plip: Move the PLIP driver
authorJeff Kirsher <jeffrey.t.kirsher@intel.com>
Wed, 3 Aug 2011 10:01:58 +0000 (03:01 -0700)
committerJeff Kirsher <jeffrey.t.kirsher@intel.com>
Sat, 27 Aug 2011 07:58:33 +0000 (00:58 -0700)
Move the Parallel Line Internet Protocol (PLIP) driver into
drivers/net/plip/ and make the necessary Kconfig and Makefile changes.

CC: Niibe Yutaka <gniibe@mri.co.jp>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Acked-by: Alan Cox <alan@linux.intel.com>
drivers/net/Kconfig
drivers/net/Makefile
drivers/net/plip.c [deleted file]
drivers/net/plip/Kconfig [new file with mode: 0644]
drivers/net/plip/Makefile [new file with mode: 0644]
drivers/net/plip/plip.c [new file with mode: 0644]

index 1d8fa955d120ba7e7a000e76865f7ef4b2a82dc5..3f72686e18041de2b3b8accbba550612fc0c1d6a 100644 (file)
@@ -197,6 +197,8 @@ source "drivers/net/ethernet/Kconfig"
 
 source "drivers/net/fddi/Kconfig"
 
+source "drivers/net/plip/Kconfig"
+
 source "drivers/net/tokenring/Kconfig"
 
 source "drivers/net/wireless/Kconfig"
@@ -272,41 +274,6 @@ config RIONET_RX_SIZE
        depends on RIONET
        default "128"
 
-config PLIP
-       tristate "PLIP (parallel port) support"
-       depends on PARPORT
-       ---help---
-         PLIP (Parallel Line Internet Protocol) is used to create a
-         reasonably fast mini network consisting of two (or, rarely, more)
-         local machines.  A PLIP link from a Linux box is a popular means to
-         install a Linux distribution on a machine which doesn't have a
-         CD-ROM drive (a minimal system has to be transferred with floppies
-         first). The kernels on both machines need to have this PLIP option
-         enabled for this to work.
-
-         The PLIP driver has two modes, mode 0 and mode 1.  The parallel
-         ports (the connectors at the computers with 25 holes) are connected
-         with "null printer" or "Turbo Laplink" cables which can transmit 4
-         bits at a time (mode 0) or with special PLIP cables, to be used on
-         bidirectional parallel ports only, which can transmit 8 bits at a
-         time (mode 1); you can find the wiring of these cables in
-         <file:Documentation/networking/PLIP.txt>.  The cables can be up to
-         15m long.  Mode 0 works also if one of the machines runs DOS/Windows
-         and has some PLIP software installed, e.g. the Crynwr PLIP packet
-         driver (<http://oak.oakland.edu/simtel.net/msdos/pktdrvr-pre.html>)
-         and winsock or NCSA's telnet.
-
-         If you want to use PLIP, say Y and read the PLIP mini-HOWTO as well
-         as the NET-3-HOWTO, both available from
-         <http://www.tldp.org/docs.html#howto>.  Note that the PLIP
-         protocol has been changed and this PLIP driver won't work together
-         with the PLIP support in Linux versions 1.0.x.  This option enlarges
-         your kernel by about 8 KB.
-
-         To compile this driver as a module, choose M here. The module
-         will be called plip. If unsure, say Y or M, in case you buy
-         a laptop later.
-
 config SLIP
        tristate "SLIP (serial line) support"
        ---help---
index f64d02ce1ed90c0dc07ea38eeca42665bb5c0375..52dae95d4ea6b49d1562ceb95b5778a2093904c6 100644 (file)
@@ -12,7 +12,6 @@ obj-$(CONFIG_VMXNET3) += vmxnet3/
 #
 # link order important here
 #
-obj-$(CONFIG_PLIP) += plip.o
 obj-$(CONFIG_RIONET) += rionet.o
 
 #
@@ -39,6 +38,7 @@ obj-$(CONFIG_DEV_APPLETALK) += appletalk/
 obj-$(CONFIG_ETHERNET) += ethernet/
 obj-$(CONFIG_FDDI) += fddi/
 obj-$(CONFIG_HIPPI) += hippi/
+obj-$(CONFIG_PLIP) += plip/
 onj-$(CONFIG_PPP) += ppp/
 obj-$(CONFIG_PPP_ASYNC) += ppp/
 obj-$(CONFIG_PPP_BSDCOMP) += ppp/
diff --git a/drivers/net/plip.c b/drivers/net/plip.c
deleted file mode 100644 (file)
index a9e9ca8..0000000
+++ /dev/null
@@ -1,1403 +0,0 @@
-/* $Id: plip.c,v 1.3.6.2 1997/04/16 15:07:56 phil Exp $ */
-/* PLIP: A parallel port "network" driver for Linux. */
-/* This driver is for parallel port with 5-bit cable (LapLink (R) cable). */
-/*
- * Authors:    Donald Becker <becker@scyld.com>
- *             Tommy Thorn <thorn@daimi.aau.dk>
- *             Tanabe Hiroyasu <hiro@sanpo.t.u-tokyo.ac.jp>
- *             Alan Cox <gw4pts@gw4pts.ampr.org>
- *             Peter Bauer <100136.3530@compuserve.com>
- *             Niibe Yutaka <gniibe@mri.co.jp>
- *             Nimrod Zimerman <zimerman@mailandnews.com>
- *
- * Enhancements:
- *             Modularization and ifreq/ifmap support by Alan Cox.
- *             Rewritten by Niibe Yutaka.
- *             parport-sharing awareness code by Philip Blundell.
- *             SMP locking by Niibe Yutaka.
- *             Support for parallel ports with no IRQ (poll mode),
- *             Modifications to use the parallel port API
- *             by Nimrod Zimerman.
- *
- * Fixes:
- *             Niibe Yutaka
- *               - Module initialization.
- *               - MTU fix.
- *               - Make sure other end is OK, before sending a packet.
- *               - Fix immediate timer problem.
- *
- *             Al Viro
- *               - Changed {enable,disable}_irq handling to make it work
- *                 with new ("stack") semantics.
- *
- *             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.
- */
-
-/*
- * Original version and the name 'PLIP' from Donald Becker <becker@scyld.com>
- * inspired by Russ Nelson's parallel port packet driver.
- *
- * NOTE:
- *     Tanabe Hiroyasu had changed the protocol, and it was in Linux v1.0.
- *     Because of the necessity to communicate to DOS machines with the
- *     Crynwr packet driver, Peter Bauer changed the protocol again
- *     back to original protocol.
- *
- *     This version follows original PLIP protocol.
- *     So, this PLIP can't communicate the PLIP of Linux v1.0.
- */
-
-/*
- *     To use with DOS box, please do (Turn on ARP switch):
- *     # ifconfig plip[0-2] arp
- */
-static const char version[] = "NET3 PLIP version 2.4-parport gniibe@mri.co.jp\n";
-
-/*
-  Sources:
-       Ideas and protocols came from Russ Nelson's <nelson@crynwr.com>
-       "parallel.asm" parallel port packet driver.
