Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / net / llc / llc_conn.c
diff --git a/kernel/net/llc/llc_conn.c b/kernel/net/llc/llc_conn.c
new file mode 100644 (file)
index 0000000..81a61fc
--- /dev/null
@@ -0,0 +1,1016 @@
+/*
+ * llc_conn.c - Driver routines for connection component.
+ *
+ * Copyright (c) 1997 by Procom Technology, Inc.
+ *              2001-2003 by Arnaldo Carvalho de Melo <acme@conectiva.com.br>
+ *
+ * This program can be redistributed or modified under the terms of the
+ * GNU General Public License as published by the Free Software Foundation.
+ * This program is distributed without any warranty or implied warranty
+ * of merchantability or fitness for a particular purpose.
+ *
+ * See the GNU General Public License for more details.
+ */
+
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <net/llc_sap.h>
+#include <net/llc_conn.h>
+#include <net/sock.h>
+#include <net/tcp_states.h>
+#include <net/llc_c_ev.h>
+#include <net/llc_c_ac.h>
+#include <net/llc_c_st.h>
+#include <net/llc_pdu.h>
+
+#if 0
+#define dprintk(args...) printk(KERN_DEBUG args)
+#else
+#define dprintk(args...)
+#endif
+
+static int llc_find_offset(int state, int ev_type);
+static void llc_conn_send_pdus(struct sock *sk);
+static int llc_conn_service(struct sock *sk, struct sk_buff *skb);
+static int llc_exec_conn_trans_actions(struct sock *sk,
+                                      struct llc_conn_state_trans *trans,
+                                      struct sk_buff *ev);
+static struct llc_conn_state_trans *llc_qualify_conn_ev(struct sock *sk,
+                                                       struct sk_buff *skb);
+
+/* Offset table on connection states transition diagram */
+static int llc_offset_table[NBR_CONN_STATES][NBR_CONN_EV];
+
+int sysctl_llc2_ack_timeout = LLC2_ACK_TIME * HZ;
+int sysctl_llc2_p_timeout = LLC2_P_TIME * HZ;
+int sysctl_llc2_rej_timeout = LLC2_REJ_TIME * HZ;
+int sysctl_llc2_busy_timeout = LLC2_BUSY_TIME * HZ;
+
+/**
+ *     llc_conn_state_process - sends event to connection state machine
+ *     @sk: connection
+ *     @skb: occurred event
+ *
+ *     Sends an event to connection state machine. After processing event
+ *     (executing it's actions and changing state), upper layer will be
+ *     indicated or confirmed, if needed. Returns 0 for success, 1 for
+ *     failure. The socket lock has to be held before calling this function.
+ */
+int llc_conn_state_process(struct sock *sk, struct sk_buff *skb)
+{
+       int rc;
+       struct llc_sock *llc = llc_sk(skb->sk);
+       struct llc_conn_state_ev *ev = llc_conn_ev(skb);
+
+       /*
+        * We have to hold the skb, because llc_conn_service will kfree it in
+        * the sending path and we need to look at the skb->cb, where we encode
+        * llc_conn_state_ev.
+        */
+       skb_get(skb);
+       ev->ind_prim = ev->cfm_prim = 0;
+       /*
+        * Send event to state machine
+        */
+       rc = llc_conn_service(skb->sk, skb);
+       if (unlikely(rc != 0)) {
+               printk(KERN_ERR "%s: llc_conn_service failed\n", __func__);
+               goto out_kfree_skb;
+       }
+
+       if (unlikely(!ev->ind_prim && !ev->cfm_prim)) {
+               /* indicate or confirm not required */
+               if (!skb->next)
+                       goto out_kfree_skb;
+               goto out_skb_put;
+       }
+
+       if (unlikely(ev->ind_prim && ev->cfm_prim)) /* Paranoia */
+               skb_get(skb);
+
+       switch (ev->ind_prim) {
+       case LLC_DATA_PRIM:
+               llc_save_primitive(sk, skb, LLC_DATA_PRIM);
+               if (unlikely(sock_queue_rcv_skb(sk, skb))) {
+                       /*
+                        * shouldn't happen
+                        */
+                       printk(KERN_ERR "%s: sock_queue_rcv_skb failed!\n",
+                              __func__);
+                       kfree_skb(skb);
+               }
+               break;
+       case LLC_CONN_PRIM:
+               /*
+                * Can't be sock_queue_rcv_skb, because we have to leave the
+                * skb->sk pointing to the newly created struct sock in
+                * llc_conn_handler. -acme
+                */
+               skb_queue_tail(&sk->sk_receive_queue, skb);
+               sk->sk_state_change(sk);
+               break;
+       case LLC_DISC_PRIM:
+               sock_hold(sk);
+               if (sk->sk_type == SOCK_STREAM &&
+                   sk->sk_state == TCP_ESTABLISHED) {
+                       sk->sk_shutdown       = SHUTDOWN_MASK;
