[CAN]: Add broadcast manager (bcm) protocol
authorOliver Hartkopp <oliver.hartkopp@volkswagen.de>
Fri, 16 Nov 2007 23:53:52 +0000 (15:53 -0800)
committerDavid S. Miller <davem@davemloft.net>
Mon, 28 Jan 2008 22:54:11 +0000 (14:54 -0800)
This patch adds the CAN broadcast manager (bcm) protocol.

Signed-off-by: Oliver Hartkopp <oliver.hartkopp@volkswagen.de>
Signed-off-by: Urs Thuermann <urs.thuermann@volkswagen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/linux/can/bcm.h [new file with mode: 0644]
net/can/Kconfig
net/can/Makefile
net/can/bcm.c [new file with mode: 0644]

diff --git a/include/linux/can/bcm.h b/include/linux/can/bcm.h
new file mode 100644 (file)
index 0000000..7ade33a
--- /dev/null
@@ -0,0 +1,65 @@
+/*
+ * linux/can/bcm.h
+ *
+ * Definitions for CAN Broadcast Manager (BCM)
+ *
+ * Author: Oliver Hartkopp <oliver.hartkopp@volkswagen.de>
+ * Copyright (c) 2002-2007 Volkswagen Group Electronic Research
+ * All rights reserved.
+ *
+ * Send feedback to <socketcan-users@lists.berlios.de>
+ *
+ */
+
+#ifndef CAN_BCM_H
+#define CAN_BCM_H
+
+/**
+ * struct bcm_msg_head - head of messages to/from the broadcast manager
+ * @opcode:    opcode, see enum below.
+ * @flags:     special flags, see below.
+ * @count:     number of frames to send before changing interval.
+ * @ival1:     interval for the first @count frames.
+ * @ival2:     interval for the following frames.
+ * @can_id:    CAN ID of frames to be sent or received.
+ * @nframes:   number of frames appended to the message head.
+ * @frames:    array of CAN frames.
+ */
+struct bcm_msg_head {
+       int opcode;
+       int flags;
+       int count;
+       struct timeval ival1, ival2;
+       canid_t can_id;
+       int nframes;
+       struct can_frame frames[0];
+};
+
+enum {
+       TX_SETUP = 1,   /* create (cyclic) transmission task */
+       TX_DELETE,      /* remove (cyclic) transmission task */
+       TX_READ,        /* read properties of (cyclic) transmission task */
+       TX_SEND,        /* send one CAN frame */
+       RX_SETUP,       /* create RX content filter subscription */
+       RX_DELETE,      /* remove RX content filter subscription */
+       RX_READ,        /* read properties of RX content filter subscription */
+       TX_STATUS,      /* reply to TX_READ request */
+       TX_EXPIRED,     /* notification on performed transmissions (count=0) */
+       RX_STATUS,      /* reply to RX_READ request */
+       RX_TIMEOUT,     /* cyclic message is absent */
+       RX_CHANGED      /* updated CAN frame (detected content change) */
+};
+
+#define SETTIMER            0x0001
+#define STARTTIMER          0x0002
+#define TX_COUNTEVT         0x0004
+#define TX_ANNOUNCE         0x0008
+#define TX_CP_CAN_ID        0x0010
+#define RX_FILTER_ID        0x0020
+#define RX_CHECK_DLC        0x0040
+#define RX_NO_AUTOTIMER     0x0080
+#define RX_ANNOUNCE_RESUME  0x0100
+#define TX_RESET_MULTI_IDX  0x0200
+#define RX_RTR_FRAME        0x0400
+
+#endif /* CAN_BCM_H */
index 4718d1f..182b96b 100644 (file)
@@ -26,3 +26,16 @@ config CAN_RAW
          most cases where no higher level protocol is being used. The raw
          socket has several filter options e.g. ID masking / error frames.
          To receive/send raw CAN messages, use AF_CAN with protocol CAN_RAW.
+
+config CAN_BCM
+       tristate "Broadcast Manager CAN Protocol (with content filtering)"
+       depends on CAN
+       default N
+       ---help---
+         The Broadcast Manager offers content filtering, timeout monitoring,
+         sending of RTR frames, and cyclic CAN messages without permanent user
+         interaction. The BCM can be 'programmed' via the BSD socket API and
+         informs you on demand e.g. only on content updates / timeouts.
+         You probably want to use the bcm socket in most cases where cyclic
+         CAN messages are used on the bus (e.g. in automotive environments).
+         To use the Broadcast Manager, use AF_CAN with protocol CAN_BCM.
index 86f1cf2..9cd3c4b 100644 (file)
@@ -7,3 +7,6 @@ can-objs                := af_can.o proc.o
 
 obj-$(CONFIG_CAN_RAW)  += can-raw.o
 can-raw-objs           := raw.o
+
+obj-$(CONFIG_CAN_BCM)  += can-bcm.o
+can-bcm-objs           := bcm.o
diff --git a/net/can/bcm.c b/net/can/bcm.c
new file mode 100644 (file)
index 0000000..bd4282d
--- /dev/null
@@ -0,0 +1,1561 @@
+/*
+ * bcm.c - Broadcast Manager to filter/send (cyclic) CAN content
+ *
+ * Copyright (c) 2002-2007 Volkswagen Group Electronic Research
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the name of Volkswagen nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * Alternatively, provided that this notice is retained in full, this
+ * software may be distributed under the terms of the GNU General
+ * Public License ("GPL") version 2, in which case the provisions of the
+ * GPL apply INSTEAD OF those given above.
+ *
+ * The provided data structures and external interfaces from this code
+ * are not restricted to be used by modules with a GPL compatible license.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
+ * DAMAGE.