-
-  The "Crynwr" parallel port standard specifies the following protocol:
-    Trigger by sending nibble '0x8' (this causes interrupt on other end)
-    count-low octet
-    count-high octet
-    ... data octets
-    checksum octet
-  Each octet is sent as <wait for rx. '0x1?'> <send 0x10+(octet&0x0F)>
-                       <wait for rx. '0x0?'> <send 0x00+((octet>>4)&0x0F)>
-
-  The packet is encapsulated as if it were ethernet.
-
-  The cable used is a de facto standard parallel null cable -- sold as
-  a "LapLink" cable by various places.  You'll need a 12-conductor cable to
-  make one yourself.  The wiring is:
-    SLCTIN     17 - 17
-    GROUND     25 - 25
-    D0->ERROR  2 - 15          15 - 2
-    D1->SLCT   3 - 13          13 - 3
-    D2->PAPOUT 4 - 12          12 - 4
-    D3->ACK    5 - 10          10 - 5
-    D4->BUSY   6 - 11          11 - 6
-  Do not connect the other pins.  They are
-    D5,D6,D7 are 7,8,9
-    STROBE is 1, FEED is 14, INIT is 16
-    extra grounds are 18,19,20,21,22,23,24
-*/
-
-#include <linux/module.h>
-#include <linux/kernel.h>
-#include <linux/types.h>
-#include <linux/fcntl.h>
-#include <linux/interrupt.h>
-#include <linux/string.h>
-#include <linux/slab.h>
-#include <linux/if_ether.h>
-#include <linux/in.h>
-#include <linux/errno.h>
-#include <linux/delay.h>
-#include <linux/init.h>
-#include <linux/netdevice.h>
-#include <linux/etherdevice.h>
-#include <linux/inetdevice.h>
-#include <linux/skbuff.h>
-#include <linux/if_plip.h>
-#include <linux/workqueue.h>
-#include <linux/spinlock.h>
-#include <linux/completion.h>
-#include <linux/parport.h>
-#include <linux/bitops.h>
-
-#include <net/neighbour.h>
-
-#include <asm/system.h>
-#include <asm/irq.h>
-#include <asm/byteorder.h>
-
-/* Maximum number of devices to support. */
-#define PLIP_MAX  8
-
-/* Use 0 for production, 1 for verification, >2 for debug */
-#ifndef NET_DEBUG
-#define NET_DEBUG 1
-#endif
-static const unsigned int net_debug = NET_DEBUG;
-
-#define ENABLE(irq)  if (irq != -1) enable_irq(irq)
-#define DISABLE(irq) if (irq != -1) disable_irq(irq)
-
-/* In micro second */
-#define PLIP_DELAY_UNIT                   1
-
-/* Connection time out = PLIP_TRIGGER_WAIT * PLIP_DELAY_UNIT usec */
-#define PLIP_TRIGGER_WAIT       500
-
-/* Nibble time out = PLIP_NIBBLE_WAIT * PLIP_DELAY_UNIT usec */
-#define PLIP_NIBBLE_WAIT        3000
-
-/* Bottom halves */
-static void plip_kick_bh(struct work_struct *work);
-static void plip_bh(struct work_struct *work);
-static void plip_timer_bh(struct work_struct *work);
-
-/* Interrupt handler */
-static void plip_interrupt(void *dev_id);
-
-/* Functions for DEV methods */
-static int plip_tx_packet(struct sk_buff *skb, struct net_device *dev);
-static int plip_hard_header(struct sk_buff *skb, struct net_device *dev,
-                            unsigned short type, const void *daddr,
-                           const void *saddr, unsigned len);
-static int plip_hard_header_cache(const struct neighbour *neigh,
-                                  struct hh_cache *hh, __be16 type);
-static int plip_open(struct net_device *dev);
-static int plip_close(struct net_device *dev);
-static int plip_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd);
-static int plip_preempt(void *handle);
-static void plip_wakeup(void *handle);
-
-enum plip_connection_state {
-       PLIP_CN_NONE=0,
-       PLIP_CN_RECEIVE,
-       PLIP_CN_SEND,
-       PLIP_CN_CLOSING,
-       PLIP_CN_ERROR
-};
-
-enum plip_packet_state {
-       PLIP_PK_DONE=0,
-       PLIP_PK_TRIGGER,
-       PLIP_PK_LENGTH_LSB,
-       PLIP_PK_LENGTH_MSB,
-       PLIP_PK_DATA,
-       PLIP_PK_CHECKSUM
-};
-
-enum plip_nibble_state {
-       PLIP_NB_BEGIN,
-       PLIP_NB_1,
-       PLIP_NB_2,
-};
-
-struct plip_local {
-       enum plip_packet_state state;
-       enum plip_nibble_state nibble;
-       union {
-               struct {
-#if defined(__LITTLE_ENDIAN)
-                       unsigned char lsb;
-                       unsigned char msb;
-#elif defined(__BIG_ENDIAN)
-                       unsigned char msb;
-                       unsigned char lsb;
-#else
-#error "Please fix the endianness defines in <asm/byteorder.h>"
-#endif
-               } b;
-               unsigned short h;
-       } length;
-       unsigned short byte;
-       unsigned char  checksum;
-       unsigned char  data;
-       struct sk_buff *skb;
-};
-
-struct net_local {
-       struct net_device *dev;
-       struct work_struct immediate;
-       struct delayed_work deferred;
-       struct delayed_work timer;
-       struct plip_local snd_data;
-       struct plip_local rcv_data;
-       struct pardevice *pardev;
-       unsigned long  trigger;
-       unsigned long  nibble;
-       enum plip_connection_state connection;
-       unsigned short timeout_count;
-       int is_deferred;
-       int port_owner;
-       int should_relinquish;
-       spinlock_t lock;
-       atomic_t kill_timer;
-       struct completion killed_timer_cmp;
-};
-
-static inline void enable_parport_interrupts (struct net_device *dev)
-{
-       if (dev->irq != -1)
-       {
-               struct parport *port =
-                  ((struct net_local *)netdev_priv(dev))->pardev->port;
-               port->ops->enable_irq (port);
-       }
-}
-
-static inline void disable_parport_interrupts (struct net_device *dev)
-{
-       if (dev->irq != -1)
-       {
-               struct parport *port =
-                  ((struct net_local *)netdev_priv(dev))->pardev->port;
-               port->ops->disable_irq (port);
-       }
-}
-
-static inline void write_data (struct net_device *dev, unsigned char data)
-{
-       struct parport *port =
-          ((struct net_local *)netdev_priv(dev))->pardev->port;
-
-       port->ops->write_data (port, data);
-}
-
-static inline unsigned char read_status (struct net_device *dev)
-{
-       struct parport *port =
-          ((struct net_local *)netdev_priv(dev))->pardev->port;
-
-       return port->ops->read_status (port);
-}
-
-static const struct header_ops plip_header_ops = {
-       .create = plip_hard_header,
-       .cache  = plip_hard_header_cache,
-};
-
-static const struct net_device_ops plip_netdev_ops = {
-       .ndo_open                = plip_open,
-       .ndo_stop                = plip_close,
-       .ndo_start_xmit          = plip_tx_packet,
-       .ndo_do_ioctl            = plip_ioctl,
-       .ndo_change_mtu          = eth_change_mtu,
-       .ndo_set_mac_address     = eth_mac_addr,
-       .ndo_validate_addr       = eth_validate_addr,
-};
-
-/* Entry point of PLIP driver.
-   Probe the hardware, and register/initialize the driver.