+                       sk->sk_socket->state  = SS_UNCONNECTED;
+                       sk->sk_state          = TCP_CLOSE;
+                       if (!sock_flag(sk, SOCK_DEAD)) {
+                               sock_set_flag(sk, SOCK_DEAD);
+                               sk->sk_state_change(sk);
+                       }
+               }
+               kfree_skb(skb);
+               sock_put(sk);
+               break;
+       case LLC_RESET_PRIM:
+               /*
+                * FIXME:
+                * RESET is not being notified to upper layers for now
+                */
+               printk(KERN_INFO "%s: received a reset ind!\n", __func__);
+               kfree_skb(skb);
+               break;
+       default:
+               if (ev->ind_prim) {
+                       printk(KERN_INFO "%s: received unknown %d prim!\n",
+                               __func__, ev->ind_prim);
+                       kfree_skb(skb);
+               }
+               /* No indication */
+               break;
+       }
+
+       switch (ev->cfm_prim) {
+       case LLC_DATA_PRIM:
+               if (!llc_data_accept_state(llc->state))
+                       sk->sk_write_space(sk);
+               else
+                       rc = llc->failed_data_req = 1;
+               break;
+       case LLC_CONN_PRIM:
+               if (sk->sk_type == SOCK_STREAM &&
+                   sk->sk_state == TCP_SYN_SENT) {
+                       if (ev->status) {
+                               sk->sk_socket->state = SS_UNCONNECTED;
+                               sk->sk_state         = TCP_CLOSE;
+                       } else {
+                               sk->sk_socket->state = SS_CONNECTED;
+                               sk->sk_state         = TCP_ESTABLISHED;
+                       }
+                       sk->sk_state_change(sk);
+               }
+               break;
+       case LLC_DISC_PRIM:
+               sock_hold(sk);
+               if (sk->sk_type == SOCK_STREAM && sk->sk_state == TCP_CLOSING) {
+                       sk->sk_socket->state = SS_UNCONNECTED;
+                       sk->sk_state         = TCP_CLOSE;
+                       sk->sk_state_change(sk);
+               }
+               sock_put(sk);
+               break;
+       case LLC_RESET_PRIM:
+               /*
+                * FIXME:
+                * RESET is not being notified to upper layers for now
+                */
+               printk(KERN_INFO "%s: received a reset conf!\n", __func__);
+               break;
+       default:
+               if (ev->cfm_prim) {
+                       printk(KERN_INFO "%s: received unknown %d prim!\n",
+                                       __func__, ev->cfm_prim);
+                       break;
+               }
+               goto out_skb_put; /* No confirmation */
+       }
+out_kfree_skb:
+       kfree_skb(skb);
+out_skb_put:
+       kfree_skb(skb);
+       return rc;
+}
+
+void llc_conn_send_pdu(struct sock *sk, struct sk_buff *skb)
+{
+       /* queue PDU to send to MAC layer */
+       skb_queue_tail(&sk->sk_write_queue, skb);
+       llc_conn_send_pdus(sk);
+}
+
+/**
+ *     llc_conn_rtn_pdu - sends received data pdu to upper layer
+ *     @sk: Active connection
+ *     @skb: Received data frame
+ *
+ *     Sends received data pdu to upper layer (by using indicate function).
+ *     Prepares service parameters (prim and prim_data). calling indication
+ *     function will be done in llc_conn_state_process.
+ */
+void llc_conn_rtn_pdu(struct sock *sk, struct sk_buff *skb)
+{
+       struct llc_conn_state_ev *ev = llc_conn_ev(skb);
+
+       ev->ind_prim = LLC_DATA_PRIM;
+}
+
+/**
+ *     llc_conn_resend_i_pdu_as_cmd - resend all all unacknowledged I PDUs
+ *     @sk: active connection
+ *     @nr: NR
+ *     @first_p_bit: p_bit value of first pdu
+ *
+ *     Resend all unacknowledged I PDUs, starting with the NR; send first as
+ *     command PDU with P bit equal first_p_bit; if more than one send
+ *     subsequent as command PDUs with P bit equal zero (0).
+ */
+void llc_conn_resend_i_pdu_as_cmd(struct sock *sk, u8 nr, u8 first_p_bit)
+{
+       struct sk_buff *skb;
+       struct llc_pdu_sn *pdu;
+       u16 nbr_unack_pdus;
+       struct llc_sock *llc;
+       u8 howmany_resend = 0;
+
+       llc_conn_remove_acked_pdus(sk, nr, &nbr_unack_pdus);
+       if (!nbr_unack_pdus)
+               goto out;
+       /*
+        * Process unack PDUs only if unack queue is not empty; remove
+        * appropriate PDUs, fix them up, and put them on mac_pdu_q.
