Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / net / irda / irlap.c
diff --git a/kernel/net/irda/irlap.c b/kernel/net/irda/irlap.c
new file mode 100644 (file)
index 0000000..1cde711
--- /dev/null
@@ -0,0 +1,1207 @@
+/*********************************************************************
+ *
+ * Filename:      irlap.c
+ * Version:       1.0
+ * Description:   IrLAP implementation for Linux
+ * Status:        Stable
+ * Author:        Dag Brattli <dagb@cs.uit.no>
+ * Created at:    Mon Aug  4 20:40:53 1997
+ * Modified at:   Tue Dec 14 09:26:44 1999
+ * Modified by:   Dag Brattli <dagb@cs.uit.no>
+ *
+ *     Copyright (c) 1998-1999 Dag Brattli, All Rights Reserved.
+ *     Copyright (c) 2000-2003 Jean Tourrilhes <jt@hpl.hp.com>
+ *
+ *     This program is free software; you can redistribute it and/or
+ *     modify it under the terms of the GNU General Public License as
+ *     published by the Free Software Foundation; either version 2 of
+ *     the License, or (at your option) any later version.
+ *
+ *     This program is distributed in the hope that it will be useful,
+ *     but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ *     GNU General Public License for more details.
+ *
+ *     You should have received a copy of the GNU General Public License
+ *     along with this program; if not, see <http://www.gnu.org/licenses/>.
+ *
+ ********************************************************************/
+
+#include <linux/slab.h>
+#include <linux/string.h>
+#include <linux/skbuff.h>
+#include <linux/delay.h>
+#include <linux/proc_fs.h>
+#include <linux/init.h>
+#include <linux/random.h>
+#include <linux/module.h>
+#include <linux/seq_file.h>
+
+#include <net/irda/irda.h>
+#include <net/irda/irda_device.h>
+#include <net/irda/irqueue.h>
+#include <net/irda/irlmp.h>
+#include <net/irda/irlmp_frame.h>
+#include <net/irda/irlap_frame.h>
+#include <net/irda/irlap.h>
+#include <net/irda/timer.h>
+#include <net/irda/qos.h>
+
+static hashbin_t *irlap = NULL;
+int sysctl_slot_timeout = SLOT_TIMEOUT * 1000 / HZ;
+
+/* This is the delay of missed pf period before generating an event
+ * to the application. The spec mandate 3 seconds, but in some cases
+ * it's way too long. - Jean II */
+int sysctl_warn_noreply_time = 3;
+
+extern void irlap_queue_xmit(struct irlap_cb *self, struct sk_buff *skb);
+static void __irlap_close(struct irlap_cb *self);
+static void irlap_init_qos_capabilities(struct irlap_cb *self,
+                                       struct qos_info *qos_user);
+
+static const char *const lap_reasons[] __maybe_unused = {
+       "ERROR, NOT USED",
+       "LAP_DISC_INDICATION",
+       "LAP_NO_RESPONSE",
+       "LAP_RESET_INDICATION",
+       "LAP_FOUND_NONE",
+       "LAP_MEDIA_BUSY",
+       "LAP_PRIMARY_CONFLICT",
+       "ERROR, NOT USED",
+};
+
+int __init irlap_init(void)
+{
+       /* Check if the compiler did its job properly.
+        * May happen on some ARM configuration, check with Russell King. */
+       IRDA_ASSERT(sizeof(struct xid_frame) == 14, ;);
+       IRDA_ASSERT(sizeof(struct test_frame) == 10, ;);
+       IRDA_ASSERT(sizeof(struct ua_frame) == 10, ;);
+       IRDA_ASSERT(sizeof(struct snrm_frame) == 11, ;);
+
+       /* Allocate master array */
+       irlap = hashbin_new(HB_LOCK);
+       if (irlap == NULL) {
+               net_err_ratelimited("%s: can't allocate irlap hashbin!\n",
+                                   __func__);
+               return -ENOMEM;
+       }
+
+       return 0;
+}
+
+void irlap_cleanup(void)
+{
+       IRDA_ASSERT(irlap != NULL, return;);
+
+       hashbin_delete(irlap, (FREE_FUNC) __irlap_close);
+}
+
+/*
+ * Function irlap_open (driver)
+ *
+ *    Initialize IrLAP layer
+ *
+ */
+struct irlap_cb *irlap_open(struct net_device *dev, struct qos_info *qos,
+                           const char *hw_name)
+{
+       struct irlap_cb *self;
+
+       /* Initialize the irlap structure. */
+       self = kzalloc(sizeof(struct irlap_cb), GFP_KERNEL);
+       if (self == NULL)
+               return NULL;
+
+       self->magic = LAP_MAGIC;
+
+       /* Make a binding between the layers */
+       self->netdev = dev;
+       self->qos_dev = qos;
+       /* Copy hardware name */
+       if(hw_name != NULL) {
+               strlcpy(self->hw_name, hw_name, sizeof(self->hw_name));
+       } else {
+               self->hw_name[0] = '\0';
+       }
+
+       /* FIXME: should we get our own field? */
+       dev->atalk_ptr = self;
+
+       self->state = LAP_OFFLINE;
+
+       /* Initialize transmit queue */
+       skb_queue_head_init(&self->txq);
+       skb_queue_head_init(&self->txq_ultra);
+       skb_queue_head_init(&self->wx_list);
+
+       /* My unique IrLAP device address! */
+       /* We don't want the broadcast address, neither the NULL address
+        * (most often used to signify "invalid"), and we don't want an
+        * address already in use (otherwise connect won't be able
+        * to select the proper link). - Jean II */
+       do {
+               get_random_bytes(&self->saddr, sizeof(self->saddr));
+       } while ((self->saddr == 0x0) || (self->saddr == BROADCAST) ||
+                (hashbin_lock_find(irlap, self->saddr, NULL)) );
+       /* Copy to the driver */
+       memcpy(dev->dev_addr, &self->saddr, 4);
+
+       init_timer(&self->slot_timer);
+       init_timer(&self->query_timer);
+       init_timer(&self->discovery_timer);
+       init_timer(&self->final_timer);
+       init_timer(&self->poll_timer);
+       init_timer(&self->wd_timer);
+       init_timer(&self->backoff_timer);
+       init_timer(&self->media_busy_timer);
+
+       irlap_apply_default_connection_parameters(self);
+
+       self->N3 = 3; /* # connections attempts to try before giving up */
+
+       self->state = LAP_NDM;
+
+       hashbin_insert(irlap, (irda_queue_t *) self, self->saddr, NULL);
+
+       irlmp_register_link(self, self->saddr, &self->notify);
+
+       return self;
+}
+EXPORT_SYMBOL(irlap_open);
+
+/*
+ * Function __irlap_close (self)
+ *
+ *    Remove IrLAP and all allocated memory. Stop any pending timers.