-
-   PLIP is rather weird, because of the way it interacts with the parport
-   system.  It is _not_ initialised from Space.c.  Instead, plip_init()
-   is called, and that function makes up a "struct net_device" for each port, and
-   then calls us here.
-
-   */
-static void
-plip_init_netdev(struct net_device *dev)
-{
-       struct net_local *nl = netdev_priv(dev);
-
-       /* Then, override parts of it */
-       dev->tx_queue_len        = 10;
-       dev->flags               = IFF_POINTOPOINT|IFF_NOARP;
-       memset(dev->dev_addr, 0xfc, ETH_ALEN);
-
-       dev->netdev_ops          = &plip_netdev_ops;
-       dev->header_ops          = &plip_header_ops;
-
-
-       nl->port_owner = 0;
-
-       /* Initialize constants */
-       nl->trigger     = PLIP_TRIGGER_WAIT;
-       nl->nibble      = PLIP_NIBBLE_WAIT;
-
-       /* Initialize task queue structures */
-       INIT_WORK(&nl->immediate, plip_bh);
-       INIT_DELAYED_WORK(&nl->deferred, plip_kick_bh);
-
-       if (dev->irq == -1)
-               INIT_DELAYED_WORK(&nl->timer, plip_timer_bh);
-
-       spin_lock_init(&nl->lock);
-}
-
-/* Bottom half handler for the delayed request.
-   This routine is kicked by do_timer().
-   Request `plip_bh' to be invoked. */
-static void
-plip_kick_bh(struct work_struct *work)
-{
-       struct net_local *nl =
-               container_of(work, struct net_local, deferred.work);
-
-       if (nl->is_deferred)
-               schedule_work(&nl->immediate);
-}
-
-/* Forward declarations of internal routines */
-static int plip_none(struct net_device *, struct net_local *,
-                    struct plip_local *, struct plip_local *);
-static int plip_receive_packet(struct net_device *, struct net_local *,
-                              struct plip_local *, struct plip_local *);
-static int plip_send_packet(struct net_device *, struct net_local *,
-                           struct plip_local *, struct plip_local *);
-static int plip_connection_close(struct net_device *, struct net_local *,
-                                struct plip_local *, struct plip_local *);
-static int plip_error(struct net_device *, struct net_local *,
-                     struct plip_local *, struct plip_local *);
-static int plip_bh_timeout_error(struct net_device *dev, struct net_local *nl,
-                                struct plip_local *snd,
-                                struct plip_local *rcv,
-                                int error);
-
-#define OK        0
-#define TIMEOUT   1
-#define ERROR     2
-#define HS_TIMEOUT     3
-
-typedef int (*plip_func)(struct net_device *dev, struct net_local *nl,
-                        struct plip_local *snd, struct plip_local *rcv);
-
-static const plip_func connection_state_table[] =
-{
-       plip_none,
-       plip_receive_packet,
-       plip_send_packet,
-       plip_connection_close,
-       plip_error
-};
-
-/* Bottom half handler of PLIP. */
-static void
-plip_bh(struct work_struct *work)
-{
-       struct net_local *nl = container_of(work, struct net_local, immediate);
-       struct plip_local *snd = &nl->snd_data;
-       struct plip_local *rcv = &nl->rcv_data;
-       plip_func f;
-       int r;
-
-       nl->is_deferred = 0;
-       f = connection_state_table[nl->connection];
-       if ((r = (*f)(nl->dev, nl, snd, rcv)) != OK &&
-           (r = plip_bh_timeout_error(nl->dev, nl, snd, rcv, r)) != OK) {
-               nl->is_deferred = 1;
-               schedule_delayed_work(&nl->deferred, 1);
-       }
-}
-
-static void
-plip_timer_bh(struct work_struct *work)
-{
-       struct net_local *nl =
-               container_of(work, struct net_local, timer.work);
-
-       if (!(atomic_read (&nl->kill_timer))) {
-               plip_interrupt (nl->dev);
-
-               schedule_delayed_work(&nl->timer, 1);
-       }
-       else {
-               complete(&nl->killed_timer_cmp);
-       }
-}
-
-static int
-plip_bh_timeout_error(struct net_device *dev, struct net_local *nl,
-                     struct plip_local *snd, struct plip_local *rcv,
-                     int error)
-{
-       unsigned char c0;
-       /*
-        * This is tricky. If we got here from the beginning of send (either
-        * with ERROR or HS_TIMEOUT) we have IRQ enabled. Otherwise it's
-        * already disabled. With the old variant of {enable,disable}_irq()
-        * extra disable_irq() was a no-op. Now it became mortal - it's
-        * unbalanced and thus we'll never re-enable IRQ (until rmmod plip,
-        * that is). So we have to treat HS_TIMEOUT and ERROR from send
-        * in a special way.
-        */
-
-       spin_lock_irq(&nl->lock);
-       if (nl->connection == PLIP_CN_SEND) {
-
-               if (error != ERROR) { /* Timeout */
-                       nl->timeout_count++;
-                       if ((error == HS_TIMEOUT && nl->timeout_count <= 10) ||
-                           nl->timeout_count <= 3) {
-                               spin_unlock_irq(&nl->lock);
-                               /* Try again later */
-                               return TIMEOUT;
-                       }
-                       c0 = read_status(dev);
-                       printk(KERN_WARNING "%s: transmit timeout(%d,%02x)\n",
-                              dev->name, snd->state, c0);
-               } else
-                       error = HS_TIMEOUT;
-               dev->stats.tx_errors++;
-               dev->stats.tx_aborted_errors++;
-       } else if (nl->connection == PLIP_CN_RECEIVE) {
-               if (rcv->state == PLIP_PK_TRIGGER) {
-                       /* Transmission was interrupted. */
-                       spin_unlock_irq(&nl->lock);
-                       return OK;
-               }
-               if (error != ERROR) { /* Timeout */
-                       if (++nl->timeout_count <= 3) {
-                               spin_unlock_irq(&nl->lock);
-                               /* Try again later */
-                               return TIMEOUT;
-                       }
-                       c0 = read_status(dev);
-                       printk(KERN_WARNING "%s: receive timeout(%d,%02x)\n",
-                              dev->name, rcv->state, c0);
-               }
-               dev->stats.rx_dropped++;
-       }
-       rcv->state = PLIP_PK_DONE;
-       if (rcv->skb) {
-               kfree_skb(rcv->skb);
-               rcv->skb = NULL;
-       }
-       snd->state = PLIP_PK_DONE;
-       if (snd->skb) {
-               dev_kfree_skb(snd->skb);
-               snd->skb = NULL;
-       }
-       spin_unlock_irq(&nl->lock);
-       if (error == HS_TIMEOUT) {
-               DISABLE(dev->irq);
-               synchronize_irq(dev->irq);
-       }
-       disable_parport_interrupts (dev);
-       netif_stop_queue (dev);
-       nl->connection = PLIP_CN_ERROR;
-       write_data (dev, 0x00);
-
-       return TIMEOUT;
-}
-
-static int
-plip_none(struct net_device *dev, struct net_local *nl,
-         struct plip_local *snd, struct plip_local *rcv)
-{
-       return OK;
-}
-
-/* PLIP_RECEIVE --- receive a byte(two nibbles)
-   Returns OK on success, TIMEOUT on timeout */
-static inline int
-plip_receive(unsigned short nibble_timeout, struct net_device *dev,
-            enum plip_nibble_state *ns_p, unsigned char *data_p)
-{
-       unsigned char c0, c1;
-       unsigned int cx;
-
-       switch (*ns_p) {
-       case PLIP_NB_BEGIN:
-               cx = nibble_timeout;
-               while (1) {
-                       c0 = read_status(dev);
-                       udelay(PLIP_DELAY_UNIT);
-                       if ((c0 & 0x80) == 0) {
-                               c1 = read_status(dev);
-                               if (c0 == c1)
-                                       break;
-                       }
-                       if (--cx == 0)
-                               return TIMEOUT;
-               }
-               *data_p = (c0 >> 3) & 0x0f;
-               write_data (dev, 0x10); /* send ACK */
-               *ns_p = PLIP_NB_1;
-
-       case PLIP_NB_1:
-               cx = nibble_timeout;
-               while (1) {
-                       c0 = read_status(dev);
-                       udelay(PLIP_DELAY_UNIT);
-                       if (c0 & 0x80) {
-                               c1 = read_status(dev);
-                               if (c0 == c1)
-                                       break;
-                       }
-                       if (--cx == 0)
-                               return TIMEOUT;
-               }
-               *data_p |= (c0 << 1) & 0xf0;
-               write_data (dev, 0x00); /* send ACK */
-               *ns_p = PLIP_NB_BEGIN;
-       case PLIP_NB_2:
-               break;
-       }
-       return OK;
-}
-
-/*
- *     Determine the packet's protocol ID. The rule here is that we
- *     assume 802.3 if the type field is short enough to be a length.