+        */
+       llc = llc_sk(sk);
+
+       while ((skb = skb_dequeue(&llc->pdu_unack_q)) != NULL) {
+               pdu = llc_pdu_sn_hdr(skb);
+               llc_pdu_set_cmd_rsp(skb, LLC_PDU_CMD);
+               llc_pdu_set_pf_bit(skb, first_p_bit);
+               skb_queue_tail(&sk->sk_write_queue, skb);
+               first_p_bit = 0;
+               llc->vS = LLC_I_GET_NS(pdu);
+               howmany_resend++;
+       }
+       if (howmany_resend > 0)
+               llc->vS = (llc->vS + 1) % LLC_2_SEQ_NBR_MODULO;
+       /* any PDUs to re-send are queued up; start sending to MAC */
+       llc_conn_send_pdus(sk);
+out:;
+}
+
+/**
+ *     llc_conn_resend_i_pdu_as_rsp - Resend all unacknowledged I PDUs
+ *     @sk: active connection.
+ *     @nr: NR
+ *     @first_f_bit: f_bit value of first pdu.
+ *
+ *     Resend all unacknowledged I PDUs, starting with the NR; send first as
+ *     response PDU with F bit equal first_f_bit; if more than one send
+ *     subsequent as response PDUs with F bit equal zero (0).
+ */
+void llc_conn_resend_i_pdu_as_rsp(struct sock *sk, u8 nr, u8 first_f_bit)
+{
+       struct sk_buff *skb;
+       u16 nbr_unack_pdus;
+       struct llc_sock *llc = llc_sk(sk);
+       u8 howmany_resend = 0;
+
+       llc_conn_remove_acked_pdus(sk, nr, &nbr_unack_pdus);
+       if (!nbr_unack_pdus)
+               goto out;
+       /*
+        * Process unack PDUs only if unack queue is not empty; remove
+        * appropriate PDUs, fix them up, and put them on mac_pdu_q
+        */
+       while ((skb = skb_dequeue(&llc->pdu_unack_q)) != NULL) {
+               struct llc_pdu_sn *pdu = llc_pdu_sn_hdr(skb);
+
+               llc_pdu_set_cmd_rsp(skb, LLC_PDU_RSP);
+               llc_pdu_set_pf_bit(skb, first_f_bit);
+               skb_queue_tail(&sk->sk_write_queue, skb);
+               first_f_bit = 0;
+               llc->vS = LLC_I_GET_NS(pdu);
+               howmany_resend++;
+       }
+       if (howmany_resend > 0)
+               llc->vS = (llc->vS + 1) % LLC_2_SEQ_NBR_MODULO;
+       /* any PDUs to re-send are queued up; start sending to MAC */
+       llc_conn_send_pdus(sk);
+out:;
+}
+
+/**
+ *     llc_conn_remove_acked_pdus - Removes acknowledged pdus from tx queue
+ *     @sk: active connection
+ *     nr: NR
+ *     how_many_unacked: size of pdu_unack_q after removing acked pdus
+ *
+ *     Removes acknowledged pdus from transmit queue (pdu_unack_q). Returns
+ *     the number of pdus that removed from queue.
+ */
+int llc_conn_remove_acked_pdus(struct sock *sk, u8 nr, u16 *how_many_unacked)
+{
+       int pdu_pos, i;
+       struct sk_buff *skb;
+       struct llc_pdu_sn *pdu;
+       int nbr_acked = 0;
+       struct llc_sock *llc = llc_sk(sk);
+       int q_len = skb_queue_len(&llc->pdu_unack_q);
+
+       if (!q_len)
+               goto out;
+       skb = skb_peek(&llc->pdu_unack_q);
+       pdu = llc_pdu_sn_hdr(skb);
+
+       /* finding position of last acked pdu in queue */
+       pdu_pos = ((int)LLC_2_SEQ_NBR_MODULO + (int)nr -
+                       (int)LLC_I_GET_NS(pdu)) % LLC_2_SEQ_NBR_MODULO;
+
+       for (i = 0; i < pdu_pos && i < q_len; i++) {
+               skb = skb_dequeue(&llc->pdu_unack_q);
+               kfree_skb(skb);
+               nbr_acked++;
+       }
+out:
+       *how_many_unacked = skb_queue_len(&llc->pdu_unack_q);
+       return nbr_acked;
+}
+
+/**
+ *     llc_conn_send_pdus - Sends queued PDUs
+ *     @sk: active connection
+ *
+ *     Sends queued pdus to MAC layer for transmission.
+ */
+static void llc_conn_send_pdus(struct sock *sk)
+{
+       struct sk_buff *skb;
+
+       while ((skb = skb_dequeue(&sk->sk_write_queue)) != NULL) {
+               struct llc_pdu_sn *pdu = llc_pdu_sn_hdr(skb);
+
+               if (LLC_PDU_TYPE_IS_I(pdu) &&
+                   !(skb->dev->flags & IFF_LOOPBACK)) {
+                       struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
+
+                       skb_queue_tail(&llc_sk(sk)->pdu_unack_q, skb);
+                       if (!skb2)
+                               break;
+                       skb = skb2;
+               }
+               dev_queue_xmit(skb);
+       }
+}
+
+/**
+ *     llc_conn_service - finds transition and changes state of connection
+ *     @sk: connection
+ *     @skb: happened event
+ *
+ *     This function finds transition that matches with happened event, then
+ *     executes related actions and finally changes state of connection.