+ *
+ */
+static void __irlap_close(struct irlap_cb *self)
+{
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       /* Stop timers */
+       del_timer(&self->slot_timer);
+       del_timer(&self->query_timer);
+       del_timer(&self->discovery_timer);
+       del_timer(&self->final_timer);
+       del_timer(&self->poll_timer);
+       del_timer(&self->wd_timer);
+       del_timer(&self->backoff_timer);
+       del_timer(&self->media_busy_timer);
+
+       irlap_flush_all_queues(self);
+
+       self->magic = 0;
+
+       kfree(self);
+}
+
+/*
+ * Function irlap_close (self)
+ *
+ *    Remove IrLAP instance
+ *
+ */
+void irlap_close(struct irlap_cb *self)
+{
+       struct irlap_cb *lap;
+
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       /* We used to send a LAP_DISC_INDICATION here, but this was
+        * racy. This has been move within irlmp_unregister_link()
+        * itself. Jean II */
+
+       /* Kill the LAP and all LSAPs on top of it */
+       irlmp_unregister_link(self->saddr);
+       self->notify.instance = NULL;
+
+       /* Be sure that we manage to remove ourself from the hash */
+       lap = hashbin_remove(irlap, self->saddr, NULL);
+       if (!lap) {
+               pr_debug("%s(), Didn't find myself!\n", __func__);
+               return;
+       }
+       __irlap_close(lap);
+}
+EXPORT_SYMBOL(irlap_close);
+
+/*
+ * Function irlap_connect_indication (self, skb)
+ *
+ *    Another device is attempting to make a connection
+ *
+ */
+void irlap_connect_indication(struct irlap_cb *self, struct sk_buff *skb)
+{
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       irlap_init_qos_capabilities(self, NULL); /* No user QoS! */
+
+       irlmp_link_connect_indication(self->notify.instance, self->saddr,
+                                     self->daddr, &self->qos_tx, skb);
+}
+
+/*
+ * Function irlap_connect_response (self, skb)
+ *
+ *    Service user has accepted incoming connection
+ *
+ */
+void irlap_connect_response(struct irlap_cb *self, struct sk_buff *userdata)
+{
+       irlap_do_event(self, CONNECT_RESPONSE, userdata, NULL);
+}
+
+/*
+ * Function irlap_connect_request (self, daddr, qos_user, sniff)
+ *
+ *    Request connection with another device, sniffing is not implemented
+ *    yet.