- *     This is normal practice and works for any 'now in use' protocol.
- *
- *     PLIP is ethernet ish but the daddr might not be valid if unicast.
- *     PLIP fortunately has no bus architecture (its Point-to-point).
- *
- *     We can't fix the daddr thing as that quirk (more bug) is embedded
- *     in far too many old systems not all even running Linux.
- */
-
-static __be16 plip_type_trans(struct sk_buff *skb, struct net_device *dev)
-{
-       struct ethhdr *eth;
-       unsigned char *rawp;
-
-       skb_reset_mac_header(skb);
-       skb_pull(skb,dev->hard_header_len);
-       eth = eth_hdr(skb);
-
-       if(*eth->h_dest&1)
-       {
-               if(memcmp(eth->h_dest,dev->broadcast, ETH_ALEN)==0)
-                       skb->pkt_type=PACKET_BROADCAST;
-               else
-                       skb->pkt_type=PACKET_MULTICAST;
-       }
-
-       /*
-        *      This ALLMULTI check should be redundant by 1.4
-        *      so don't forget to remove it.
-        */
-
-       if (ntohs(eth->h_proto) >= 1536)
-               return eth->h_proto;
-
-       rawp = skb->data;
-
-       /*
-        *      This is a magic hack to spot IPX packets. Older Novell breaks
-        *      the protocol design and runs IPX over 802.3 without an 802.2 LLC
-        *      layer. We look for FFFF which isn't a used 802.2 SSAP/DSAP. This
-        *      won't work for fault tolerant netware but does for the rest.
-        */
-       if (*(unsigned short *)rawp == 0xFFFF)
-               return htons(ETH_P_802_3);
-
-       /*
-        *      Real 802.2 LLC
-        */
-       return htons(ETH_P_802_2);
-}
-
-/* PLIP_RECEIVE_PACKET --- receive a packet */
-static int
-plip_receive_packet(struct net_device *dev, struct net_local *nl,
-                   struct plip_local *snd, struct plip_local *rcv)
-{
-       unsigned short nibble_timeout = nl->nibble;
-       unsigned char *lbuf;
-
-       switch (rcv->state) {
-       case PLIP_PK_TRIGGER:
-               DISABLE(dev->irq);
-               /* Don't need to synchronize irq, as we can safely ignore it */
-               disable_parport_interrupts (dev);
-               write_data (dev, 0x01); /* send ACK */
-               if (net_debug > 2)
-                       printk(KERN_DEBUG "%s: receive start\n", dev->name);
-               rcv->state = PLIP_PK_LENGTH_LSB;
-               rcv->nibble = PLIP_NB_BEGIN;
-
-       case PLIP_PK_LENGTH_LSB:
-               if (snd->state != PLIP_PK_DONE) {
-                       if (plip_receive(nl->trigger, dev,
-                                        &rcv->nibble, &rcv->length.b.lsb)) {
-                               /* collision, here dev->tbusy == 1 */
-                               rcv->state = PLIP_PK_DONE;
-                               nl->is_deferred = 1;
-                               nl->connection = PLIP_CN_SEND;
-                               schedule_delayed_work(&nl->deferred, 1);
-                               enable_parport_interrupts (dev);
-                               ENABLE(dev->irq);
-                               return OK;
-                       }
-               } else {
-                       if (plip_receive(nibble_timeout, dev,
-                                        &rcv->nibble, &rcv->length.b.lsb))
-                               return TIMEOUT;
-               }
-               rcv->state = PLIP_PK_LENGTH_MSB;
-
-       case PLIP_PK_LENGTH_MSB:
-               if (plip_receive(nibble_timeout, dev,
-                                &rcv->nibble, &rcv->length.b.msb))
-                       return TIMEOUT;
-               if (rcv->length.h > dev->mtu + dev->hard_header_len ||
-                   rcv->length.h < 8) {
-                       printk(KERN_WARNING "%s: bogus packet size %d.\n", dev->name, rcv->length.h);
-                       return ERROR;
-               }
-               /* Malloc up new buffer. */
-               rcv->skb = dev_alloc_skb(rcv->length.h + 2);
-               if (rcv->skb == NULL) {
-                       printk(KERN_ERR "%s: Memory squeeze.\n", dev->name);
-                       return ERROR;
-               }
-               skb_reserve(rcv->skb, 2);       /* Align IP on 16 byte boundaries */
-               skb_put(rcv->skb,rcv->length.h);
-               rcv->skb->dev = dev;
-               rcv->state = PLIP_PK_DATA;
-               rcv->byte = 0;
-               rcv->checksum = 0;
-
-       case PLIP_PK_DATA:
-               lbuf = rcv->skb->data;
-               do {
-                       if (plip_receive(nibble_timeout, dev,
-                                        &rcv->nibble, &lbuf[rcv->byte]))
-                               return TIMEOUT;
-               } while (++rcv->byte < rcv->length.h);
-               do {
-                       rcv->checksum += lbuf[--rcv->byte];
-               } while (rcv->byte);
-               rcv->state = PLIP_PK_CHECKSUM;
-
-       case PLIP_PK_CHECKSUM:
-               if (plip_receive(nibble_timeout, dev,
-                                &rcv->nibble, &rcv->data))
-                       return TIMEOUT;
-               if (rcv->data != rcv->checksum) {
-                       dev->stats.rx_crc_errors++;
-                       if (net_debug)
-                               printk(KERN_DEBUG "%s: checksum error\n", dev->name);
-                       return ERROR;
-               }
-               rcv->state = PLIP_PK_DONE;
-
-       case PLIP_PK_DONE:
-               /* Inform the upper layer for the arrival of a packet. */
-               rcv->skb->protocol=plip_type_trans(rcv->skb, dev);
-               netif_rx_ni(rcv->skb);
-               dev->stats.rx_bytes += rcv->length.h;
-               dev->stats.rx_packets++;
-               rcv->skb = NULL;
-               if (net_debug > 2)
-                       printk(KERN_DEBUG "%s: receive end\n", dev->name);
-
-               /* Close the connection. */
-               write_data (dev, 0x00);
-               spin_lock_irq(&nl->lock);
-               if (snd->state != PLIP_PK_DONE) {
-                       nl->connection = PLIP_CN_SEND;
-                       spin_unlock_irq(&nl->lock);
-                       schedule_work(&nl->immediate);
-                       enable_parport_interrupts (dev);
-                       ENABLE(dev->irq);
-                       return OK;
-               } else {
-                       nl->connection = PLIP_CN_NONE;
-                       spin_unlock_irq(&nl->lock);
-                       enable_parport_interrupts (dev);
-                       ENABLE(dev->irq);
-                       return OK;
-               }
-       }
-       return OK;
-}
-
-/* PLIP_SEND --- send a byte (two nibbles)