+ *
+ * Send feedback to <socketcan-users@lists.berlios.de>
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/list.h>
+#include <linux/proc_fs.h>
+#include <linux/uio.h>
+#include <linux/net.h>
+#include <linux/netdevice.h>
+#include <linux/socket.h>
+#include <linux/if_arp.h>
+#include <linux/skbuff.h>
+#include <linux/can.h>
+#include <linux/can/core.h>
+#include <linux/can/bcm.h>
+#include <net/sock.h>
+#include <net/net_namespace.h>
+
+/* use of last_frames[index].can_dlc */
+#define RX_RECV    0x40 /* received data for this element */
+#define RX_THR     0x80 /* element not been sent due to throttle feature */
+#define BCM_CAN_DLC_MASK 0x0F /* clean private flags in can_dlc by masking */
+
+/* get best masking value for can_rx_register() for a given single can_id */
+#define REGMASK(id) ((id & CAN_RTR_FLAG) | ((id & CAN_EFF_FLAG) ? \
+                       (CAN_EFF_MASK | CAN_EFF_FLAG) : CAN_SFF_MASK))
+
+#define CAN_BCM_VERSION CAN_VERSION
+static __initdata const char banner[] = KERN_INFO
+       "can: broadcast manager protocol (rev " CAN_BCM_VERSION ")\n";
+
+MODULE_DESCRIPTION("PF_CAN broadcast manager protocol");
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_AUTHOR("Oliver Hartkopp <oliver.hartkopp@volkswagen.de>");
+
+/* easy access to can_frame payload */
+static inline u64 GET_U64(const struct can_frame *cp)
+{
+       return *(u64 *)cp->data;
+}
+
+struct bcm_op {
+       struct list_head list;
+       int ifindex;
+       canid_t can_id;
+       int flags;
+       unsigned long j_ival1, j_ival2, j_lastmsg;
+       unsigned long frames_abs, frames_filtered;
+       struct timer_list timer, thrtimer;
+       struct timeval ival1, ival2;
+       ktime_t rx_stamp;
+       int rx_ifindex;
+       int count;
+       int nframes;
+       int currframe;
+       struct can_frame *frames;
+       struct can_frame *last_frames;
+       struct can_frame sframe;
+       struct can_frame last_sframe;
+       struct sock *sk;
+       struct net_device *rx_reg_dev;
+};
+
+static struct proc_dir_entry *proc_dir;
+
+struct bcm_sock {
+       struct sock sk;
+       int bound;
+       int ifindex;
+       struct notifier_block notifier;
+       struct list_head rx_ops;
+       struct list_head tx_ops;
+       unsigned long dropped_usr_msgs;
+       struct proc_dir_entry *bcm_proc_read;
+       char procname [9]; /* pointer printed in ASCII with \0 */
+};
+
+static inline struct bcm_sock *bcm_sk(const struct sock *sk)
+{
+       return (struct bcm_sock *)sk;
+}
+
+#define CFSIZ sizeof(struct can_frame)
+#define OPSIZ sizeof(struct bcm_op)
+#define MHSIZ sizeof(struct bcm_msg_head)
+
+/*
+ * rounded_tv2jif - calculate jiffies from timeval including optional up
+ * @tv: pointer to timeval
+ *
+ * Description:
+ * Unlike timeval_to_jiffies() provided in include/linux/jiffies.h, this
+ * function is intentionally more relaxed on precise timer ticks to get
+ * exact one jiffy for requested 1000us on a 1000HZ machine.
+ * This code is to be removed when upgrading to kernel hrtimer.
+ *
+ * Return:
+ *  calculated jiffies (max: ULONG_MAX)
+ */
+static unsigned long rounded_tv2jif(const struct timeval *tv)
+{
+       unsigned long sec  = tv->tv_sec;
+       unsigned long usec = tv->tv_usec;
+       unsigned long jif;
+
+       if (sec > ULONG_MAX / HZ)
+               return ULONG_MAX;
+
+       /* round up to get at least the requested time */
+       usec += 1000000 / HZ - 1;
+
+       jif  = usec / (1000000 / HZ);
+
+       if (sec * HZ > ULONG_MAX - jif)
+               return ULONG_MAX;
+
+       return jif + sec * HZ;
+}
+
+/*
+ * procfs functions
+ */
+static char *bcm_proc_getifname(int ifindex)
+{
+       struct net_device *dev;
+
+       if (!ifindex)
+               return "any";
+
+       /* no usage counting */
+       dev = __dev_get_by_index(&init_net, ifindex);
+       if (dev)
+               return dev->name;
+
+       return "???";
+}
+
+static int bcm_read_proc(char *page, char **start, off_t off,
+                        int count, int *eof, void *data)
+{
+       int len = 0;
+       struct sock *sk = (struct sock *)data;
+       struct bcm_sock *bo = bcm_sk(sk);
+       struct bcm_op *op;
+
+       len += snprintf(page + len, PAGE_SIZE - len, ">>> socket %p",
+                       sk->sk_socket);
+       len += snprintf(page + len, PAGE_SIZE - len, " / sk %p", sk);
+       len += snprintf(page + len, PAGE_SIZE - len, " / bo %p", bo);
+       len += snprintf(page + len, PAGE_SIZE - len, " / dropped %lu",
+                       bo->dropped_usr_msgs);
+       len += snprintf(page + len, PAGE_SIZE - len, " / bound %s",
+                       bcm_proc_getifname(bo->ifindex));
+       len += snprintf(page + len, PAGE_SIZE - len, " <<<\n");
+
+       list_for_each_entry(op, &bo->rx_ops, list) {
+
+               unsigned long reduction;
+
+               /* print only active entries & prevent division by zero */
+               if (!op->frames_abs)
+                       continue;
+
+               len += snprintf(page + len, PAGE_SIZE - len,
+                               "rx_op: %03X %-5s ",
+                               op->can_id, bcm_proc_getifname(op->ifindex));
+               len += snprintf(page + len, PAGE_SIZE - len, "[%d]%c ",
+                               op->nframes,
+                               (op->flags & RX_CHECK_DLC)?'d':' ');
+               if (op->j_ival1)
+                       len += snprintf(page + len, PAGE_SIZE - len,
+                                       "timeo=%ld ", op->j_ival1);
+
+               if (op->j_ival2)
+                       len += snprintf(page + len, PAGE_SIZE - len,
+                                       "thr=%ld ", op->j_ival2);
+
+               len += snprintf(page + len, PAGE_SIZE - len,
+                               "# recv %ld (%ld) => reduction: ",
+                               op->frames_filtered, op->frames_abs);
+
+               reduction = 100 - (op->frames_filtered * 100) / op->frames_abs;
+
+               len += snprintf(page + len, PAGE_SIZE - len, "%s%ld%%\n",
+                               (reduction == 100)?"near ":"", reduction);
+
+               if (len > PAGE_SIZE - 200) {
+                       /* mark output cut off */
+                       len += snprintf(page + len, PAGE_SIZE - len, "(..)\n");
+                       break;
+               }
+       }
+
+       list_for_each_entry(op, &bo->tx_ops, list) {
+
+               len += snprintf(page + len, PAGE_SIZE - len,
+                               "tx_op: %03X %s [%d] ",
+                               op->can_id, bcm_proc_getifname(op->ifindex),
+                               op->nframes);
+               if (op->j_ival1)
+                       len += snprintf(page + len, PAGE_SIZE - len, "t1=%ld ",
+                                       op->j_ival1);
+
+               if (op->j_ival2)
+                       len += snprintf(page + len, PAGE_SIZE - len, "t2=%ld ",
+                                       op->j_ival2);
+
+               len += snprintf(page + len, PAGE_SIZE - len, "# sent %ld\n",
+                               op->frames_abs);
+
+               if (len > PAGE_SIZE - 100) {
+                       /* mark output cut off */