+ *
+ */
+void irlap_connect_request(struct irlap_cb *self, __u32 daddr,
+                          struct qos_info *qos_user, int sniff)
+{
+       pr_debug("%s(), daddr=0x%08x\n", __func__, daddr);
+
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       self->daddr = daddr;
+
+       /*
+        *  If the service user specifies QoS values for this connection,
+        *  then use them
+        */
+       irlap_init_qos_capabilities(self, qos_user);
+
+       if ((self->state == LAP_NDM) && !self->media_busy)
+               irlap_do_event(self, CONNECT_REQUEST, NULL, NULL);
+       else
+               self->connect_pending = TRUE;
+}
+
+/*
+ * Function irlap_connect_confirm (self, skb)
+ *
+ *    Connection request has been accepted
+ *
+ */
+void irlap_connect_confirm(struct irlap_cb *self, struct sk_buff *skb)
+{
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       irlmp_link_connect_confirm(self->notify.instance, &self->qos_tx, skb);
+}
+
+/*
+ * Function irlap_data_indication (self, skb)
+ *
+ *    Received data frames from IR-port, so we just pass them up to
+ *    IrLMP for further processing
+ *
+ */
+void irlap_data_indication(struct irlap_cb *self, struct sk_buff *skb,
+                          int unreliable)
+{
+       /* Hide LAP header from IrLMP layer */
+       skb_pull(skb, LAP_ADDR_HEADER+LAP_CTRL_HEADER);
+
+       irlmp_link_data_indication(self->notify.instance, skb, unreliable);
+}
+
+
+/*
+ * Function irlap_data_request (self, skb)
+ *
+ *    Queue data for transmission, must wait until XMIT state
+ *
+ */
+void irlap_data_request(struct irlap_cb *self, struct sk_buff *skb,
+                       int unreliable)
+{
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       IRDA_ASSERT(skb_headroom(skb) >= (LAP_ADDR_HEADER+LAP_CTRL_HEADER),
+                   return;);
+       skb_push(skb, LAP_ADDR_HEADER+LAP_CTRL_HEADER);
+
+       /*
+        *  Must set frame format now so that the rest of the code knows
+        *  if its dealing with an I or an UI frame
+        */
+       if (unreliable)
+               skb->data[1] = UI_FRAME;
+       else
+               skb->data[1] = I_FRAME;
+
+       /* Don't forget to refcount it - see irlmp_connect_request(). */
+       skb_get(skb);
+
+       /* Add at the end of the queue (keep ordering) - Jean II */
+       skb_queue_tail(&self->txq, skb);
+
+       /*
+        *  Send event if this frame only if we are in the right state
+        *  FIXME: udata should be sent first! (skb_queue_head?)
+        */
+       if ((self->state == LAP_XMIT_P) || (self->state == LAP_XMIT_S)) {
+               /* If we are not already processing the Tx queue, trigger
+                * transmission immediately - Jean II */
+               if((skb_queue_len(&self->txq) <= 1) && (!self->local_busy))
+                       irlap_do_event(self, DATA_REQUEST, skb, NULL);
+               /* Otherwise, the packets will be sent normally at the
+                * next pf-poll - Jean II */
+       }
+}
+
+/*
+ * Function irlap_unitdata_request (self, skb)
+ *
+ *    Send Ultra data. This is data that must be sent outside any connection
+ *
+ */
+#ifdef CONFIG_IRDA_ULTRA
+void irlap_unitdata_request(struct irlap_cb *self, struct sk_buff *skb)
+{
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       IRDA_ASSERT(skb_headroom(skb) >= (LAP_ADDR_HEADER+LAP_CTRL_HEADER),
+              return;);
+       skb_push(skb, LAP_ADDR_HEADER+LAP_CTRL_HEADER);
+
+       skb->data[0] = CBROADCAST;
+       skb->data[1] = UI_FRAME;
+
+       /* Don't need to refcount, see irlmp_connless_data_request() */
+
+       skb_queue_tail(&self->txq_ultra, skb);
+
+       irlap_do_event(self, SEND_UI_FRAME, NULL, NULL);
+}
+#endif /*CONFIG_IRDA_ULTRA */
+
+/*
+ * Function irlap_udata_indication (self, skb)
+ *
+ *    Receive Ultra data. This is data that is received outside any connection
+ *
+ */
+#ifdef CONFIG_IRDA_ULTRA
+void irlap_unitdata_indication(struct irlap_cb *self, struct sk_buff *skb)
+{
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+       IRDA_ASSERT(skb != NULL, return;);
+
+       /* Hide LAP header from IrLMP layer */
+       skb_pull(skb, LAP_ADDR_HEADER+LAP_CTRL_HEADER);
+
+       irlmp_link_unitdata_indication(self->notify.instance, skb);
+}
+#endif /* CONFIG_IRDA_ULTRA */
+
+/*
+ * Function irlap_disconnect_request (void)
+ *
+ *    Request to disconnect connection by service user
+ */
+void irlap_disconnect_request(struct irlap_cb *self)
+{
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       /* Don't disconnect until all data frames are successfully sent */
+       if (!skb_queue_empty(&self->txq)) {
+               self->disconnect_pending = TRUE;
+               return;
+       }
+
+       /* Check if we are in the right state for disconnecting */
+       switch (self->state) {
+       case LAP_XMIT_P:        /* FALLTHROUGH */
+       case LAP_XMIT_S:        /* FALLTHROUGH */
+       case LAP_CONN:          /* FALLTHROUGH */
+       case LAP_RESET_WAIT:    /* FALLTHROUGH */
+       case LAP_RESET_CHECK:
+               irlap_do_event(self, DISCONNECT_REQUEST, NULL, NULL);
+               break;
+       default:
+               pr_debug("%s(), disconnect pending!\n", __func__);
+               self->disconnect_pending = TRUE;
+               break;
+       }
+}
+
+/*
+ * Function irlap_disconnect_indication (void)
+ *
+ *    Disconnect request from other device
+ *
+ */
+void irlap_disconnect_indication(struct irlap_cb *self, LAP_REASON reason)
+{
+       pr_debug("%s(), reason=%s\n", __func__, lap_reasons[reason]);
+
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       /* Flush queues */
+       irlap_flush_all_queues(self);
+
+       switch (reason) {
+       case LAP_RESET_INDICATION:
+               pr_debug("%s(), Sending reset request!\n", __func__);
+               irlap_do_event(self, RESET_REQUEST, NULL, NULL);
+               break;
+       case LAP_NO_RESPONSE:      /* FALLTHROUGH */
+       case LAP_DISC_INDICATION:  /* FALLTHROUGH */
+       case LAP_FOUND_NONE:       /* FALLTHROUGH */
+       case LAP_MEDIA_BUSY:
+               irlmp_link_disconnect_indication(self->notify.instance, self,
+                                                reason, NULL);
+               break;
+       default:
+               net_err_ratelimited("%s: Unknown reason %d\n",
+                                   __func__, reason);
+       }
+}
+
+/*
+ * Function irlap_discovery_request (gen_addr_bit)
+ *
+ *    Start one single discovery operation.