-   Returns OK on success, TIMEOUT when timeout    */
-static inline int
-plip_send(unsigned short nibble_timeout, struct net_device *dev,
-         enum plip_nibble_state *ns_p, unsigned char data)
-{
-       unsigned char c0;
-       unsigned int cx;
-
-       switch (*ns_p) {
-       case PLIP_NB_BEGIN:
-               write_data (dev, data & 0x0f);
-               *ns_p = PLIP_NB_1;
-
-       case PLIP_NB_1:
-               write_data (dev, 0x10 | (data & 0x0f));
-               cx = nibble_timeout;
-               while (1) {
-                       c0 = read_status(dev);
-                       if ((c0 & 0x80) == 0)
-                               break;
-                       if (--cx == 0)
-                               return TIMEOUT;
-                       udelay(PLIP_DELAY_UNIT);
-               }
-               write_data (dev, 0x10 | (data >> 4));
-               *ns_p = PLIP_NB_2;
-
-       case PLIP_NB_2:
-               write_data (dev, (data >> 4));
-               cx = nibble_timeout;
-               while (1) {
-                       c0 = read_status(dev);
-                       if (c0 & 0x80)
-                               break;
-                       if (--cx == 0)
-                               return TIMEOUT;
-                       udelay(PLIP_DELAY_UNIT);
-               }
-               *ns_p = PLIP_NB_BEGIN;
-               return OK;
-       }
-       return OK;
-}
-
-/* PLIP_SEND_PACKET --- send a packet */
-static int
-plip_send_packet(struct net_device *dev, struct net_local *nl,
-                struct plip_local *snd, struct plip_local *rcv)
-{
-       unsigned short nibble_timeout = nl->nibble;
-       unsigned char *lbuf;
-       unsigned char c0;
-       unsigned int cx;
-
-       if (snd->skb == NULL || (lbuf = snd->skb->data) == NULL) {
-               printk(KERN_DEBUG "%s: send skb lost\n", dev->name);
-               snd->state = PLIP_PK_DONE;
-               snd->skb = NULL;
-               return ERROR;
-       }
-
-       switch (snd->state) {
-       case PLIP_PK_TRIGGER:
-               if ((read_status(dev) & 0xf8) != 0x80)
-                       return HS_TIMEOUT;
-
-               /* Trigger remote rx interrupt. */
-               write_data (dev, 0x08);
-               cx = nl->trigger;
-               while (1) {
-                       udelay(PLIP_DELAY_UNIT);
-                       spin_lock_irq(&nl->lock);
-                       if (nl->connection == PLIP_CN_RECEIVE) {
-                               spin_unlock_irq(&nl->lock);
-                               /* Interrupted. */
-                               dev->stats.collisions++;
-                               return OK;
-                       }
-                       c0 = read_status(dev);
-                       if (c0 & 0x08) {
-                               spin_unlock_irq(&nl->lock);
-                               DISABLE(dev->irq);
-                               synchronize_irq(dev->irq);
-                               if (nl->connection == PLIP_CN_RECEIVE) {
-                                       /* Interrupted.
-                                          We don't need to enable irq,
-                                          as it is soon disabled.    */
-                                       /* Yes, we do. New variant of
-                                          {enable,disable}_irq *counts*
-                                          them.  -- AV  */
-                                       ENABLE(dev->irq);
-                                       dev->stats.collisions++;
-                                       return OK;
-                               }
-                               disable_parport_interrupts (dev);
-                               if (net_debug > 2)
-                                       printk(KERN_DEBUG "%s: send start\n", dev->name);
-                               snd->state = PLIP_PK_LENGTH_LSB;
-                               snd->nibble = PLIP_NB_BEGIN;
-                               nl->timeout_count = 0;
-                               break;
-                       }
-                       spin_unlock_irq(&nl->lock);
-                       if (--cx == 0) {
-                               write_data (dev, 0x00);
-                               return HS_TIMEOUT;
-                       }
-               }
-
-       case PLIP_PK_LENGTH_LSB:
-               if (plip_send(nibble_timeout, dev,
-                             &snd->nibble, snd->length.b.lsb))
-                       return TIMEOUT;
-               snd->state = PLIP_PK_LENGTH_MSB;
-
-       case PLIP_PK_LENGTH_MSB:
-               if (plip_send(nibble_timeout, dev,
-                             &snd->nibble, snd->length.b.msb))
-                       return TIMEOUT;
-               snd->state = PLIP_PK_DATA;
-               snd->byte = 0;
-               snd->checksum = 0;
-
-       case PLIP_PK_DATA:
-               do {
-                       if (plip_send(nibble_timeout, dev,
-                                     &snd->nibble, lbuf[snd->byte]))
-                               return TIMEOUT;
-               } while (++snd->byte < snd->length.h);
-               do {
-                       snd->checksum += lbuf[--snd->byte];
-               } while (snd->byte);
-               snd->state = PLIP_PK_CHECKSUM;
-
-       case PLIP_PK_CHECKSUM:
-               if (plip_send(nibble_timeout, dev,
-                             &snd->nibble, snd->checksum))
-                       return TIMEOUT;
-
-               dev->stats.tx_bytes += snd->skb->len;
-               dev_kfree_skb(snd->skb);
-               dev->stats.tx_packets++;
-               snd->state = PLIP_PK_DONE;
-
-       case PLIP_PK_DONE:
-               /* Close the connection */
-               write_data (dev, 0x00);
-               snd->skb = NULL;
-               if (net_debug > 2)
-                       printk(KERN_DEBUG "%s: send end\n", dev->name);
-               nl->connection = PLIP_CN_CLOSING;
-               nl->is_deferred = 1;
-               schedule_delayed_work(&nl->deferred, 1);
-               enable_parport_interrupts (dev);
-               ENABLE(dev->irq);
-               return OK;
-       }
-       return OK;
-}
-
-static int
-plip_connection_close(struct net_device *dev, struct net_local *nl,
-                     struct plip_local *snd, struct plip_local *rcv)
-{
-       spin_lock_irq(&nl->lock);
-       if (nl->connection == PLIP_CN_CLOSING) {
-               nl->connection = PLIP_CN_NONE;
-               netif_wake_queue (dev);
-       }
-       spin_unlock_irq(&nl->lock);
-       if (nl->should_relinquish) {
-               nl->should_relinquish = nl->port_owner = 0;