+ *     Returns 0 for success, 1 for failure.
+ */
+static int llc_conn_service(struct sock *sk, struct sk_buff *skb)
+{
+       int rc = 1;
+       struct llc_sock *llc = llc_sk(sk);
+       struct llc_conn_state_trans *trans;
+
+       if (llc->state > NBR_CONN_STATES)
+               goto out;
+       rc = 0;
+       trans = llc_qualify_conn_ev(sk, skb);
+       if (trans) {
+               rc = llc_exec_conn_trans_actions(sk, trans, skb);
+               if (!rc && trans->next_state != NO_STATE_CHANGE) {
+                       llc->state = trans->next_state;
+                       if (!llc_data_accept_state(llc->state))
+                               sk->sk_state_change(sk);
+               }
+       }
+out:
+       return rc;
+}
+
+/**
+ *     llc_qualify_conn_ev - finds transition for event
+ *     @sk: connection
+ *     @skb: happened event
+ *
+ *     This function finds transition that matches with happened event.
+ *     Returns pointer to found transition on success, %NULL otherwise.
+ */
+static struct llc_conn_state_trans *llc_qualify_conn_ev(struct sock *sk,
+                                                       struct sk_buff *skb)
+{
+       struct llc_conn_state_trans **next_trans;
+       const llc_conn_ev_qfyr_t *next_qualifier;
+       struct llc_conn_state_ev *ev = llc_conn_ev(skb);
+       struct llc_sock *llc = llc_sk(sk);
+       struct llc_conn_state *curr_state =
+                                       &llc_conn_state_table[llc->state - 1];
+
+       /* search thru events for this state until
+        * list exhausted or until no more
+        */
+       for (next_trans = curr_state->transitions +
+               llc_find_offset(llc->state - 1, ev->type);
+            (*next_trans)->ev; next_trans++) {
+               if (!((*next_trans)->ev)(sk, skb)) {
+                       /* got POSSIBLE event match; the event may require
+                        * qualification based on the values of a number of
+                        * state flags; if all qualifications are met (i.e.,
+                        * if all qualifying functions return success, or 0,
+                        * then this is THE event we're looking for
+                        */
+                       for (next_qualifier = (*next_trans)->ev_qualifiers;
+                            next_qualifier && *next_qualifier &&
+                            !(*next_qualifier)(sk, skb); next_qualifier++)
+                               /* nothing */;
+                       if (!next_qualifier || !*next_qualifier)
+                               /* all qualifiers executed successfully; this is
+                                * our transition; return it so we can perform
+                                * the associated actions & change the state
+                                */
+                               return *next_trans;
+               }
+       }
+       return NULL;
+}
+
+/**
+ *     llc_exec_conn_trans_actions - executes related actions
+ *     @sk: connection
+ *     @trans: transition that it's actions must be performed
+ *     @skb: event
+ *
+ *     Executes actions that is related to happened event. Returns 0 for
+ *     success, 1 to indicate failure of at least one action.
+ */
+static int llc_exec_conn_trans_actions(struct sock *sk,
+                                      struct llc_conn_state_trans *trans,
+                                      struct sk_buff *skb)
+{
+       int rc = 0;
+       const llc_conn_action_t *next_action;
+
+       for (next_action = trans->ev_actions;
+            next_action && *next_action; next_action++) {
+               int rc2 = (*next_action)(sk, skb);
+
+               if (rc2 == 2) {
+                       rc = rc2;
+                       break;
+               } else if (rc2)
+                       rc = 1;
+       }
+       return rc;
+}
+
+static inline bool llc_estab_match(const struct llc_sap *sap,
+                                  const struct llc_addr *daddr,
+                                  const struct llc_addr *laddr,
+                                  const struct sock *sk)
+{
+       struct llc_sock *llc = llc_sk(sk);
+
+       return llc->laddr.lsap == laddr->lsap &&
+               llc->daddr.lsap == daddr->lsap &&
+               ether_addr_equal(llc->laddr.mac, laddr->mac) &&
+               ether_addr_equal(llc->daddr.mac, daddr->mac);
+}
+
+/**
+ *     __llc_lookup_established - Finds connection for the remote/local sap/mac
+ *     @sap: SAP
+ *     @daddr: address of remote LLC (MAC + SAP)
+ *     @laddr: address of local LLC (MAC + SAP)
+ *
+ *     Search connection list of the SAP and finds connection using the remote
+ *     mac, remote sap, local mac, and local sap. Returns pointer for
+ *     connection found, %NULL otherwise.
+ *     Caller has to make sure local_bh is disabled.