+ *
+ */
+void irlap_discovery_request(struct irlap_cb *self, discovery_t *discovery)
+{
+       struct irlap_info info;
+
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+       IRDA_ASSERT(discovery != NULL, return;);
+
+       pr_debug("%s(), nslots = %d\n", __func__, discovery->nslots);
+
+       IRDA_ASSERT((discovery->nslots == 1) || (discovery->nslots == 6) ||
+                   (discovery->nslots == 8) || (discovery->nslots == 16),
+                   return;);
+
+       /* Discovery is only possible in NDM mode */
+       if (self->state != LAP_NDM) {
+               pr_debug("%s(), discovery only possible in NDM mode\n",
+                        __func__);
+               irlap_discovery_confirm(self, NULL);
+               /* Note : in theory, if we are not in NDM, we could postpone
+                * the discovery like we do for connection request.
+                * In practice, it's not worth it. If the media was busy,
+                * it's likely next time around it won't be busy. If we are
+                * in REPLY state, we will get passive discovery info & event.
+                * Jean II */
+               return;
+       }
+
+       /* Check if last discovery request finished in time, or if
+        * it was aborted due to the media busy flag. */
+       if (self->discovery_log != NULL) {
+               hashbin_delete(self->discovery_log, (FREE_FUNC) kfree);
+               self->discovery_log = NULL;
+       }
+
+       /* All operations will occur at predictable time, no need to lock */
+       self->discovery_log = hashbin_new(HB_NOLOCK);
+
+       if (self->discovery_log == NULL) {
+               net_warn_ratelimited("%s(), Unable to allocate discovery log!\n",
+                                    __func__);
+               return;
+       }
+
+       info.S = discovery->nslots; /* Number of slots */
+       info.s = 0; /* Current slot */
+
+       self->discovery_cmd = discovery;
+       info.discovery = discovery;
+
+       /* sysctl_slot_timeout bounds are checked in irsysctl.c - Jean II */
+       self->slot_timeout = msecs_to_jiffies(sysctl_slot_timeout);
+
+       irlap_do_event(self, DISCOVERY_REQUEST, NULL, &info);
+}
+
+/*
+ * Function irlap_discovery_confirm (log)
+ *
+ *    A device has been discovered in front of this station, we
+ *    report directly to LMP.
+ */
+void irlap_discovery_confirm(struct irlap_cb *self, hashbin_t *discovery_log)
+{
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       IRDA_ASSERT(self->notify.instance != NULL, return;);
+
+       /*
+        * Check for successful discovery, since we are then allowed to clear
+        * the media busy condition (IrLAP 6.13.4 - p.94). This should allow
+        * us to make connection attempts much faster and easier (i.e. no
+        * collisions).
+        * Setting media busy to false will also generate an event allowing
+        * to process pending events in NDM state machine.
+        * Note : the spec doesn't define what's a successful discovery is.
+        * If we want Ultra to work, it's successful even if there is
+        * nobody discovered - Jean II
+        */
+       if (discovery_log)
+               irda_device_set_media_busy(self->netdev, FALSE);
+
+       /* Inform IrLMP */
+       irlmp_link_discovery_confirm(self->notify.instance, discovery_log);
+}
+
+/*
+ * Function irlap_discovery_indication (log)
+ *
+ *    Somebody is trying to discover us!
+ *
+ */
+void irlap_discovery_indication(struct irlap_cb *self, discovery_t *discovery)
+{
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+       IRDA_ASSERT(discovery != NULL, return;);
+
+       IRDA_ASSERT(self->notify.instance != NULL, return;);
+
+       /* A device is very likely to connect immediately after it performs
+        * a successful discovery. This means that in our case, we are much
+        * more likely to receive a connection request over the medium.
+        * So, we backoff to avoid collisions.
+        * IrLAP spec 6.13.4 suggest 100ms...
+        * Note : this little trick actually make a *BIG* difference. If I set
+        * my Linux box with discovery enabled and one Ultra frame sent every
+        * second, my Palm has no trouble connecting to it every time !