-               parport_release(nl->pardev);
-       }
-       return OK;
-}
-
-/* PLIP_ERROR --- wait till other end settled */
-static int
-plip_error(struct net_device *dev, struct net_local *nl,
-          struct plip_local *snd, struct plip_local *rcv)
-{
-       unsigned char status;
-
-       status = read_status(dev);
-       if ((status & 0xf8) == 0x80) {
-               if (net_debug > 2)
-                       printk(KERN_DEBUG "%s: reset interface.\n", dev->name);
-               nl->connection = PLIP_CN_NONE;
-               nl->should_relinquish = 0;
-               netif_start_queue (dev);
-               enable_parport_interrupts (dev);
-               ENABLE(dev->irq);
-               netif_wake_queue (dev);
-       } else {
-               nl->is_deferred = 1;
-               schedule_delayed_work(&nl->deferred, 1);
-       }
-
-       return OK;
-}
-
-/* Handle the parallel port interrupts. */
-static void
-plip_interrupt(void *dev_id)
-{
-       struct net_device *dev = dev_id;
-       struct net_local *nl;
-       struct plip_local *rcv;
-       unsigned char c0;
-       unsigned long flags;
-
-       nl = netdev_priv(dev);
-       rcv = &nl->rcv_data;
-
-       spin_lock_irqsave (&nl->lock, flags);
-
-       c0 = read_status(dev);
-       if ((c0 & 0xf8) != 0xc0) {
-               if ((dev->irq != -1) && (net_debug > 1))
-                       printk(KERN_DEBUG "%s: spurious interrupt\n", dev->name);
-               spin_unlock_irqrestore (&nl->lock, flags);
-               return;
-       }
-
-       if (net_debug > 3)
-               printk(KERN_DEBUG "%s: interrupt.\n", dev->name);
-
-       switch (nl->connection) {
-       case PLIP_CN_CLOSING:
-               netif_wake_queue (dev);
-       case PLIP_CN_NONE:
-       case PLIP_CN_SEND:
-               rcv->state = PLIP_PK_TRIGGER;
-               nl->connection = PLIP_CN_RECEIVE;
-               nl->timeout_count = 0;
-               schedule_work(&nl->immediate);
-               break;
-
-       case PLIP_CN_RECEIVE:
-               /* May occur because there is race condition
-                  around test and set of dev->interrupt.
-                  Ignore this interrupt. */
-               break;
-
-       case PLIP_CN_ERROR:
-               printk(KERN_ERR "%s: receive interrupt in error state\n", dev->name);
-               break;
-       }
-
-       spin_unlock_irqrestore(&nl->lock, flags);
-}
-
-static int
-plip_tx_packet(struct sk_buff *skb, struct net_device *dev)
-{
-       struct net_local *nl = netdev_priv(dev);
-       struct plip_local *snd = &nl->snd_data;
-
-       if (netif_queue_stopped(dev))
-               return NETDEV_TX_BUSY;
-
-       /* We may need to grab the bus */
-       if (!nl->port_owner) {
-               if (parport_claim(nl->pardev))
-                       return NETDEV_TX_BUSY;
-               nl->port_owner = 1;
-       }
-
-       netif_stop_queue (dev);
-
-       if (skb->len > dev->mtu + dev->hard_header_len) {
-               printk(KERN_WARNING "%s: packet too big, %d.\n", dev->name, (int)skb->len);
-               netif_start_queue (dev);
-               return NETDEV_TX_BUSY;
-       }
-
-       if (net_debug > 2)
-               printk(KERN_DEBUG "%s: send request\n", dev->name);
-
-       spin_lock_irq(&nl->lock);
-       snd->skb = skb;
-       snd->length.h = skb->len;
-       snd->state = PLIP_PK_TRIGGER;
-       if (nl->connection == PLIP_CN_NONE) {
-               nl->connection = PLIP_CN_SEND;
-               nl->timeout_count = 0;
-       }
-       schedule_work(&nl->immediate);
-       spin_unlock_irq(&nl->lock);
-
-       return NETDEV_TX_OK;
-}
-
-static void
-plip_rewrite_address(const struct net_device *dev, struct ethhdr *eth)
-{
-       const struct in_device *in_dev;
-
-       rcu_read_lock();
-       in_dev = __in_dev_get_rcu(dev);
-       if (in_dev) {
-               /* Any address will do - we take the first */
-               const struct in_ifaddr *ifa = in_dev->ifa_list;
-               if (ifa) {
-                       memcpy(eth->h_source, dev->dev_addr, 6);
-                       memset(eth->h_dest, 0xfc, 2);
-                       memcpy(eth->h_dest+2, &ifa->ifa_address, 4);
-               }
-       }
-       rcu_read_unlock();
-}
-
-static int
-plip_hard_header(struct sk_buff *skb, struct net_device *dev,
-                unsigned short type, const void *daddr,
-                const void *saddr, unsigned len)
-{
-       int ret;
-
-       ret = eth_header(skb, dev, type, daddr, saddr, len);
-       if (ret >= 0)
-               plip_rewrite_address (dev, (struct ethhdr *)skb->data);
-
-       return ret;
-}
-
-static int plip_hard_header_cache(const struct neighbour *neigh,
-                                 struct hh_cache *hh, __be16 type)
-{
-       int ret;
-
-       ret = eth_header_cache(neigh, hh, type);
-       if (ret == 0) {
-               struct ethhdr *eth;
-
-               eth = (struct ethhdr*)(((u8*)hh->hh_data) +
-                                      HH_DATA_OFF(sizeof(*eth)));
-               plip_rewrite_address (neigh->dev, eth);
-       }
-
-       return ret;
-}
-
-/* Open/initialize the board.  This is called (in the current kernel)
-   sometime after booting when the 'ifconfig' program is run.
-
-   This routine gets exclusive access to the parallel port by allocating
-   its IRQ line.
- */
-static int
-plip_open(struct net_device *dev)
-{
-       struct net_local *nl = netdev_priv(dev);
-       struct in_device *in_dev;
-
-       /* Grab the port */
-       if (!nl->port_owner) {
-               if (parport_claim(nl->pardev)) return -EAGAIN;
-               nl->port_owner = 1;
-       }
-
-       nl->should_relinquish = 0;
-
-       /* Clear the data port. */
-       write_data (dev, 0x00);
-
-       /* Enable rx interrupt. */
-       enable_parport_interrupts (dev);
-       if (dev->irq == -1)
-       {
-               atomic_set (&nl->kill_timer, 0);
-               schedule_delayed_work(&nl->timer, 1);
-       }
-
-       /* Initialize the state machine. */
-       nl->rcv_data.state = nl->snd_data.state = PLIP_PK_DONE;
-       nl->rcv_data.skb = nl->snd_data.skb = NULL;
-       nl->connection = PLIP_CN_NONE;
-       nl->is_deferred = 0;
-
-       /* Fill in the MAC-level header.
-          We used to abuse dev->broadcast to store the point-to-point
-          MAC address, but we no longer do it. Instead, we fetch the
-          interface address whenever it is needed, which is cheap enough
-          because we use the hh_cache. Actually, abusing dev->broadcast
-          didn't work, because when using plip_open the point-to-point
-          address isn't yet known.