+ */
+static struct sock *__llc_lookup_established(struct llc_sap *sap,
+                                            struct llc_addr *daddr,
+                                            struct llc_addr *laddr)
+{
+       struct sock *rc;
+       struct hlist_nulls_node *node;
+       int slot = llc_sk_laddr_hashfn(sap, laddr);
+       struct hlist_nulls_head *laddr_hb = &sap->sk_laddr_hash[slot];
+
+       rcu_read_lock();
+again:
+       sk_nulls_for_each_rcu(rc, node, laddr_hb) {
+               if (llc_estab_match(sap, daddr, laddr, rc)) {
+                       /* Extra checks required by SLAB_DESTROY_BY_RCU */
+                       if (unlikely(!atomic_inc_not_zero(&rc->sk_refcnt)))
+                               goto again;
+                       if (unlikely(llc_sk(rc)->sap != sap ||
+                                    !llc_estab_match(sap, daddr, laddr, rc))) {
+                               sock_put(rc);
+                               continue;
+                       }
+                       goto found;
+               }
+       }
+       rc = NULL;
+       /*
+        * if the nulls value we got at the end of this lookup is
+        * not the expected one, we must restart lookup.
+        * We probably met an item that was moved to another chain.
+        */
+       if (unlikely(get_nulls_value(node) != slot))
+               goto again;
+found:
+       rcu_read_unlock();
+       return rc;
+}
+
+struct sock *llc_lookup_established(struct llc_sap *sap,
+                                   struct llc_addr *daddr,
+                                   struct llc_addr *laddr)
+{
+       struct sock *sk;
+
+       local_bh_disable();
+       sk = __llc_lookup_established(sap, daddr, laddr);
+       local_bh_enable();
+       return sk;
+}
+
+static inline bool llc_listener_match(const struct llc_sap *sap,
+                                     const struct llc_addr *laddr,
+                                     const struct sock *sk)
+{
+       struct llc_sock *llc = llc_sk(sk);
+
+       return sk->sk_type == SOCK_STREAM && sk->sk_state == TCP_LISTEN &&
+               llc->laddr.lsap == laddr->lsap &&
+               ether_addr_equal(llc->laddr.mac, laddr->mac);
+}
+
+static struct sock *__llc_lookup_listener(struct llc_sap *sap,
+                                         struct llc_addr *laddr)
+{
+       struct sock *rc;
+       struct hlist_nulls_node *node;
+       int slot = llc_sk_laddr_hashfn(sap, laddr);
+       struct hlist_nulls_head *laddr_hb = &sap->sk_laddr_hash[slot];
+
+       rcu_read_lock();
+again:
+       sk_nulls_for_each_rcu(rc, node, laddr_hb) {
+               if (llc_listener_match(sap, laddr, rc)) {
+                       /* Extra checks required by SLAB_DESTROY_BY_RCU */
+                       if (unlikely(!atomic_inc_not_zero(&rc->sk_refcnt)))
+                               goto again;
+                       if (unlikely(llc_sk(rc)->sap != sap ||
+                                    !llc_listener_match(sap, laddr, rc))) {
+                               sock_put(rc);
+                               continue;
+                       }
+                       goto found;
+               }
+       }
+       rc = NULL;
+       /*
+        * if the nulls value we got at the end of this lookup is
+        * not the expected one, we must restart lookup.
+        * We probably met an item that was moved to another chain.
+        */
+       if (unlikely(get_nulls_value(node) != slot))
+               goto again;
+found:
+       rcu_read_unlock();
+       return rc;
+}
+
+/**
+ *     llc_lookup_listener - Finds listener for local MAC + SAP
+ *     @sap: SAP
+ *     @laddr: address of local LLC (MAC + SAP)
+ *
+ *     Search connection list of the SAP and finds connection listening on
+ *     local mac, and local sap. Returns pointer for parent socket found,
+ *     %NULL otherwise.
+ *     Caller has to make sure local_bh is disabled.
+ */
+static struct sock *llc_lookup_listener(struct llc_sap *sap,
+                                       struct llc_addr *laddr)
+{
+       static struct llc_addr null_addr;
+       struct sock *rc = __llc_lookup_listener(sap, laddr);
+
+       if (!rc)
+               rc = __llc_lookup_listener(sap, &null_addr);
+
+       return rc;
+}
+
+static struct sock *__llc_lookup(struct llc_sap *sap,
+                                struct llc_addr *daddr,
+                                struct llc_addr *laddr)
+{
+       struct sock *sk = __llc_lookup_established(sap, daddr, laddr);
+
+       return sk ? : llc_lookup_listener(sap, laddr);
+}
+
+/**
+ *     llc_data_accept_state - designates if in this state data can be sent.
+ *     @state: state of connection.
+ *
+ *     Returns 0 if data can be sent, 1 otherwise.
+ */
+u8 llc_data_accept_state(u8 state)
+{
+       return state != LLC_CONN_STATE_NORMAL && state != LLC_CONN_STATE_BUSY &&
+              state != LLC_CONN_STATE_REJ;
+}
+
+/**
+ *     llc_find_next_offset - finds offset for next category of transitions
+ *     @state: state table.
