (status == -DLM_EUNLOCK || status == -DLM_ECANCEL)
+               lksb->lksb_conn->cc_proto->lp_unlock_ast(lksb, 0);
+       else
+               lksb->lksb_conn->cc_proto->lp_lock_ast(lksb);
+}
+
+static void fsdlm_blocking_ast_wrapper(void *astarg, int level)
+{
+       struct ocfs2_dlm_lksb *lksb = astarg;
+
+       lksb->lksb_conn->cc_proto->lp_blocking_ast(lksb, level);
+}
+
+static int user_dlm_lock(struct ocfs2_cluster_connection *conn,
+                        int mode,
+                        struct ocfs2_dlm_lksb *lksb,
+                        u32 flags,
+                        void *name,
+                        unsigned int namelen)
+{
+       int ret;
+
+       if (!lksb->lksb_fsdlm.sb_lvbptr)
+               lksb->lksb_fsdlm.sb_lvbptr = (char *)lksb +
+                                            sizeof(struct dlm_lksb);
+
+       ret = dlm_lock(conn->cc_lockspace, mode, &lksb->lksb_fsdlm,
+                      flags|DLM_LKF_NODLCKWT, name, namelen, 0,
+                      fsdlm_lock_ast_wrapper, lksb,
+                      fsdlm_blocking_ast_wrapper);
+       return ret;
+}
+
+static int user_dlm_unlock(struct ocfs2_cluster_connection *conn,
+                          struct ocfs2_dlm_lksb *lksb,
+                          u32 flags)
+{
+       int ret;
+
+       ret = dlm_unlock(conn->cc_lockspace, lksb->lksb_fsdlm.sb_lkid,
+                        flags, &lksb->lksb_fsdlm, lksb);
+       return ret;
+}
+
+static int user_dlm_lock_status(struct ocfs2_dlm_lksb *lksb)
+{
+       return lksb->lksb_fsdlm.sb_status;
+}
+
+static int user_dlm_lvb_valid(struct ocfs2_dlm_lksb *lksb)
+{
+       int invalid = lksb->lksb_fsdlm.sb_flags & DLM_SBF_VALNOTVALID;
+
+       return !invalid;
+}
+
+static void *user_dlm_lvb(struct ocfs2_dlm_lksb *lksb)
+{
+       if (!lksb->lksb_fsdlm.sb_lvbptr)
+               lksb->lksb_fsdlm.sb_lvbptr = (char *)lksb +
+                                            sizeof(struct dlm_lksb);
+       return (void *)(lksb->lksb_fsdlm.sb_lvbptr);
+}
+
+static void user_dlm_dump_lksb(struct ocfs2_dlm_lksb *lksb)
 {
-       return ocfs2_control_init();
 }
 
-static void __exit user_stack_exit(void)
+static int user_plock(struct ocfs2_cluster_connection *conn,
+                     u64 ino,
+                     struct file *file,
+                     int cmd,
+                     struct file_lock *fl)
+{
+       /*
+        * This more or less just demuxes the plock request into any
+        * one of three dlm calls.
+        *
+        * Internally, fs/dlm will pass these to a misc device, which
+        * a userspace daemon will read and write to.
+        *
+        * For now, cancel requests (which happen internally only),
+        * are turned into unlocks. Most of this function taken from
+        * gfs2_lock.
+        */
+
+       if (cmd == F_CANCELLK) {
+               cmd = F_SETLK;
+               fl->fl_type = F_UNLCK;
+       }
+
+       if (IS_GETLK(cmd))
+               return dlm_posix_get(conn->cc_lockspace, ino, file, fl);
+       else if (fl->fl_type == F_UNLCK)
+               return dlm_posix_unlock(conn->cc_lockspace, ino, file, fl);
+       else
+               return dlm_posix_lock(conn->cc_lockspace, ino, file, cmd, fl);
+}
+
+/*
+ * Compare a requested locking protocol version against the current one.
+ *
+ * If the major numbers are different, they are incompatible.
+ * If the current minor is greater than the request, they are incompatible.
+ * If the current minor is less than or equal to the request, they are
+ * compatible, and the requester should run at the current minor version.