+        * Jean II */
+       irda_device_set_media_busy(self->netdev, SMALL);
+
+       irlmp_link_discovery_indication(self->notify.instance, discovery);
+}
+
+/*
+ * Function irlap_status_indication (quality_of_link)
+ */
+void irlap_status_indication(struct irlap_cb *self, int quality_of_link)
+{
+       switch (quality_of_link) {
+       case STATUS_NO_ACTIVITY:
+               net_info_ratelimited("IrLAP, no activity on link!\n");
+               break;
+       case STATUS_NOISY:
+               net_info_ratelimited("IrLAP, noisy link!\n");
+               break;
+       default:
+               break;
+       }
+       irlmp_status_indication(self->notify.instance,
+                               quality_of_link, LOCK_NO_CHANGE);
+}
+
+/*
+ * Function irlap_reset_indication (void)
+ */
+void irlap_reset_indication(struct irlap_cb *self)
+{
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       if (self->state == LAP_RESET_WAIT)
+               irlap_do_event(self, RESET_REQUEST, NULL, NULL);
+       else
+               irlap_do_event(self, RESET_RESPONSE, NULL, NULL);
+}
+
+/*
+ * Function irlap_reset_confirm (void)
+ */
+void irlap_reset_confirm(void)
+{
+}
+
+/*
+ * Function irlap_generate_rand_time_slot (S, s)
+ *
+ *    Generate a random time slot between s and S-1 where
+ *    S = Number of slots (0 -> S-1)
+ *    s = Current slot
+ */
+int irlap_generate_rand_time_slot(int S, int s)
+{
+       static int rand;
+       int slot;
+
+       IRDA_ASSERT((S - s) > 0, return 0;);
+
+       rand += jiffies;
+       rand ^= (rand << 12);
+       rand ^= (rand >> 20);
+
+       slot = s + rand % (S-s);
+
+       IRDA_ASSERT((slot >= s) || (slot < S), return 0;);
+
+       return slot;
+}
+
+/*
+ * Function irlap_update_nr_received (nr)
+ *
+ *    Remove all acknowledged frames in current window queue. This code is
+ *    not intuitive and you should not try to change it. If you think it
+ *    contains bugs, please mail a patch to the author instead.
+ */
+void irlap_update_nr_received(struct irlap_cb *self, int nr)
+{
+       struct sk_buff *skb = NULL;
+       int count = 0;
+
+       /*
+        * Remove all the ack-ed frames from the window queue.
+        */
+
+       /*
+        *  Optimize for the common case. It is most likely that the receiver
+        *  will acknowledge all the frames we have sent! So in that case we
+        *  delete all frames stored in window.
+        */
+       if (nr == self->vs) {
+               while ((skb = skb_dequeue(&self->wx_list)) != NULL) {
+                       dev_kfree_skb(skb);
+               }
+               /* The last acked frame is the next to send minus one */
+               self->va = nr - 1;
+       } else {
+               /* Remove all acknowledged frames in current window */
+               while ((skb_peek(&self->wx_list) != NULL) &&
+                      (((self->va+1) % 8) != nr))
+               {
+                       skb = skb_dequeue(&self->wx_list);
+                       dev_kfree_skb(skb);
+
+                       self->va = (self->va + 1) % 8;
+                       count++;
+               }
+       }
+
+       /* Advance window */
+       self->window = self->window_size - skb_queue_len(&self->wx_list);
+}
+
+/*
+ * Function irlap_validate_ns_received (ns)
+ *
+ *    Validate the next to send (ns) field from received frame.
+ */
+int irlap_validate_ns_received(struct irlap_cb *self, int ns)
+{
+       /*  ns as expected?  */
+       if (ns == self->vr)
+               return NS_EXPECTED;
+       /*
+        *  Stations are allowed to treat invalid NS as unexpected NS
+        *  IrLAP, Recv ... with-invalid-Ns. p. 84
+        */
+       return NS_UNEXPECTED;
+
+       /* return NR_INVALID; */
+}
+/*
+ * Function irlap_validate_nr_received (nr)
+ *
+ *    Validate the next to receive (nr) field from received frame.
+ *
+ */
+int irlap_validate_nr_received(struct irlap_cb *self, int nr)
+{
+       /*  nr as expected?  */
+       if (nr == self->vs) {
+               pr_debug("%s(), expected!\n", __func__);
+               return NR_EXPECTED;
+       }
+
+       /*
+        *  unexpected nr? (but within current window), first we check if the
+        *  ns numbers of the frames in the current window wrap.
+        */
+       if (self->va < self->vs) {
+               if ((nr >= self->va) && (nr <= self->vs))
+                       return NR_UNEXPECTED;
+       } else {
+               if ((nr >= self->va) || (nr <= self->vs))
+                       return NR_UNEXPECTED;
+       }
+
+       /* Invalid nr!  */
+       return NR_INVALID;
+}
+
+/*
+ * Function irlap_initiate_connection_state ()
+ *
+ *    Initialize the connection state parameters
+ *
+ */
+void irlap_initiate_connection_state(struct irlap_cb *self)
+{
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       /* Next to send and next to receive */
+       self->vs = self->vr = 0;
+
+       /* Last frame which got acked (0 - 1) % 8 */
+       self->va = 7;
+
+       self->window = 1;
+
+       self->remote_busy = FALSE;
+       self->retry_count = 0;
+}
+
+/*
+ * Function irlap_wait_min_turn_around (self, qos)
+ *
+ *    Wait negotiated minimum turn around time, this function actually sets
+ *    the number of BOS's that must be sent before the next transmitted
+ *    frame in order to delay for the specified amount of time. This is
+ *    done to avoid using timers, and the forbidden udelay!