+ *     @offset: start offset.
+ *
+ *     Finds offset of next category of transitions in transition table.
+ *     Returns the start index of next category.
+ */
+static u16 __init llc_find_next_offset(struct llc_conn_state *state, u16 offset)
+{
+       u16 cnt = 0;
+       struct llc_conn_state_trans **next_trans;
+
+       for (next_trans = state->transitions + offset;
+            (*next_trans)->ev; next_trans++)
+               ++cnt;
+       return cnt;
+}
+
+/**
+ *     llc_build_offset_table - builds offset table of connection
+ *
+ *     Fills offset table of connection state transition table
+ *     (llc_offset_table).
+ */
+void __init llc_build_offset_table(void)
+{
+       struct llc_conn_state *curr_state;
+       int state, ev_type, next_offset;
+
+       for (state = 0; state < NBR_CONN_STATES; state++) {
+               curr_state = &llc_conn_state_table[state];
+               next_offset = 0;
+               for (ev_type = 0; ev_type < NBR_CONN_EV; ev_type++) {
+                       llc_offset_table[state][ev_type] = next_offset;
+                       next_offset += llc_find_next_offset(curr_state,
+                                                           next_offset) + 1;
+               }
+       }
+}
+
+/**
+ *     llc_find_offset - finds start offset of category of transitions
+ *     @state: state of connection
+ *     @ev_type: type of happened event
+ *
+ *     Finds start offset of desired category of transitions. Returns the
+ *     desired start offset.
+ */
+static int llc_find_offset(int state, int ev_type)
+{
+       int rc = 0;
+       /* at this stage, llc_offset_table[..][2] is not important. it is for
+        * init_pf_cycle and I don't know what is it.
+        */
+       switch (ev_type) {
+       case LLC_CONN_EV_TYPE_PRIM:
+               rc = llc_offset_table[state][0]; break;
+       case LLC_CONN_EV_TYPE_PDU:
+               rc = llc_offset_table[state][4]; break;
+       case LLC_CONN_EV_TYPE_SIMPLE:
+               rc = llc_offset_table[state][1]; break;
+       case LLC_CONN_EV_TYPE_P_TMR:
+       case LLC_CONN_EV_TYPE_ACK_TMR:
+       case LLC_CONN_EV_TYPE_REJ_TMR:
+       case LLC_CONN_EV_TYPE_BUSY_TMR:
+               rc = llc_offset_table[state][3]; break;
+       }
+       return rc;
+}
+
+/**
+ *     llc_sap_add_socket - adds a socket to a SAP
+ *     @sap: SAP
+ *     @sk: socket
+ *
+ *     This function adds a socket to the hash tables of a SAP.
+ */
+void llc_sap_add_socket(struct llc_sap *sap, struct sock *sk)
+{
+       struct llc_sock *llc = llc_sk(sk);
+       struct hlist_head *dev_hb = llc_sk_dev_hash(sap, llc->dev->ifindex);
+       struct hlist_nulls_head *laddr_hb = llc_sk_laddr_hash(sap, &llc->laddr);
+
+       llc_sap_hold(sap);
+       llc_sk(sk)->sap = sap;
+
+       spin_lock_bh(&sap->sk_lock);
+       sap->sk_count++;
+       sk_nulls_add_node_rcu(sk, laddr_hb);
+       hlist_add_head(&llc->dev_hash_node, dev_hb);
+       spin_unlock_bh(&sap->sk_lock);
+}
+
+/**
+ *     llc_sap_remove_socket - removes a socket from SAP
+ *     @sap: SAP
+ *     @sk: socket
+ *
+ *     This function removes a connection from the hash tables of a SAP if
+ *     the connection was in this list.
+ */
+void llc_sap_remove_socket(struct llc_sap *sap, struct sock *sk)
+{
+       struct llc_sock *llc = llc_sk(sk);
+
+       spin_lock_bh(&sap->sk_lock);
+       sk_nulls_del_node_init_rcu(sk);
+       hlist_del(&llc->dev_hash_node);
+       sap->sk_count--;
+       spin_unlock_bh(&sap->sk_lock);
+       llc_sap_put(sap);
+}
+
+/**
+ *     llc_conn_rcv - sends received pdus to the connection state machine
+ *     @sk: current connection structure.
+ *     @skb: received frame.
+ *
+ *     Sends received pdus to the connection state machine.