+                       len += snprintf(page + len, PAGE_SIZE - len, "(..)\n");
+                       break;
+               }
+       }
+
+       len += snprintf(page + len, PAGE_SIZE - len, "\n");
+
+       *eof = 1;
+       return len;
+}
+
+/*
+ * bcm_can_tx - send the (next) CAN frame to the appropriate CAN interface
+ *              of the given bcm tx op
+ */
+static void bcm_can_tx(struct bcm_op *op)
+{
+       struct sk_buff *skb;
+       struct net_device *dev;
+       struct can_frame *cf = &op->frames[op->currframe];
+
+       /* no target device? => exit */
+       if (!op->ifindex)
+               return;
+
+       dev = dev_get_by_index(&init_net, op->ifindex);
+       if (!dev) {
+               /* RFC: should this bcm_op remove itself here? */
+               return;
+       }
+
+       skb = alloc_skb(CFSIZ, gfp_any());
+       if (!skb)
+               goto out;
+
+       memcpy(skb_put(skb, CFSIZ), cf, CFSIZ);
+
+       /* send with loopback */
+       skb->dev = dev;
+       skb->sk = op->sk;
+       can_send(skb, 1);
+
+       /* update statistics */
+       op->currframe++;
+       op->frames_abs++;
+
+       /* reached last frame? */
+       if (op->currframe >= op->nframes)
+               op->currframe = 0;
+ out:
+       dev_put(dev);
+}
+
+/*
+ * bcm_send_to_user - send a BCM message to the userspace
+ *                    (consisting of bcm_msg_head + x CAN frames)
+ */
+static void bcm_send_to_user(struct bcm_op *op, struct bcm_msg_head *head,
+                            struct can_frame *frames, int has_timestamp)
+{
+       struct sk_buff *skb;
+       struct can_frame *firstframe;
+       struct sockaddr_can *addr;
+       struct sock *sk = op->sk;
+       int datalen = head->nframes * CFSIZ;
+       int err;
+
+       skb = alloc_skb(sizeof(*head) + datalen, gfp_any());
+       if (!skb)
+               return;
+
+       memcpy(skb_put(skb, sizeof(*head)), head, sizeof(*head));
+
+       if (head->nframes) {
+               /* can_frames starting here */
+               firstframe = (struct can_frame *) skb_tail_pointer(skb);
+
+               memcpy(skb_put(skb, datalen), frames, datalen);
+
+               /*
+                * the BCM uses the can_dlc-element of the can_frame
+                * structure for internal purposes. This is only
+                * relevant for updates that are generated by the
+                * BCM, where nframes is 1
+                */
+               if (head->nframes == 1)
+                       firstframe->can_dlc &= BCM_CAN_DLC_MASK;
+       }
+
+       if (has_timestamp) {
+               /* restore rx timestamp */
+               skb->tstamp = op->rx_stamp;
+       }
+
+       /*
+        *  Put the datagram to the queue so that bcm_recvmsg() can
+        *  get it from there.  We need to pass the interface index to
+        *  bcm_recvmsg().  We pass a whole struct sockaddr_can in skb->cb
+        *  containing the interface index.
+        */
+
+       BUILD_BUG_ON(sizeof(skb->cb) < sizeof(struct sockaddr_can));
+       addr = (struct sockaddr_can *)skb->cb;
+       memset(addr, 0, sizeof(*addr));
+       addr->can_family  = AF_CAN;
+       addr->can_ifindex = op->rx_ifindex;
+
+       err = sock_queue_rcv_skb(sk, skb);
+       if (err < 0) {
+               struct bcm_sock *bo = bcm_sk(sk);
+
+               kfree_skb(skb);
+               /* don't care about overflows in this statistic */
+               bo->dropped_usr_msgs++;
+       }
+}
+
+/*
+ * bcm_tx_timeout_handler - performes cyclic CAN frame transmissions
+ */
+static void bcm_tx_timeout_handler(unsigned long data)
+{
+       struct bcm_op *op = (struct bcm_op *)data;
+
+       if (op->j_ival1 && (op->count > 0)) {
+
+               op->count--;
+               if (!op->count && (op->flags & TX_COUNTEVT)) {
+                       struct bcm_msg_head msg_head;
+
+                       /* create notification to user */
+                       msg_head.opcode  = TX_EXPIRED;
+                       msg_head.flags   = op->flags;
+                       msg_head.count   = op->count;
+                       msg_head.ival1   = op->ival1;
+                       msg_head.ival2   = op->ival2;
+                       msg_head.can_id  = op->can_id;
+                       msg_head.nframes = 0;
+
+                       bcm_send_to_user(op, &msg_head, NULL, 0);
+               }
+       }
+
+       if (op->j_ival1 && (op->count > 0)) {
+
+               /* send (next) frame */
+               bcm_can_tx(op);
+               mod_timer(&op->timer, jiffies + op->j_ival1);
+
+       } else {
+               if (op->j_ival2) {
+
+                       /* send (next) frame */
+                       bcm_can_tx(op);
+                       mod_timer(&op->timer, jiffies + op->j_ival2);
+               }
+       }
+
+       return;
+}
+
+/*
+ * bcm_rx_changed - create a RX_CHANGED notification due to changed content
+ */
+static void bcm_rx_changed(struct bcm_op *op, struct can_frame *data)
+{
+       struct bcm_msg_head head;
+
+       op->j_lastmsg = jiffies;
+
+       /* update statistics */
+       op->frames_filtered++;
+
+       /* prevent statistics overflow */
+       if (op->frames_filtered > ULONG_MAX/100)
+               op->frames_filtered = op->frames_abs = 0;
+
+       head.opcode  = RX_CHANGED;
+       head.flags   = op->flags;
+       head.count   = op->count;
+       head.ival1   = op->ival1;
+       head.ival2   = op->ival2;
+       head.can_id  = op->can_id;
+       head.nframes = 1;
+
+       bcm_send_to_user(op, &head, data, 1);
+}
+
+/*
+ * bcm_rx_update_and_send - process a detected relevant receive content change
+ *                          1. update the last received data
+ *                          2. send a notification to the user (if possible)
+ */
+static void bcm_rx_update_and_send(struct bcm_op *op,
+                                  struct can_frame *lastdata,
+                                  struct can_frame *rxdata)
+{
+       unsigned long nexttx = op->j_lastmsg + op->j_ival2;
+
+       memcpy(lastdata, rxdata, CFSIZ);
+
+       /* mark as used */
+       lastdata->can_dlc |= RX_RECV;
+
+       /* throttle bcm_rx_changed ? */
+       if ((op->thrtimer.expires) ||
+           ((op->j_ival2) && (nexttx > jiffies))) {
+               /* we are already waiting OR we have to start waiting */
+
+               /* mark as 'throttled' */
+               lastdata->can_dlc |= RX_THR;
+
+               if (!(op->thrtimer.expires)) {
+                       /* start the timer only the first time */
+                       mod_timer(&op->thrtimer, nexttx);
+               }
+
+       } else {
+               /* send RX_CHANGED to the user immediately */
+               bcm_rx_changed(op, rxdata);
+       }
+}
+
+/*
+ * bcm_rx_cmp_to_index - (bit)compares the currently received data to formerly
+ *                       received data stored in op->last_frames[]
+ */
+static void bcm_rx_cmp_to_index(struct bcm_op *op, int index,
+                               struct can_frame *rxdata)
+{
+       /*
+        * no one uses the MSBs of can_dlc for comparation,
+        * so we use it here to detect the first time of reception
+        */
+
+       if (!(op->last_frames[index].can_dlc & RX_RECV)) {