+ */
+void irlap_wait_min_turn_around(struct irlap_cb *self, struct qos_info *qos)
+{
+       __u32 min_turn_time;
+       __u32 speed;
+
+       /* Get QoS values.  */
+       speed = qos->baud_rate.value;
+       min_turn_time = qos->min_turn_time.value;
+
+       /* No need to calculate XBOFs for speeds over 115200 bps */
+       if (speed > 115200) {
+               self->mtt_required = min_turn_time;
+               return;
+       }
+
+       /*
+        *  Send additional BOF's for the next frame for the requested
+        *  min turn time, so now we must calculate how many chars (XBOF's) we
+        *  must send for the requested time period (min turn time)
+        */
+       self->xbofs_delay = irlap_min_turn_time_in_bytes(speed, min_turn_time);
+}
+
+/*
+ * Function irlap_flush_all_queues (void)
+ *
+ *    Flush all queues
+ *
+ */
+void irlap_flush_all_queues(struct irlap_cb *self)
+{
+       struct sk_buff* skb;
+
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       /* Free transmission queue */
+       while ((skb = skb_dequeue(&self->txq)) != NULL)
+               dev_kfree_skb(skb);
+
+       while ((skb = skb_dequeue(&self->txq_ultra)) != NULL)
+               dev_kfree_skb(skb);
+
+       /* Free sliding window buffered packets */
+       while ((skb = skb_dequeue(&self->wx_list)) != NULL)
+               dev_kfree_skb(skb);
+}
+
+/*
+ * Function irlap_setspeed (self, speed)
+ *
+ *    Change the speed of the IrDA port
+ *
+ */
+static void irlap_change_speed(struct irlap_cb *self, __u32 speed, int now)
+{
+       struct sk_buff *skb;
+
+       pr_debug("%s(), setting speed to %d\n", __func__, speed);
+
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       self->speed = speed;
+
+       /* Change speed now, or just piggyback speed on frames */
+       if (now) {
+               /* Send down empty frame to trigger speed change */
+               skb = alloc_skb(0, GFP_ATOMIC);
+               if (skb)
+                       irlap_queue_xmit(self, skb);
+       }
+}
+
+/*
+ * Function irlap_init_qos_capabilities (self, qos)
+ *
+ *    Initialize QoS for this IrLAP session, What we do is to compute the
+ *    intersection of the QoS capabilities for the user, driver and for
+ *    IrLAP itself. Normally, IrLAP will not specify any values, but it can
+ *    be used to restrict certain values.
+ */
+static void irlap_init_qos_capabilities(struct irlap_cb *self,
+                                       struct qos_info *qos_user)
+{
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+       IRDA_ASSERT(self->netdev != NULL, return;);
+
+       /* Start out with the maximum QoS support possible */
+       irda_init_max_qos_capabilies(&self->qos_rx);
+
+       /* Apply drivers QoS capabilities */
+       irda_qos_compute_intersection(&self->qos_rx, self->qos_dev);
+
+       /*
+        *  Check for user supplied QoS parameters. The service user is only
+        *  allowed to supply these values. We check each parameter since the
+        *  user may not have set all of them.
+        */
+       if (qos_user) {
+               pr_debug("%s(), Found user specified QoS!\n", __func__);
+
+               if (qos_user->baud_rate.bits)
+                       self->qos_rx.baud_rate.bits &= qos_user->baud_rate.bits;
+
+               if (qos_user->max_turn_time.bits)
+                       self->qos_rx.max_turn_time.bits &= qos_user->max_turn_time.bits;
+               if (qos_user->data_size.bits)
+                       self->qos_rx.data_size.bits &= qos_user->data_size.bits;
+
+               if (qos_user->link_disc_time.bits)
+                       self->qos_rx.link_disc_time.bits &= qos_user->link_disc_time.bits;
+       }
+
+       /* Use 500ms in IrLAP for now */
+       self->qos_rx.max_turn_time.bits &= 0x01;
+
+       /* Set data size */
+       /*self->qos_rx.data_size.bits &= 0x03;*/
+
+       irda_qos_bits_to_value(&self->qos_rx);
+}
+
+/*
+ * Function irlap_apply_default_connection_parameters (void, now)
+ *
+ *    Use the default connection and transmission parameters
+ */
+void irlap_apply_default_connection_parameters(struct irlap_cb *self)
+{
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       /* xbofs : Default value in NDM */
+       self->next_bofs   = 12;
+       self->bofs_count  = 12;
+
+       /* NDM Speed is 9600 */
+       irlap_change_speed(self, 9600, TRUE);
+
+       /* Set mbusy when going to NDM state */
+       irda_device_set_media_busy(self->netdev, TRUE);
+
+       /*
+        * Generate random connection address for this session, which must
+        * be 7 bits wide and different from 0x00 and 0xfe
+        */
+       while ((self->caddr == 0x00) || (self->caddr == 0xfe)) {
+               get_random_bytes(&self->caddr, sizeof(self->caddr));
+               self->caddr &= 0xfe;
+       }
+
+       /* Use default values until connection has been negitiated */
+       self->slot_timeout = sysctl_slot_timeout;
+       self->final_timeout = FINAL_TIMEOUT;
+       self->poll_timeout = POLL_TIMEOUT;
+       self->wd_timeout = WD_TIMEOUT;
+
+       /* Set some default values */
+       self->qos_tx.baud_rate.value = 9600;
+       self->qos_rx.baud_rate.value = 9600;
+       self->qos_tx.max_turn_time.value = 0;
+       self->qos_rx.max_turn_time.value = 0;
+       self->qos_tx.min_turn_time.value = 0;
+       self->qos_rx.min_turn_time.value = 0;
+       self->qos_tx.data_size.value = 64;
+       self->qos_rx.data_size.value = 64;
+       self->qos_tx.window_size.value = 1;
+       self->qos_rx.window_size.value = 1;
+       self->qos_tx.additional_bofs.value = 12;
+       self->qos_rx.additional_bofs.value = 12;
+       self->qos_tx.link_disc_time.value = 0;
+       self->qos_rx.link_disc_time.value = 0;
+
+       irlap_flush_all_queues(self);
+
+       self->disconnect_pending = FALSE;
+       self->connect_pending = FALSE;
+}
+
+/*
+ * Function irlap_apply_connection_parameters (qos, now)
+ *
+ *    Initialize IrLAP with the negotiated QoS values
+ *
+ * If 'now' is false, the speed and xbofs will be changed after the next
+ * frame is sent.