+ */
+static int llc_conn_rcv(struct sock *sk, struct sk_buff *skb)
+{
+       struct llc_conn_state_ev *ev = llc_conn_ev(skb);
+
+       ev->type   = LLC_CONN_EV_TYPE_PDU;
+       ev->reason = 0;
+       return llc_conn_state_process(sk, skb);
+}
+
+static struct sock *llc_create_incoming_sock(struct sock *sk,
+                                            struct net_device *dev,
+                                            struct llc_addr *saddr,
+                                            struct llc_addr *daddr)
+{
+       struct sock *newsk = llc_sk_alloc(sock_net(sk), sk->sk_family, GFP_ATOMIC,
+                                         sk->sk_prot);
+       struct llc_sock *newllc, *llc = llc_sk(sk);
+
+       if (!newsk)
+               goto out;
+       newllc = llc_sk(newsk);
+       memcpy(&newllc->laddr, daddr, sizeof(newllc->laddr));
+       memcpy(&newllc->daddr, saddr, sizeof(newllc->daddr));
+       newllc->dev = dev;
+       dev_hold(dev);
+       llc_sap_add_socket(llc->sap, newsk);
+       llc_sap_hold(llc->sap);
+out:
+       return newsk;
+}
+
+void llc_conn_handler(struct llc_sap *sap, struct sk_buff *skb)
+{
+       struct llc_addr saddr, daddr;
+       struct sock *sk;
+
+       llc_pdu_decode_sa(skb, saddr.mac);
+       llc_pdu_decode_ssap(skb, &saddr.lsap);
+       llc_pdu_decode_da(skb, daddr.mac);
+       llc_pdu_decode_dsap(skb, &daddr.lsap);
+
+       sk = __llc_lookup(sap, &saddr, &daddr);
+       if (!sk)
+               goto drop;
+
+       bh_lock_sock(sk);
+       /*
+        * This has to be done here and not at the upper layer ->accept
+        * method because of the way the PROCOM state machine works:
+        * it needs to set several state variables (see, for instance,
+        * llc_adm_actions_2 in net/llc/llc_c_st.c) and send a packet to
+        * the originator of the new connection, and this state has to be
+        * in the newly created struct sock private area. -acme
+        */
+       if (unlikely(sk->sk_state == TCP_LISTEN)) {
+               struct sock *newsk = llc_create_incoming_sock(sk, skb->dev,
+                                                             &saddr, &daddr);
+               if (!newsk)
+                       goto drop_unlock;
+               skb_set_owner_r(skb, newsk);
+       } else {
+               /*
+                * Can't be skb_set_owner_r, this will be done at the
+                * llc_conn_state_process function, later on, when we will use
+                * skb_queue_rcv_skb to send it to upper layers, this is
+                * another trick required to cope with how the PROCOM state
+                * machine works. -acme
+                */
+               skb->sk = sk;
+       }
+       if (!sock_owned_by_user(sk))
+               llc_conn_rcv(sk, skb);
+       else {
+               dprintk("%s: adding to backlog...\n", __func__);
+               llc_set_backlog_type(skb, LLC_PACKET);
+               if (sk_add_backlog(sk, skb, sk->sk_rcvbuf))
+                       goto drop_unlock;
+       }
+out:
+       bh_unlock_sock(sk);
+       sock_put(sk);
+       return;
+drop:
+       kfree_skb(skb);
+       return;
+drop_unlock:
+       kfree_skb(skb);
+       goto out;
+}
+
+#undef LLC_REFCNT_DEBUG
+#ifdef LLC_REFCNT_DEBUG
+static atomic_t llc_sock_nr;
+#endif
+
+/**
+ *     llc_backlog_rcv - Processes rx frames and expired timers.
+ *     @sk: LLC sock (p8022 connection)
+ *     @skb: queued rx frame or event
+ *
+ *     This function processes frames that has received and timers that has
+ *     expired during sending an I pdu (refer to data_req_handler).  frames
+ *     queue by llc_rcv function (llc_mac.c) and timers queue by timer
+ *     callback functions(llc_c_ac.c).
+ */
+static int llc_backlog_rcv(struct sock *sk, struct sk_buff *skb)
+{
+       int rc = 0;
+       struct llc_sock *llc = llc_sk(sk);
+
+       if (likely(llc_backlog_type(skb) == LLC_PACKET)) {
+               if (likely(llc->state > 1)) /* not closed */
+                       rc = llc_conn_rcv(sk, skb);
+               else
+                       goto out_kfree_skb;
+       } else if (llc_backlog_type(skb) == LLC_EVENT) {
+               /* timer expiration event */
+               if (likely(llc->state > 1))  /* not closed */
+                       rc = llc_conn_state_process(sk, skb);
+               else
+                       goto out_kfree_skb;
+       } else {
+               printk(KERN_ERR "%s: invalid skb in backlog\n", __func__);
+               goto out_kfree_skb;
+       }
+out:
+       return rc;
+out_kfree_skb:
+       kfree_skb(skb);
+       goto out;
+}
+
+/**
+ *     llc_sk_init - Initializes a socket with default llc values.
+ *     @sk: socket to initialize.
+ *
+ *     Initializes a socket with default llc values.