+               /* received data for the first time => send update to user */
+               bcm_rx_update_and_send(op, &op->last_frames[index], rxdata);
+               return;
+       }
+
+       /* do a real check in can_frame data section */
+
+       if ((GET_U64(&op->frames[index]) & GET_U64(rxdata)) !=
+           (GET_U64(&op->frames[index]) & GET_U64(&op->last_frames[index]))) {
+               bcm_rx_update_and_send(op, &op->last_frames[index], rxdata);
+               return;
+       }
+
+       if (op->flags & RX_CHECK_DLC) {
+               /* do a real check in can_frame dlc */
+               if (rxdata->can_dlc != (op->last_frames[index].can_dlc &
+                                       BCM_CAN_DLC_MASK)) {
+                       bcm_rx_update_and_send(op, &op->last_frames[index],
+                                              rxdata);
+                       return;
+               }
+       }
+}
+
+/*
+ * bcm_rx_starttimer - enable timeout monitoring for CAN frame receiption
+ */
+static void bcm_rx_starttimer(struct bcm_op *op)
+{
+       if (op->flags & RX_NO_AUTOTIMER)
+               return;
+
+       if (op->j_ival1)
+               mod_timer(&op->timer, jiffies + op->j_ival1);
+}
+
+/*
+ * bcm_rx_timeout_handler - when the (cyclic) CAN frame receiption timed out
+ */
+static void bcm_rx_timeout_handler(unsigned long data)
+{
+       struct bcm_op *op = (struct bcm_op *)data;
+       struct bcm_msg_head msg_head;
+
+       msg_head.opcode  = RX_TIMEOUT;
+       msg_head.flags   = op->flags;
+       msg_head.count   = op->count;
+       msg_head.ival1   = op->ival1;
+       msg_head.ival2   = op->ival2;
+       msg_head.can_id  = op->can_id;
+       msg_head.nframes = 0;
+
+       bcm_send_to_user(op, &msg_head, NULL, 0);
+
+       /* no restart of the timer is done here! */
+
+       /* if user wants to be informed, when cyclic CAN-Messages come back */
+       if ((op->flags & RX_ANNOUNCE_RESUME) && op->last_frames) {
+               /* clear received can_frames to indicate 'nothing received' */
+               memset(op->last_frames, 0, op->nframes * CFSIZ);
+       }
+}
+
+/*
+ * bcm_rx_thr_handler - the time for blocked content updates is over now:
+ *                      Check for throttled data and send it to the userspace
+ */
+static void bcm_rx_thr_handler(unsigned long data)
+{
+       struct bcm_op *op = (struct bcm_op *)data;
+       int i = 0;
+
+       /* mark disabled / consumed timer */
+       op->thrtimer.expires = 0;
+
+       if (op->nframes > 1) {
+               /* for MUX filter we start at index 1 */
+               for (i = 1; i < op->nframes; i++) {
+                       if ((op->last_frames) &&
+                           (op->last_frames[i].can_dlc & RX_THR)) {
+                               op->last_frames[i].can_dlc &= ~RX_THR;
+                               bcm_rx_changed(op, &op->last_frames[i]);
+                       }
+               }
+
+       } else {
+               /* for RX_FILTER_ID and simple filter */
+               if (op->last_frames && (op->last_frames[0].can_dlc & RX_THR)) {
+                       op->last_frames[0].can_dlc &= ~RX_THR;
+                       bcm_rx_changed(op, &op->last_frames[0]);
+               }
+       }
+}
+
+/*
+ * bcm_rx_handler - handle a CAN frame receiption
+ */
+static void bcm_rx_handler(struct sk_buff *skb, void *data)
+{
+       struct bcm_op *op = (struct bcm_op *)data;
+       struct can_frame rxframe;
+       int i;
+
+       /* disable timeout */
+       del_timer(&op->timer);
+
+       if (skb->len == sizeof(rxframe)) {
+               memcpy(&rxframe, skb->data, sizeof(rxframe));
+               /* save rx timestamp */
+               op->rx_stamp = skb->tstamp;
+               /* save originator for recvfrom() */
+               op->rx_ifindex = skb->dev->ifindex;
+               /* update statistics */
+               op->frames_abs++;
+               kfree_skb(skb);
+
+       } else {
+               kfree_skb(skb);
+               return;
+       }
+
+       if (op->can_id != rxframe.can_id)
+               return;
+
+       if (op->flags & RX_RTR_FRAME) {
+               /* send reply for RTR-request (placed in op->frames[0]) */
+               bcm_can_tx(op);
+               return;
+       }
+
+       if (op->flags & RX_FILTER_ID) {
+               /* the easiest case */
+               bcm_rx_update_and_send(op, &op->last_frames[0], &rxframe);
+               bcm_rx_starttimer(op);
+               return;
+       }
+
+       if (op->nframes == 1) {
+               /* simple compare with index 0 */
+               bcm_rx_cmp_to_index(op, 0, &rxframe);
+               bcm_rx_starttimer(op);
+               return;
+       }
+
+       if (op->nframes > 1) {
+               /*
+                * multiplex compare
+                *
+                * find the first multiplex mask that fits.
+                * Remark: The MUX-mask is stored in index 0
+                */
+
+               for (i = 1; i < op->nframes; i++) {
+                       if ((GET_U64(&op->frames[0]) & GET_U64(&rxframe)) ==
+                           (GET_U64(&op->frames[0]) &
+                            GET_U64(&op->frames[i]))) {
+                               bcm_rx_cmp_to_index(op, i, &rxframe);
+                               break;
+                       }
+               }
+               bcm_rx_starttimer(op);
+       }
+}
+
+/*
+ * helpers for bcm_op handling: find & delete bcm [rx|tx] op elements
+ */
+static struct bcm_op *bcm_find_op(struct list_head *ops, canid_t can_id,
+                                 int ifindex)
+{
+       struct bcm_op *op;
+
+       list_for_each_entry(op, ops, list) {
+               if ((op->can_id == can_id) && (op->ifindex == ifindex))
+                       return op;
+       }
+
+       return NULL;
+}
+
+static void bcm_remove_op(struct bcm_op *op)
+{
+       del_timer(&op->timer);
+       del_timer(&op->thrtimer);
+
+       if ((op->frames) && (op->frames != &op->sframe))
+               kfree(op->frames);
+
+       if ((op->last_frames) && (op->last_frames != &op->last_sframe))
+               kfree(op->last_frames);
+
+       kfree(op);
+
+       return;
+}
+
+static void bcm_rx_unreg(struct net_device *dev, struct bcm_op *op)
+{
+       if (op->rx_reg_dev == dev) {
+               can_rx_unregister(dev, op->can_id, REGMASK(op->can_id),
+                                 bcm_rx_handler, op);
+
+               /* mark as removed subscription */
+               op->rx_reg_dev = NULL;
+       } else
+               printk(KERN_ERR "can-bcm: bcm_rx_unreg: registered device "
+                      "mismatch %p %p\n", op->rx_reg_dev, dev);
+}
+
+/*
+ * bcm_delete_rx_op - find and remove a rx op (returns number of removed ops)
+ */
+static int bcm_delete_rx_op(struct list_head *ops, canid_t can_id, int ifindex)
+{
+       struct bcm_op *op, *n;
+
+       list_for_each_entry_safe(op, n, ops, list) {
+               if ((op->can_id == can_id) && (op->ifindex == ifindex)) {
+
+                       /*
+                        * Don't care if we're bound or not (due to netdev
+                        * problems) can_rx_unregister() is always a save
+                        * thing to do here.