+ * If 'now' is true, the speed and xbofs is changed immediately
+ */
+void irlap_apply_connection_parameters(struct irlap_cb *self, int now)
+{
+       IRDA_ASSERT(self != NULL, return;);
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return;);
+
+       /* Set the negotiated xbofs value */
+       self->next_bofs   = self->qos_tx.additional_bofs.value;
+       if (now)
+               self->bofs_count = self->next_bofs;
+
+       /* Set the negotiated link speed (may need the new xbofs value) */
+       irlap_change_speed(self, self->qos_tx.baud_rate.value, now);
+
+       self->window_size = self->qos_tx.window_size.value;
+       self->window      = self->qos_tx.window_size.value;
+
+#ifdef CONFIG_IRDA_DYNAMIC_WINDOW
+       /*
+        *  Calculate how many bytes it is possible to transmit before the
+        *  link must be turned around
+        */
+       self->line_capacity =
+               irlap_max_line_capacity(self->qos_tx.baud_rate.value,
+                                       self->qos_tx.max_turn_time.value);
+       self->bytes_left = self->line_capacity;
+#endif /* CONFIG_IRDA_DYNAMIC_WINDOW */
+
+
+       /*
+        *  Initialize timeout values, some of the rules are listed on
+        *  page 92 in IrLAP.
+        */
+       IRDA_ASSERT(self->qos_tx.max_turn_time.value != 0, return;);
+       IRDA_ASSERT(self->qos_rx.max_turn_time.value != 0, return;);
+       /* The poll timeout applies only to the primary station.
+        * It defines the maximum time the primary stay in XMIT mode
+        * before timeout and turning the link around (sending a RR).
+        * Or, this is how much we can keep the pf bit in primary mode.
+        * Therefore, it must be lower or equal than our *OWN* max turn around.
+        * Jean II */
+       self->poll_timeout = msecs_to_jiffies(
+                               self->qos_tx.max_turn_time.value);
+       /* The Final timeout applies only to the primary station.
+        * It defines the maximum time the primary wait (mostly in RECV mode)
+        * for an answer from the secondary station before polling it again.
+        * Therefore, it must be greater or equal than our *PARTNER*
+        * max turn around time - Jean II */
+       self->final_timeout = msecs_to_jiffies(
+                               self->qos_rx.max_turn_time.value);
+       /* The Watchdog Bit timeout applies only to the secondary station.
+        * It defines the maximum time the secondary wait (mostly in RECV mode)
+        * for poll from the primary station before getting annoyed.
+        * Therefore, it must be greater or equal than our *PARTNER*
+        * max turn around time - Jean II */
+       self->wd_timeout = self->final_timeout * 2;
+
+       /*
+        * N1 and N2 are maximum retry count for *both* the final timer
+        * and the wd timer (with a factor 2) as defined above.
+        * After N1 retry of a timer, we give a warning to the user.
+        * After N2 retry, we consider the link dead and disconnect it.
+        * Jean II
+        */
+
+       /*
+        *  Set N1 to 0 if Link Disconnect/Threshold Time = 3 and set it to
+        *  3 seconds otherwise. See page 71 in IrLAP for more details.