+ */
+static void llc_sk_init(struct sock *sk)
+{
+       struct llc_sock *llc = llc_sk(sk);
+
+       llc->state    = LLC_CONN_STATE_ADM;
+       llc->inc_cntr = llc->dec_cntr = 2;
+       llc->dec_step = llc->connect_step = 1;
+
+       setup_timer(&llc->ack_timer.timer, llc_conn_ack_tmr_cb,
+                       (unsigned long)sk);
+       llc->ack_timer.expire         = sysctl_llc2_ack_timeout;
+
+       setup_timer(&llc->pf_cycle_timer.timer, llc_conn_pf_cycle_tmr_cb,
+                       (unsigned long)sk);
+       llc->pf_cycle_timer.expire         = sysctl_llc2_p_timeout;
+
+       setup_timer(&llc->rej_sent_timer.timer, llc_conn_rej_tmr_cb,
+                       (unsigned long)sk);
+       llc->rej_sent_timer.expire         = sysctl_llc2_rej_timeout;
+
+       setup_timer(&llc->busy_state_timer.timer, llc_conn_busy_tmr_cb,
+                       (unsigned long)sk);
+       llc->busy_state_timer.expire         = sysctl_llc2_busy_timeout;
+
+       llc->n2 = 2;   /* max retransmit */
+       llc->k  = 2;   /* tx win size, will adjust dynam */
+       llc->rw = 128; /* rx win size (opt and equal to
+                       * tx_win of remote LLC) */
+       skb_queue_head_init(&llc->pdu_unack_q);
+       sk->sk_backlog_rcv = llc_backlog_rcv;
+}
+
+/**
+ *     llc_sk_alloc - Allocates LLC sock
+ *     @family: upper layer protocol family
+ *     @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
+ *
+ *     Allocates a LLC sock and initializes it. Returns the new LLC sock
+ *     or %NULL if there's no memory available for one
+ */
+struct sock *llc_sk_alloc(struct net *net, int family, gfp_t priority, struct proto *prot)
+{
+       struct sock *sk = sk_alloc(net, family, priority, prot);
+
+       if (!sk)
+               goto out;
+       llc_sk_init(sk);
+       sock_init_data(NULL, sk);
+#ifdef LLC_REFCNT_DEBUG
+       atomic_inc(&llc_sock_nr);
+       printk(KERN_DEBUG "LLC socket %p created in %s, now we have %d alive\n", sk,
+               __func__, atomic_read(&llc_sock_nr));
+#endif
+out:
+       return sk;
+}
+
+/**
+ *     llc_sk_free - Frees a LLC socket
+ *     @sk - socket to free
+ *
+ *     Frees a LLC socket
+ */
+void llc_sk_free(struct sock *sk)
+{
+       struct llc_sock *llc = llc_sk(sk);
+
+       llc->state = LLC_CONN_OUT_OF_SVC;
+       /* Stop all (possibly) running timers */
+       llc_conn_ac_stop_all_timers(sk, NULL);
+#ifdef DEBUG_LLC_CONN_ALLOC
+       printk(KERN_INFO "%s: unackq=%d, txq=%d\n", __func__,
+               skb_queue_len(&llc->pdu_unack_q),
+               skb_queue_len(&sk->sk_write_queue));
+#endif
+       skb_queue_purge(&sk->sk_receive_queue);
+       skb_queue_purge(&sk->sk_write_queue);
+       skb_queue_purge(&llc->pdu_unack_q);
+#ifdef LLC_REFCNT_DEBUG
+       if (atomic_read(&sk->sk_refcnt) != 1) {
+               printk(KERN_DEBUG "Destruction of LLC sock %p delayed in %s, cnt=%d\n",
+                       sk, __func__, atomic_read(&sk->sk_refcnt));
+               printk(KERN_DEBUG "%d LLC sockets are still alive\n",
+                       atomic_read(&llc_sock_nr));
+       } else {
+               atomic_dec(&llc_sock_nr);
+               printk(KERN_DEBUG "LLC socket %p released in %s, %d are still alive\n", sk,
+                       __func__, atomic_read(&llc_sock_nr));
+       }
+#endif
+       sock_put(sk);
+}
+
+/**
+ *     llc_sk_reset - resets a connection
+ *     @sk: LLC socket to reset
+ *
+ *     Resets a connection to the out of service state. Stops its timers
+ *     and frees any frames in the queues of the connection.
+ */
+void llc_sk_reset(struct sock *sk)
+{
+       struct llc_sock *llc = llc_sk(sk);
+
+       llc_conn_ac_stop_all_timers(sk, NULL);
+       skb_queue_purge(&sk->sk_write_queue);
+       skb_queue_purge(&llc->pdu_unack_q);
+       llc->remote_busy_flag   = 0;
+       llc->cause_flag         = 0;
+       llc->retry_count        = 0;
+       llc_conn_set_p_flag(sk, 0);
+       llc->f_flag             = 0;
+       llc->s_flag             = 0;
+       llc->ack_pf             = 0;
+       llc->first_pdu_Ns       = 0;
+       llc->ack_must_be_send   = 0;
+       llc->dec_step           = 1;
+       llc->inc_cntr           = 2;
+       llc->dec_cntr           = 2;
+       llc->X                  = 0;
+       llc->failed_data_req    = 0 ;
+       llc->last_nr            = 0;
+}