+                        */
+                       if (op->ifindex) {
+                               /*
+                                * Only remove subscriptions that had not
+                                * been removed due to NETDEV_UNREGISTER
+                                * in bcm_notifier()
+                                */
+                               if (op->rx_reg_dev) {
+                                       struct net_device *dev;
+
+                                       dev = dev_get_by_index(&init_net,
+                                                              op->ifindex);
+                                       if (dev) {
+                                               bcm_rx_unreg(dev, op);
+                                               dev_put(dev);
+                                       }
+                               }
+                       } else
+                               can_rx_unregister(NULL, op->can_id,
+                                                 REGMASK(op->can_id),
+                                                 bcm_rx_handler, op);
+
+                       list_del(&op->list);
+                       bcm_remove_op(op);
+                       return 1; /* done */
+               }
+       }
+
+       return 0; /* not found */
+}
+
+/*
+ * bcm_delete_tx_op - find and remove a tx op (returns number of removed ops)
+ */
+static int bcm_delete_tx_op(struct list_head *ops, canid_t can_id, int ifindex)
+{
+       struct bcm_op *op, *n;
+
+       list_for_each_entry_safe(op, n, ops, list) {
+               if ((op->can_id == can_id) && (op->ifindex == ifindex)) {
+                       list_del(&op->list);
+                       bcm_remove_op(op);
+                       return 1; /* done */
+               }
+       }
+
+       return 0; /* not found */
+}
+
+/*
+ * bcm_read_op - read out a bcm_op and send it to the user (for bcm_sendmsg)
+ */
+static int bcm_read_op(struct list_head *ops, struct bcm_msg_head *msg_head,
+                      int ifindex)
+{
+       struct bcm_op *op = bcm_find_op(ops, msg_head->can_id, ifindex);
+
+       if (!op)
+               return -EINVAL;
+
+       /* put current values into msg_head */
+       msg_head->flags   = op->flags;
+       msg_head->count   = op->count;
+       msg_head->ival1   = op->ival1;
+       msg_head->ival2   = op->ival2;
+       msg_head->nframes = op->nframes;
+
+       bcm_send_to_user(op, msg_head, op->frames, 0);
+
+       return MHSIZ;
+}
+
+/*
+ * bcm_tx_setup - create or update a bcm tx op (for bcm_sendmsg)
+ */
+static int bcm_tx_setup(struct bcm_msg_head *msg_head, struct msghdr *msg,
+                       int ifindex, struct sock *sk)
+{
+       struct bcm_sock *bo = bcm_sk(sk);
+       struct bcm_op *op;
+       int i, err;
+
+       /* we need a real device to send frames */
+       if (!ifindex)
+               return -ENODEV;
+
+       /* we need at least one can_frame */
+       if (msg_head->nframes < 1)
+               return -EINVAL;
+
+       /* check the given can_id */
+       op = bcm_find_op(&bo->tx_ops, msg_head->can_id, ifindex);
+
+       if (op) {
+               /* update existing BCM operation */
+
+               /*
+                * Do we need more space for the can_frames than currently
+                * allocated? -> This is a _really_ unusual use-case and
+                * therefore (complexity / locking) it is not supported.
+                */
+               if (msg_head->nframes > op->nframes)
+                       return -E2BIG;
+
+               /* update can_frames content */
+               for (i = 0; i < msg_head->nframes; i++) {
+                       err = memcpy_fromiovec((u8 *)&op->frames[i],
+                                              msg->msg_iov, CFSIZ);
+                       if (err < 0)
+                               return err;
+
+                       if (msg_head->flags & TX_CP_CAN_ID) {
+                               /* copy can_id into frame */
+                               op->frames[i].can_id = msg_head->can_id;
+                       }
+               }
+
+       } else {
+               /* insert new BCM operation for the given can_id */
+
+               op = kzalloc(OPSIZ, GFP_KERNEL);
+               if (!op)
+                       return -ENOMEM;
+
+               op->can_id    = msg_head->can_id;
+
+               /* create array for can_frames and copy the data */
+               if (msg_head->nframes > 1) {
+                       op->frames = kmalloc(msg_head->nframes * CFSIZ,
+                                            GFP_KERNEL);
+                       if (!op->frames) {
+                               kfree(op);
+                               return -ENOMEM;
+                       }
+               } else
+                       op->frames = &op->sframe;
+
+               for (i = 0; i < msg_head->nframes; i++) {
+                       err = memcpy_fromiovec((u8 *)&op->frames[i],
+                                              msg->msg_iov, CFSIZ);
+                       if (err < 0) {
+                               if (op->frames != &op->sframe)
+                                       kfree(op->frames);
+                               kfree(op);
+                               return err;
+                       }
+
+                       if (msg_head->flags & TX_CP_CAN_ID) {
+                               /* copy can_id into frame */
+                               op->frames[i].can_id = msg_head->can_id;
+                       }
+               }
+
+               /* tx_ops never compare with previous received messages */
+               op->last_frames = NULL;
+
+               /* bcm_can_tx / bcm_tx_timeout_handler needs this */
+               op->sk = sk;
+               op->ifindex = ifindex;
+
+               /* initialize uninitialized (kzalloc) structure */
+               setup_timer(&op->timer, bcm_tx_timeout_handler,
+                           (unsigned long)op);
+
+               /* currently unused in tx_ops */
+               init_timer(&op->thrtimer);
+
+               /* add this bcm_op to the list of the tx_ops */
+               list_add(&op->list, &bo->tx_ops);
+
+       } /* if ((op = bcm_find_op(&bo->tx_ops, msg_head->can_id, ifindex))) */
+
+       if (op->nframes != msg_head->nframes) {
+               op->nframes   = msg_head->nframes;
+               /* start multiple frame transmission with index 0 */
+               op->currframe = 0;
+       }
+
+       /* check flags */
+
+       op->flags = msg_head->flags;
+
+       if (op->flags & TX_RESET_MULTI_IDX) {
+               /* start multiple frame transmission with index 0 */
+               op->currframe = 0;
+       }
+
+       if (op->flags & SETTIMER) {
+               /* set timer values */
+               op->count = msg_head->count;
+               op->ival1 = msg_head->ival1;
+               op->ival2 = msg_head->ival2;
+               op->j_ival1 = rounded_tv2jif(&msg_head->ival1);
+               op->j_ival2 = rounded_tv2jif(&msg_head->ival2);
+
+               /* disable an active timer due to zero values? */
+               if (!op->j_ival1 && !op->j_ival2)
+                       del_timer(&op->timer);
+       }
+
+       if ((op->flags & STARTTIMER) &&
+           ((op->j_ival1 && op->count) || op->j_ival2)) {
+
+               /* spec: send can_frame when starting timer */
+               op->flags |= TX_ANNOUNCE;
+
+               if (op->j_ival1 && (op->count > 0)) {
+                       /* op->count-- is done in bcm_tx_timeout_handler */
+                       mod_timer(&op->timer, jiffies + op->j_ival1);