+ */
+static int fs_protocol_compare(struct ocfs2_protocol_version *existing,
+                              struct ocfs2_protocol_version *request)
+{
+       if (existing->pv_major != request->pv_major)
+               return 1;
+
+       if (existing->pv_minor > request->pv_minor)
+               return 1;
+
+       if (existing->pv_minor < request->pv_minor)
+               request->pv_minor = existing->pv_minor;
+
+       return 0;
+}
+
+static int user_cluster_connect(struct ocfs2_cluster_connection *conn)
+{
+       dlm_lockspace_t *fsdlm;
+       struct ocfs2_live_connection *uninitialized_var(control);
+       int rc = 0;
+
+       BUG_ON(conn == NULL);
+
+       rc = ocfs2_live_connection_new(conn, &control);
+       if (rc)
+               goto out;
+
+       /*
+        * running_proto must have been set before we allowed any mounts
+        * to proceed.
+        */
+       if (fs_protocol_compare(&running_proto, &conn->cc_version)) {
+               printk(KERN_ERR
+                      "Unable to mount with fs locking protocol version "
+                      "%u.%u because the userspace control daemon has "
+                      "negotiated %u.%u\n",
+                      conn->cc_version.pv_major, conn->cc_version.pv_minor,
+                      running_proto.pv_major, running_proto.pv_minor);
+               rc = -EPROTO;
+               ocfs2_live_connection_drop(control);
+               goto out;
+       }
+
+       rc = dlm_new_lockspace(conn->cc_name, strlen(conn->cc_name),
+                              &fsdlm, DLM_LSFL_FS, DLM_LVB_LEN);
+       if (rc) {
+               ocfs2_live_connection_drop(control);
+               goto out;
+       }
+
+       conn->cc_private = control;
+       conn->cc_lockspace = fsdlm;
+out:
+       return rc;
+}
+
+static int user_cluster_disconnect(struct ocfs2_cluster_connection *conn)
+{
+       dlm_release_lockspace(conn->cc_lockspace, 2);
+       conn->cc_lockspace = NULL;
+       ocfs2_live_connection_drop(conn->cc_private);
+       conn->cc_private = NULL;
+       return 0;
+}
+
+static int user_cluster_this_node(unsigned int *this_node)
+{
+       int rc;
+
+       rc = ocfs2_control_get_this_node();
+       if (rc < 0)
+               return rc;
+
+       *this_node = rc;
+       return 0;
+}
+
+static struct ocfs2_stack_operations ocfs2_user_plugin_ops = {
+       .connect        = user_cluster_connect,
+       .disconnect     = user_cluster_disconnect,
+       .this_node      = user_cluster_this_node,
+       .dlm_lock       = user_dlm_lock,
+       .dlm_unlock     = user_dlm_unlock,
+       .lock_status    = user_dlm_lock_status,
+       .lvb_valid      = user_dlm_lvb_valid,
+       .lock_lvb       = user_dlm_lvb,
+       .plock          = user_plock,
+       .dump_lksb      = user_dlm_dump_lksb,
+};
+
+static struct ocfs2_stack_plugin ocfs2_user_plugin = {
+       .sp_name        = "user",
+       .sp_ops         = &ocfs2_user_plugin_ops,
+       .sp_owner       = THIS_MODULE,
+};
+
+
+static int __init ocfs2_user_plugin_init(void)
+{
+       int rc;
+
+       rc = ocfs2_control_init();
+       if (!rc) {
+               rc = ocfs2_stack_glue_register(&ocfs2_user_plugin);
+               if (rc)
+                       ocfs2_control_exit();
+       }
+
+       return rc;
+}
+
+static void __exit ocfs2_user_plugin_exit(void)
 {
+       ocfs2_stack_glue_unregister(&ocfs2_user_plugin);
        ocfs2_control_exit();
 }
 
 MODULE_AUTHOR("Oracle");
 MODULE_DESCRIPTION("ocfs2 driver for userspace cluster stacks");
 MODULE_LICENSE("GPL");
-module_init(user_stack_init);
-module_exit(user_stack_exit);
+module_init(ocfs2_user_plugin_init);
+module_exit(ocfs2_user_plugin_exit);