+        *  Actually, it's not always 3 seconds, as we allow to set
+        *  it via sysctl... Max maxtt is 500ms, and N1 need to be multiple
+        *  of 2, so 1 second is minimum we can allow. - Jean II
+        */
+       if (self->qos_tx.link_disc_time.value == sysctl_warn_noreply_time)
+               /*
+                * If we set N1 to 0, it will trigger immediately, which is
+                * not what we want. What we really want is to disable it,
+                * Jean II
+                */
+               self->N1 = -2; /* Disable - Need to be multiple of 2*/
+       else
+               self->N1 = sysctl_warn_noreply_time * 1000 /
+                 self->qos_rx.max_turn_time.value;
+
+       pr_debug("Setting N1 = %d\n", self->N1);
+
+       /* Set N2 to match our own disconnect time */
+       self->N2 = self->qos_tx.link_disc_time.value * 1000 /
+               self->qos_rx.max_turn_time.value;
+       pr_debug("Setting N2 = %d\n", self->N2);
+}
+
+#ifdef CONFIG_PROC_FS
+struct irlap_iter_state {
+       int id;
+};
+
+static void *irlap_seq_start(struct seq_file *seq, loff_t *pos)
+{
+       struct irlap_iter_state *iter = seq->private;
+       struct irlap_cb *self;
+
+       /* Protect our access to the tsap list */
+       spin_lock_irq(&irlap->hb_spinlock);
+       iter->id = 0;
+
+       for (self = (struct irlap_cb *) hashbin_get_first(irlap);
+            self; self = (struct irlap_cb *) hashbin_get_next(irlap)) {
+               if (iter->id == *pos)
+                       break;
+               ++iter->id;
+       }
+
+       return self;
+}
+
+static void *irlap_seq_next(struct seq_file *seq, void *v, loff_t *pos)
+{
+       struct irlap_iter_state *iter = seq->private;
+
+       ++*pos;
+       ++iter->id;
+       return (void *) hashbin_get_next(irlap);
+}
+
+static void irlap_seq_stop(struct seq_file *seq, void *v)
+{
+       spin_unlock_irq(&irlap->hb_spinlock);
+}
+
+static int irlap_seq_show(struct seq_file *seq, void *v)
+{
+       const struct irlap_iter_state *iter = seq->private;
+       const struct irlap_cb *self = v;
+
+       IRDA_ASSERT(self->magic == LAP_MAGIC, return -EINVAL;);
+
+       seq_printf(seq, "irlap%d ", iter->id);
+       seq_printf(seq, "state: %s\n",
+                  irlap_state[self->state]);
+
+       seq_printf(seq, "  device name: %s, ",
+                  (self->netdev) ? self->netdev->name : "bug");
+       seq_printf(seq, "hardware name: %s\n", self->hw_name);
+
+       seq_printf(seq, "  caddr: %#02x, ", self->caddr);
+       seq_printf(seq, "saddr: %#08x, ", self->saddr);
+       seq_printf(seq, "daddr: %#08x\n", self->daddr);
+
+       seq_printf(seq, "  win size: %d, ",
+                  self->window_size);
+       seq_printf(seq, "win: %d, ", self->window);
+#ifdef CONFIG_IRDA_DYNAMIC_WINDOW
+       seq_printf(seq, "line capacity: %d, ",
+                  self->line_capacity);
+       seq_printf(seq, "bytes left: %d\n", self->bytes_left);
+#endif /* CONFIG_IRDA_DYNAMIC_WINDOW */
+       seq_printf(seq, "  tx queue len: %d ",
+                  skb_queue_len(&self->txq));
+       seq_printf(seq, "win queue len: %d ",
+                  skb_queue_len(&self->wx_list));
+       seq_printf(seq, "rbusy: %s", self->remote_busy ?
+                  "TRUE" : "FALSE");
+       seq_printf(seq, " mbusy: %s\n", self->media_busy ?
+                  "TRUE" : "FALSE");
+
+       seq_printf(seq, "  retrans: %d ", self->retry_count);
+       seq_printf(seq, "vs: %d ", self->vs);
+       seq_printf(seq, "vr: %d ", self->vr);
+       seq_printf(seq, "va: %d\n", self->va);
+
+       seq_printf(seq, "  qos\tbps\tmaxtt\tdsize\twinsize\taddbofs\tmintt\tldisc\tcomp\n");
+
+       seq_printf(seq, "  tx\t%d\t",
+                  self->qos_tx.baud_rate.value);
+       seq_printf(seq, "%d\t",
+                  self->qos_tx.max_turn_time.value);
+       seq_printf(seq, "%d\t",
+                  self->qos_tx.data_size.value);
+       seq_printf(seq, "%d\t",
+                  self->qos_tx.window_size.value);
+       seq_printf(seq, "%d\t",
+                  self->qos_tx.additional_bofs.value);
+       seq_printf(seq, "%d\t",
+                  self->qos_tx.min_turn_time.value);
+       seq_printf(seq, "%d\t",
+                  self->qos_tx.link_disc_time.value);
+       seq_printf(seq, "\n");
+
+       seq_printf(seq, "  rx\t%d\t",
+                  self->qos_rx.baud_rate.value);
+       seq_printf(seq, "%d\t",
+                  self->qos_rx.max_turn_time.value);
+       seq_printf(seq, "%d\t",
+                  self->qos_rx.data_size.value);
+       seq_printf(seq, "%d\t",
+                  self->qos_rx.window_size.value);
+       seq_printf(seq, "%d\t",
+                  self->qos_rx.additional_bofs.value);
+       seq_printf(seq, "%d\t",
+                  self->qos_rx.min_turn_time.value);
+       seq_printf(seq, "%d\n",
+                  self->qos_rx.link_disc_time.value);
+
+       return 0;
+}
+
+static const struct seq_operations irlap_seq_ops = {
+       .start  = irlap_seq_start,
+       .next   = irlap_seq_next,
+       .stop   = irlap_seq_stop,
+       .show   = irlap_seq_show,
+};
+
+static int irlap_seq_open(struct inode *inode, struct file *file)
+{
+       if (irlap == NULL)
+               return -EINVAL;
+
+       return seq_open_private(file, &irlap_seq_ops,
+                       sizeof(struct irlap_iter_state));
+}
+
+const struct file_operations irlap_seq_fops = {
+       .owner          = THIS_MODULE,
+       .open           = irlap_seq_open,
+       .read           = seq_read,
+       .llseek         = seq_lseek,
+       .release        = seq_release_private,
+};
+
+#endif /* CONFIG_PROC_FS */