+               } else
+                       mod_timer(&op->timer, jiffies + op->j_ival2);
+       }
+
+       if (op->flags & TX_ANNOUNCE)
+               bcm_can_tx(op);
+
+       return msg_head->nframes * CFSIZ + MHSIZ;
+}
+
+/*
+ * bcm_rx_setup - create or update a bcm rx op (for bcm_sendmsg)
+ */
+static int bcm_rx_setup(struct bcm_msg_head *msg_head, struct msghdr *msg,
+                       int ifindex, struct sock *sk)
+{
+       struct bcm_sock *bo = bcm_sk(sk);
+       struct bcm_op *op;
+       int do_rx_register;
+       int err = 0;
+
+       if ((msg_head->flags & RX_FILTER_ID) || (!(msg_head->nframes))) {
+               /* be robust against wrong usage ... */
+               msg_head->flags |= RX_FILTER_ID;
+               /* ignore trailing garbage */
+               msg_head->nframes = 0;
+       }
+
+       if ((msg_head->flags & RX_RTR_FRAME) &&
+           ((msg_head->nframes != 1) ||
+            (!(msg_head->can_id & CAN_RTR_FLAG))))
+               return -EINVAL;
+
+       /* check the given can_id */
+       op = bcm_find_op(&bo->rx_ops, msg_head->can_id, ifindex);
+       if (op) {
+               /* update existing BCM operation */
+
+               /*
+                * Do we need more space for the can_frames than currently
+                * allocated? -> This is a _really_ unusual use-case and
+                * therefore (complexity / locking) it is not supported.
+                */
+               if (msg_head->nframes > op->nframes)
+                       return -E2BIG;
+
+               if (msg_head->nframes) {
+                       /* update can_frames content */
+                       err = memcpy_fromiovec((u8 *)op->frames,
+                                              msg->msg_iov,
+                                              msg_head->nframes * CFSIZ);
+                       if (err < 0)
+                               return err;
+
+                       /* clear last_frames to indicate 'nothing received' */
+                       memset(op->last_frames, 0, msg_head->nframes * CFSIZ);
+               }
+
+               op->nframes = msg_head->nframes;
+
+               /* Only an update -> do not call can_rx_register() */
+               do_rx_register = 0;
+
+       } else {
+               /* insert new BCM operation for the given can_id */
+               op = kzalloc(OPSIZ, GFP_KERNEL);
+               if (!op)
+                       return -ENOMEM;
+
+               op->can_id    = msg_head->can_id;
+               op->nframes   = msg_head->nframes;
+
+               if (msg_head->nframes > 1) {
+                       /* create array for can_frames and copy the data */
+                       op->frames = kmalloc(msg_head->nframes * CFSIZ,
+                                            GFP_KERNEL);
+                       if (!op->frames) {
+                               kfree(op);
+                               return -ENOMEM;
+                       }
+
+                       /* create and init array for received can_frames */
+                       op->last_frames = kzalloc(msg_head->nframes * CFSIZ,
+                                                 GFP_KERNEL);
+                       if (!op->last_frames) {
+                               kfree(op->frames);
+                               kfree(op);
+                               return -ENOMEM;
+                       }
+
+               } else {
+                       op->frames = &op->sframe;
+                       op->last_frames = &op->last_sframe;
+               }
+
+               if (msg_head->nframes) {
+                       err = memcpy_fromiovec((u8 *)op->frames, msg->msg_iov,
+                                              msg_head->nframes * CFSIZ);
+                       if (err < 0) {
+                               if (op->frames != &op->sframe)
+                                       kfree(op->frames);
+                               if (op->last_frames != &op->last_sframe)
+                                       kfree(op->last_frames);
+                               kfree(op);
+                               return err;
+                       }
+               }
+
+               /* bcm_can_tx / bcm_tx_timeout_handler needs this */
+               op->sk = sk;
+               op->ifindex = ifindex;
+
+               /* initialize uninitialized (kzalloc) structure */
+               setup_timer(&op->timer, bcm_rx_timeout_handler,
+                           (unsigned long)op);
+
+               /* init throttle timer for RX_CHANGED */
+               setup_timer(&op->thrtimer, bcm_rx_thr_handler,
+                           (unsigned long)op);
+
+               /* mark disabled timer */
+               op->thrtimer.expires = 0;
+
+               /* add this bcm_op to the list of the rx_ops */
+               list_add(&op->list, &bo->rx_ops);
+
+               /* call can_rx_register() */
+               do_rx_register = 1;
+
+       } /* if ((op = bcm_find_op(&bo->rx_ops, msg_head->can_id, ifindex))) */
+
+       /* check flags */
+       op->flags = msg_head->flags;
+
+       if (op->flags & RX_RTR_FRAME) {
+
+               /* no timers in RTR-mode */
+               del_timer(&op->thrtimer);
+               del_timer(&op->timer);
+
+               /*
+                * funny feature in RX(!)_SETUP only for RTR-mode:
+                * copy can_id into frame BUT without RTR-flag to
+                * prevent a full-load-loopback-test ... ;-]
+                */
+               if ((op->flags & TX_CP_CAN_ID) ||
+                   (op->frames[0].can_id == op->can_id))
+                       op->frames[0].can_id = op->can_id & ~CAN_RTR_FLAG;
+
+       } else {
+               if (op->flags & SETTIMER) {
+
+                       /* set timer value */
+                       op->ival1 = msg_head->ival1;
+                       op->ival2 = msg_head->ival2;
+                       op->j_ival1 = rounded_tv2jif(&msg_head->ival1);
+                       op->j_ival2 = rounded_tv2jif(&msg_head->ival2);
+
+                       /* disable an active timer due to zero value? */
+                       if (!op->j_ival1)
+                               del_timer(&op->timer);
+
+                       /* free currently blocked msgs ? */
+                       if (op->thrtimer.expires) {
+                               /* send blocked msgs hereafter */
+                               mod_timer(&op->thrtimer, jiffies + 2);
+                       }
+
+                       /*
+                        * if (op->j_ival2) is zero, no (new) throttling
+                        * will happen. For details see functions
+                        * bcm_rx_update_and_send() and bcm_rx_thr_handler()
+                        */
+               }
+
+               if ((op->flags & STARTTIMER) && op->j_ival1)
+                       mod_timer(&op->timer, jiffies + op->j_ival1);
+       }
+
+       /* now we can register for can_ids, if we added a new bcm_op */
+       if (do_rx_register) {
+               if (ifindex) {
+                       struct net_device *dev;
+
+                       dev = dev_get_by_index(&init_net, ifindex);
+                       if (dev) {
+                               err = can_rx_register(dev, op->can_id,
+                                                     REGMASK(op->can_id),
+                                                     bcm_rx_handler, op,
+                                                     "bcm");
+
+                               op->rx_reg_dev = dev;
+                               dev_put(dev);
+                       }
+
+               } else
+                       err = can_rx_register(NULL, op->can_id,
+                                             REGMASK(op->can_id),
+                                             bcm_rx_handler, op, "bcm");
+               if (err) {
+                       /* this bcm rx op is broken -> remove it */
+                       list_del(&op->list);
+                       bcm_remove_op(op);
+                       return err;
+               }
+       }
+
+       return msg_head->nframes * CFSIZ + MHSIZ;
+}
+
+/*
+ * bcm_tx_send - send a single CAN frame to the CAN interface (for bcm_sendmsg)
+ */
+static int bcm_tx_send(struct msghdr *msg, int ifindex, struct sock *sk)
+{
+       struct sk_buff *skb;
+       struct net_device *dev;
+       int err;
+
+       /* we need a real device to send frames */
+       if (!ifindex)
+               return -ENODEV;
+
+       skb = alloc_skb(CFSIZ, GFP_KERNEL);
+
+       if (!skb)
+               return -ENOMEM;
+
+       err = memcpy_fromiovec(skb_put(skb, CFSIZ), msg->msg_iov, CFSIZ);
+       if (err < 0) {
+               kfree_skb(skb);
+               return err;
+       }
+
+       dev = dev_get_by_index(&init_net, ifindex);
+       if (!dev) {
+               kfree_skb(skb);
+               return -ENODEV;
+       }
+
+       skb->dev = dev;
+       skb->sk  = sk;
+       can_send(skb, 1); /* send with loopback */
+       dev_put(dev);
+
+       return CFSIZ + MHSIZ;
+}
+
+/*
+ * bcm_sendmsg - process BCM commands (opcodes) from the userspace
+ */
+static int bcm_sendmsg(struct kiocb *iocb, struct socket *sock,
+                      struct msghdr *msg, size_t size)
+{
+       struct sock *sk = sock->sk;
+       struct bcm_sock *bo = bcm_sk(sk);
+       int ifindex = bo->ifindex; /* default ifindex for this bcm_op */
+       struct bcm_msg_head msg_head;
+       int ret; /* read bytes or error codes as return value */
+
+       if (!bo->bound)
+               return -ENOTCONN;
+
+       /* check for alternative ifindex for this bcm_op */
+
+       if (!ifindex && msg->msg_name) {
+               /* no bound device as default => check msg_name */
+               struct sockaddr_can *addr =
+                       (struct sockaddr_can *)msg->msg_name;
+
+               if (addr->can_family != AF_CAN)
+                       return -EINVAL;
+
+               /* ifindex from sendto() */
+               ifindex = addr->can_ifindex;
+
+               if (ifindex) {
+                       struct net_device *dev;
+
+                       dev = dev_get_by_index(&init_net, ifindex);
+                       if (!dev)
+                               return -ENODEV;
+
+                       if (dev->type != ARPHRD_CAN) {
+                               dev_put(dev);
+                               return -ENODEV;
+                       }
+
+                       dev_put(dev);
+               }
+       }
+
+       /* read message head information */
+
+       ret = memcpy_fromiovec((u8 *)&msg_head, msg->msg_iov, MHSIZ);
+       if (ret < 0)
+               return ret;
+
+       lock_sock(sk);
+
+       switch (msg_head.opcode) {
+
+       case TX_SETUP:
+               ret = bcm_tx_setup(&msg_head, msg, ifindex, sk);
+               break;
+
+       case RX_SETUP:
+               ret = bcm_rx_setup(&msg_head, msg, ifindex, sk);
+               break;
+
+       case TX_DELETE:
+               if (bcm_delete_tx_op(&bo->tx_ops, msg_head.can_id, ifindex))
+                       ret = MHSIZ;
+               else
+                       ret = -EINVAL;
+               break;
+
+       case RX_DELETE:
+               if (bcm_delete_rx_op(&bo->rx_ops, msg_head.can_id, ifindex))
+                       ret = MHSIZ;
+               else
+                       ret = -EINVAL;
+               break;
+
+       case TX_READ:
+               /* reuse msg_head for the reply to TX_READ */
+               msg_head.opcode  = TX_STATUS;
+               ret = bcm_read_op(&bo->tx_ops, &msg_head, ifindex);
+               break;
+
+       case RX_READ:
+               /* reuse msg_head for the reply to RX_READ */
+               msg_head.opcode  = RX_STATUS;
+               ret = bcm_read_op(&bo->rx_ops, &msg_head, ifindex);
+               break;
+
+       case TX_SEND:
+               /* we need at least one can_frame */
+               if (msg_head.nframes < 1)
+                       ret = -EINVAL;
+               else
+                       ret = bcm_tx_send(msg, ifindex, sk);
+               break;
+
+       default:
+               ret = -EINVAL;
+               break;
+       }
+
+       release_sock(sk);
+
+       return ret;
+}
+
+/*
+ * notification handler for netdevice status changes
+ */
+static int bcm_notifier(struct notifier_block *nb, unsigned long msg,
+                       void *data)
+{
+       struct net_device *dev = (struct net_device *)data;
+       struct bcm_sock *bo = container_of(nb, struct bcm_sock, notifier);
+       struct sock *sk = &bo->sk;
+       struct bcm_op *op;
+       int notify_enodev = 0;
+
+       if (dev->nd_net != &init_net)
+               return NOTIFY_DONE;
+
+       if (dev->type != ARPHRD_CAN)
+               return NOTIFY_DONE;
+
+       switch (msg) {
+
+       case NETDEV_UNREGISTER:
+               lock_sock(sk);
+
+               /* remove device specific receive entries */
+               list_for_each_entry(op, &bo->rx_ops, list)
+                       if (op->rx_reg_dev == dev)
+                               bcm_rx_unreg(dev, op);
+
+               /* remove device reference, if this is our bound device */
+               if (bo->bound && bo->ifindex == dev->ifindex) {
+                       bo->bound   = 0;
+                       bo->ifindex = 0;
+                       notify_enodev = 1;
+               }
+
+               release_sock(sk);
+
+               if (notify_enodev) {
+                       sk->sk_err = ENODEV;
+                       if (!sock_flag(sk, SOCK_DEAD))
+                               sk->sk_error_report(sk);
+               }
+               break;
+
+       case NETDEV_DOWN:
+               if (bo->bound && bo->ifindex == dev->ifindex) {
+                       sk->sk_err = ENETDOWN;
+                       if (!sock_flag(sk, SOCK_DEAD))
+                               sk->sk_error_report(sk);
+               }
+       }
+
+       return NOTIFY_DONE;
+}
+
+/*
+ * initial settings for all BCM sockets to be set at socket creation time
+ */
+static int bcm_init(struct sock *sk)
+{
+       struct bcm_sock *bo = bcm_sk(sk);
+
+       bo->bound            = 0;
+       bo->ifindex          = 0;
+       bo->dropped_usr_msgs = 0;
+       bo->bcm_proc_read    = NULL;
+
+       INIT_LIST_HEAD(&bo->tx_ops);
+       INIT_LIST_HEAD(&bo->rx_ops);
+
+       /* set notifier */
+       bo->notifier.notifier_call = bcm_notifier;
+
+       register_netdevice_notifier(&bo->notifier);
+
+       return 0;
+}
+
+/*
+ * standard socket functions
+ */
+static int bcm_release(struct socket *sock)
+{
+       struct sock *sk = sock->sk;
+       struct bcm_sock *bo = bcm_sk(sk);
+       struct bcm_op *op, *next;
+
+       /* remove bcm_ops, timer, rx_unregister(), etc. */
+
+       unregister_netdevice_notifier(&bo->notifier);
+
+       lock_sock(sk);
+
+       list_for_each_entry_safe(op, next, &bo->tx_ops, list)
+               bcm_remove_op(op);
+
+       list_for_each_entry_safe(op, next, &bo->rx_ops, list) {
+               /*
+                * Don't care if we're bound or not (due to netdev problems)
+                * can_rx_unregister() is always a save thing to do here.
+                */
+               if (op->ifindex) {
+                       /*
+                        * Only remove subscriptions that had not
+                        * been removed due to NETDEV_UNREGISTER
+                        * in bcm_notifier()
+                        */
+                       if (op->rx_reg_dev) {
+                               struct net_device *dev;
+
+                               dev = dev_get_by_index(&init_net, op->ifindex);
+                               if (dev) {
+                                       bcm_rx_unreg(dev, op);
+                                       dev_put(dev);
+                               }
+                       }
+               } else
+                       can_rx_unregister(NULL, op->can_id,
+                                         REGMASK(op->can_id),
+                                         bcm_rx_handler, op);
+
+               bcm_remove_op(op);
+       }
+
+       /* remove procfs entry */
+       if (proc_dir && bo->bcm_proc_read)
+               remove_proc_entry(bo->procname, proc_dir);
+
+       /* remove device reference */
+       if (bo->bound) {
+               bo->bound   = 0;
+               bo->ifindex = 0;
+       }
+
+       release_sock(sk);
+       sock_put(sk);
+
+       return 0;
+}
+
+static int bcm_connect(struct socket *sock, struct sockaddr *uaddr, int len,
+                      int flags)
+{
+       struct sockaddr_can *addr = (struct sockaddr_can *)uaddr;
+       struct sock *sk = sock->sk;
+       struct bcm_sock *bo = bcm_sk(sk);
+
+       if (bo->bound)
+               return -EISCONN;
+
+       /* bind a device to this socket */
+       if (addr->can_ifindex) {
+               struct net_device *dev;
+
+               dev = dev_get_by_index(&init_net, addr->can_ifindex);
+               if (!dev)
+                       return -ENODEV;
+
+               if (dev->type != ARPHRD_CAN) {
+                       dev_put(dev);
+                       return -ENODEV;
+               }
+
+               bo->ifindex = dev->ifindex;
+               dev_put(dev);
+
+       } else {
+               /* no interface reference for ifindex = 0 ('any' CAN device) */
+               bo->ifindex = 0;
+       }
+
+       bo->bound = 1;
+
+       if (proc_dir) {
+               /* unique socket address as filename */
+               sprintf(bo->procname, "%p", sock);
+               bo->bcm_proc_read = create_proc_read_entry(bo->procname, 0644,
+                                                          proc_dir,
+                                                          bcm_read_proc, sk);
+       }
+
+       return 0;
+}
+
+static int bcm_recvmsg(struct kiocb *iocb, struct socket *sock,
+                      struct msghdr *msg, size_t size, int flags)
+{
+       struct sock *sk = sock->sk;
+       struct sk_buff *skb;
+       int error = 0;
+       int noblock;
+       int err;
+
+       noblock =  flags & MSG_DONTWAIT;
+       flags   &= ~MSG_DONTWAIT;
+       skb = skb_recv_datagram(sk, flags, noblock, &error);
+       if (!skb)
+               return error;
+
+       if (skb->len < size)
+               size = skb->len;
+
+       err = memcpy_toiovec(msg->msg_iov, skb->data, size);
+       if (err < 0) {
+               skb_free_datagram(sk, skb);
+               return err;
+       }
+
+       sock_recv_timestamp(msg, sk, skb);
+
+       if (msg->msg_name) {
+               msg->msg_namelen = sizeof(struct sockaddr_can);
+               memcpy(msg->msg_name, skb->cb, msg->msg_namelen);
+       }
+
+       skb_free_datagram(sk, skb);
+
+       return size;
+}
+
+static struct proto_ops bcm_ops __read_mostly = {
+       .family        = PF_CAN,
+       .release       = bcm_release,
+       .bind          = sock_no_bind,
+       .connect       = bcm_connect,
+       .socketpair    = sock_no_socketpair,
+       .accept        = sock_no_accept,
+       .getname       = sock_no_getname,
+       .poll          = datagram_poll,
+       .ioctl         = NULL,          /* use can_ioctl() from af_can.c */
+       .listen        = sock_no_listen,
+       .shutdown      = sock_no_shutdown,
+       .setsockopt    = sock_no_setsockopt,
+       .getsockopt    = sock_no_getsockopt,
+       .sendmsg       = bcm_sendmsg,
+       .recvmsg       = bcm_recvmsg,
+       .mmap          = sock_no_mmap,
+       .sendpage      = sock_no_sendpage,
+};
+
+static struct proto bcm_proto __read_mostly = {
+       .name       = "CAN_BCM",
+       .owner      = THIS_MODULE,
+       .obj_size   = sizeof(struct bcm_sock),
+       .init       = bcm_init,
+};
+
+static struct can_proto bcm_can_proto __read_mostly = {
+       .type       = SOCK_DGRAM,
+       .protocol   = CAN_BCM,
+       .capability = -1,
+       .ops        = &bcm_ops,
+       .prot       = &bcm_proto,
+};
+
+static int __init bcm_module_init(void)
+{
+       int err;
+
+       printk(banner);
+
+       err = can_proto_register(&bcm_can_proto);
+       if (err < 0) {
+               printk(KERN_ERR "can: registration of bcm protocol failed\n");
+               return err;
+       }
+
+       /* create /proc/net/can-bcm directory */
+       proc_dir = proc_mkdir("can-bcm", init_net.proc_net);
+
+       if (proc_dir)
+               proc_dir->owner = THIS_MODULE;
+
+       return 0;
+}
+
+static void __exit bcm_module_exit(void)
+{
+       can_proto_unregister(&bcm_can_proto);
+
+       if (proc_dir)
+               proc_net_remove(&init_net, "can-bcm");
+}
+
+module_init(bcm_module_init);
+module_exit(bcm_module_exit);