-          PLIP doesn't have a real MAC address, but we need it to be
-          DOS compatible, and to properly support taps (otherwise,
-          when the device address isn't identical to the address of a
-          received frame, the kernel incorrectly drops it).             */
-
-       in_dev=__in_dev_get_rtnl(dev);
-       if (in_dev) {
-               /* Any address will do - we take the first. We already
-                  have the first two bytes filled with 0xfc, from
-                  plip_init_dev(). */
-               struct in_ifaddr *ifa=in_dev->ifa_list;
-               if (ifa != NULL) {
-                       memcpy(dev->dev_addr+2, &ifa->ifa_local, 4);
-               }
-       }
-
-       netif_start_queue (dev);
-
-       return 0;
-}
-
-/* The inverse routine to plip_open (). */
-static int
-plip_close(struct net_device *dev)
-{
-       struct net_local *nl = netdev_priv(dev);
-       struct plip_local *snd = &nl->snd_data;
-       struct plip_local *rcv = &nl->rcv_data;
-
-       netif_stop_queue (dev);
-       DISABLE(dev->irq);
-       synchronize_irq(dev->irq);
-
-       if (dev->irq == -1)
-       {
-               init_completion(&nl->killed_timer_cmp);
-               atomic_set (&nl->kill_timer, 1);
-               wait_for_completion(&nl->killed_timer_cmp);
-       }
-
-#ifdef NOTDEF
-       outb(0x00, PAR_DATA(dev));
-#endif
-       nl->is_deferred = 0;
-       nl->connection = PLIP_CN_NONE;
-       if (nl->port_owner) {
-               parport_release(nl->pardev);
-               nl->port_owner = 0;
-       }
-
-       snd->state = PLIP_PK_DONE;
-       if (snd->skb) {
-               dev_kfree_skb(snd->skb);
-               snd->skb = NULL;
-       }
-       rcv->state = PLIP_PK_DONE;
-       if (rcv->skb) {
-               kfree_skb(rcv->skb);
-               rcv->skb = NULL;
-       }
-
-#ifdef NOTDEF
-       /* Reset. */
-       outb(0x00, PAR_CONTROL(dev));
-#endif
-       return 0;
-}
-
-static int
-plip_preempt(void *handle)
-{
-       struct net_device *dev = (struct net_device *)handle;
-       struct net_local *nl = netdev_priv(dev);
-
-       /* Stand our ground if a datagram is on the wire */
-       if (nl->connection != PLIP_CN_NONE) {
-               nl->should_relinquish = 1;
-               return 1;
-       }
-
-       nl->port_owner = 0;     /* Remember that we released the bus */
-       return 0;
-}
-
-static void
-plip_wakeup(void *handle)
-{
-       struct net_device *dev = (struct net_device *)handle;
-       struct net_local *nl = netdev_priv(dev);
-
-       if (nl->port_owner) {
-               /* Why are we being woken up? */
-               printk(KERN_DEBUG "%s: why am I being woken up?\n", dev->name);
-               if (!parport_claim(nl->pardev))
-                       /* bus_owner is already set (but why?) */
-                       printk(KERN_DEBUG "%s: I'm broken.\n", dev->name);
-               else
-                       return;
-       }
-
-       if (!(dev->flags & IFF_UP))
-               /* Don't need the port when the interface is down */
-               return;
-
-       if (!parport_claim(nl->pardev)) {
-               nl->port_owner = 1;
-               /* Clear the data port. */
-               write_data (dev, 0x00);
-       }
-}
-
-static int
-plip_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
-{
-       struct net_local *nl = netdev_priv(dev);
-       struct plipconf *pc = (struct plipconf *) &rq->ifr_ifru;
-
-       if (cmd != SIOCDEVPLIP)
-               return -EOPNOTSUPP;
-
-       switch(pc->pcmd) {
-       case PLIP_GET_TIMEOUT:
-               pc->trigger = nl->trigger;
-               pc->nibble  = nl->nibble;
-               break;
-       case PLIP_SET_TIMEOUT:
-               if(!capable(CAP_NET_ADMIN))
-                       return -EPERM;
-               nl->trigger = pc->trigger;
-               nl->nibble  = pc->nibble;
-               break;
-       default:
-               return -EOPNOTSUPP;
-       }
-       return 0;
-}
-
-static int parport[PLIP_MAX] = { [0 ... PLIP_MAX-1] = -1 };
-static int timid;
-
-module_param_array(parport, int, NULL, 0);
-module_param(timid, int, 0);
-MODULE_PARM_DESC(parport, "List of parport device numbers to use by plip");
-
-static struct net_device *dev_plip[PLIP_MAX] = { NULL, };
-
-static inline int
-plip_searchfor(int list[], int a)
-{
-       int i;
-       for (i = 0; i < PLIP_MAX && list[i] != -1; i++) {
-               if (list[i] == a) return 1;
-       }
-       return 0;
-}
-
-/* plip_attach() is called (by the parport code) when a port is
- * available to use. */
-static void plip_attach (struct parport *port)
-{
-       static int unit;
-       struct net_device *dev;
-       struct net_local *nl;
-       char name[IFNAMSIZ];
-
-       if ((parport[0] == -1 && (!timid || !port->devices)) ||
-           plip_searchfor(parport, port->number)) {
-               if (unit == PLIP_MAX) {
-                       printk(KERN_ERR "plip: too many devices\n");
-                       return;
-               }
-
-               sprintf(name, "plip%d", unit);
-               dev = alloc_etherdev(sizeof(struct net_local));
-               if (!dev) {
-                       printk(KERN_ERR "plip: memory squeeze\n");
-                       return;
-               }
-
-               strcpy(dev->name, name);
-
-               dev->irq = port->irq;
-               dev->base_addr = port->base;
-               if (port->irq == -1) {
-                       printk(KERN_INFO "plip: %s has no IRQ. Using IRQ-less mode,"
-                                "which is fairly inefficient!\n", port->name);
-               }
-
-               nl = netdev_priv(dev);
-               nl->dev = dev;
-               nl->pardev = parport_register_device(port, dev->name, plip_preempt,
-                                                plip_wakeup, plip_interrupt,
-                                                0, dev);
-
-               if (!nl->pardev) {
-                       printk(KERN_ERR "%s: parport_register failed\n", name);
-                       goto err_free_dev;
-               }
-
-               plip_init_netdev(dev);
-
-               if (register_netdev(dev)) {
-                       printk(KERN_ERR "%s: network register failed\n", name);
-                       goto err_parport_unregister;
-               }
-
-               printk(KERN_INFO "%s", version);
-               if (dev->irq != -1)
-                       printk(KERN_INFO "%s: Parallel port at %#3lx, "
-                                        "using IRQ %d.\n",
-                                        dev->name, dev->base_addr, dev->irq);
-               else
-                       printk(KERN_INFO "%s: Parallel port at %#3lx, "
-                                        "not using IRQ.\n",
-                                        dev->name, dev->base_addr);
-               dev_plip[unit++] = dev;
-       }
-       return;
-
-err_parport_unregister:
-       parport_unregister_device(nl->pardev);
-err_free_dev:
-       free_netdev(dev);
-}
-
-/* plip_detach() is called (by the parport code) when a port is
- * no longer available to use. */
-static void plip_detach (struct parport *port)
-{
-       /* Nothing to do */
-}
-
-static struct parport_driver plip_driver = {
-       .name   = "plip",
-       .attach = plip_attach,
-       .detach = plip_detach
-};
-
-static void __exit plip_cleanup_module (void)
-{
-       struct net_device *dev;
-       int i;
-
-       parport_unregister_driver (&plip_driver);
-
-       for (i=0; i < PLIP_MAX; i++) {
-               if ((dev = dev_plip[i])) {
-                       struct net_local *nl = netdev_priv(dev);
-                       unregister_netdev(dev);
-                       if (nl->port_owner)
-                               parport_release(nl->pardev);
-                       parport_unregister_device(nl->pardev);
-                       free_netdev(dev);
-                       dev_plip[i] = NULL;
-               }
-       }
-}
-
-#ifndef MODULE
-
-static int parport_ptr;
-
-static int __init plip_setup(char *str)
-{
-       int ints[4];
-
-       str = get_options(str, ARRAY_SIZE(ints), ints);
-
-       /* Ugh. */
-       if (!strncmp(str, "parport", 7)) {
-               int n = simple_strtoul(str+7, NULL, 10);
-               if (parport_ptr < PLIP_MAX)
-                       parport[parport_ptr++] = n;
-               else
-                       printk(KERN_INFO "plip: too many ports, %s ignored.\n",
-                              str);
-       } else if (!strcmp(str, "timid")) {
-               timid = 1;
-       } else {
-               if (ints[0] == 0 || ints[1] == 0) {
-                       /* disable driver on "plip=" or "plip=0" */
-                       parport[0] = -2;
-               } else {
-                       printk(KERN_WARNING "warning: 'plip=0x%x' ignored\n",
-                              ints[1]);
-               }
-       }
-       return 1;
-}
-
-__setup("plip=", plip_setup);
-
-#endif /* !MODULE */
-
-static int __init plip_init (void)
-{
-       if (parport[0] == -2)
-               return 0;
-
-       if (parport[0] != -1 && timid) {
-               printk(KERN_WARNING "plip: warning, ignoring `timid' since specific ports given.\n");
-               timid = 0;
-       }
-
-       if (parport_register_driver (&plip_driver)) {
-               printk (KERN_WARNING "plip: couldn't register driver\n");
-               return 1;
-       }
-
-       return 0;
-}
-
-module_init(plip_init);
-module_exit(plip_cleanup_module);
-MODULE_LICENSE("GPL");
diff --git a/drivers/net/plip/Kconfig b/drivers/net/plip/Kconfig
new file mode 100644 (file)
index 0000000..80c4a33
--- /dev/null
@@ -0,0 +1,38 @@
+#
+# Parallel Line Internet Protocol (PLIP) network device configuration
+#
+
+config PLIP
+       tristate "PLIP (parallel port) support"
+       depends on PARPORT
+       ---help---
+         PLIP (Parallel Line Internet Protocol) is used to create a
+         reasonably fast mini network consisting of two (or, rarely, more)
+         local machines.  A PLIP link from a Linux box is a popular means to
+         install a Linux distribution on a machine which doesn't have a
+         CD-ROM drive (a minimal system has to be transferred with floppies
+         first). The kernels on both machines need to have this PLIP option
+         enabled for this to work.
+
+         The PLIP driver has two modes, mode 0 and mode 1.  The parallel
+         ports (the connectors at the computers with 25 holes) are connected
+         with "null printer" or "Turbo Laplink" cables which can transmit 4
+         bits at a time (mode 0) or with special PLIP cables, to be used on
+         bidirectional parallel ports only, which can transmit 8 bits at a
+         time (mode 1); you can find the wiring of these cables in
+         <file:Documentation/networking/PLIP.txt>.  The cables can be up to
+         15m long.  Mode 0 works also if one of the machines runs DOS/Windows
+         and has some PLIP software installed, e.g. the Crynwr PLIP packet
+         driver (<http://oak.oakland.edu/simtel.net/msdos/pktdrvr-pre.html>)
+         and winsock or NCSA's telnet.
+
+         If you want to use PLIP, say Y and read the PLIP mini-HOWTO as well
+         as the NET-3-HOWTO, both available from
+         <http://www.tldp.org/docs.html#howto>.  Note that the PLIP
+         protocol has been changed and this PLIP driver won't work together
+         with the PLIP support in Linux versions 1.0.x.  This option enlarges
+         your kernel by about 8 KB.
+
+         To compile this driver as a module, choose M here. The module
+         will be called plip. If unsure, say Y or M, in case you buy
+         a laptop later.
diff --git a/drivers/net/plip/Makefile b/drivers/net/plip/Makefile
new file mode 100644 (file)
index 0000000..ed95879
--- /dev/null
@@ -0,0 +1,5 @@
+#
+# Makefile for the PLIP network device drivers.
+#
+
+obj-$(CONFIG_PLIP) += plip.o
diff --git a/drivers/net/plip/plip.c b/drivers/net/plip/plip.c
new file mode 100644 (file)
index 0000000..a9e9ca8
--- /dev/null
@@ -0,0 +1,1403 @@
+/* $Id: plip.c,v 1.3.6.2 1997/04/16 15:07:56 phil Exp $ */
+/* PLIP: A parallel port "network" driver for Linux. */
+/* This driver is for parallel port with 5-bit cable (LapLink (R) cable). */
+/*
+ * Authors:    Donald Becker <becker@scyld.com>
+ *             Tommy Thorn <thorn@daimi.aau.dk>
+ *             Tanabe Hiroyasu <hiro@sanpo.t.u-tokyo.ac.jp>
+ *             Alan Cox <gw4pts@gw4pts.ampr.org>
+ *             Peter Bauer <100136.3530@compuserve.com>
+ *             Niibe Yutaka <gniibe@mri.co.jp>
+ *             Nimrod Zimerman <zimerman@mailandnews.com>
+ *
+ * Enhancements:
+ *             Modularization and ifreq/ifmap support by Alan Cox.
+ *             Rewritten by Niibe Yutaka.
+ *             parport-sharing awareness code by Philip Blundell.
+ *             SMP locking by Niibe Yutaka.
+ *             Support for parallel ports with no IRQ (poll mode),
+ *             Modifications to use the parallel port API
+ *             by Nimrod Zimerman.
+ *
+ * Fixes:
+ *             Niibe Yutaka
+ *               - Module initialization.
+ *               - MTU fix.
+ *               - Make sure other end is OK, before sending a packet.
+ *               - Fix immediate timer problem.
+ *
+ *             Al Viro
+ *               - Changed {enable,disable}_irq handling to make it work
+ *                 with new ("stack") semantics.
+ *
+ *             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.
+ */
+
+/*
+ * Original version and the name 'PLIP' from Donald Becker <becker@scyld.com>
+ * inspired by Russ Nelson's parallel port packet driver.
+ *
+ * NOTE:
+ *     Tanabe Hiroyasu had changed the protocol, and it was in Linux v1.0.
+ *     Because of the necessity to communicate to DOS machines with the
+ *     Crynwr packet driver, Peter Bauer changed the protocol again
+ *     back to original protocol.
+ *
+ *     This version follows original PLIP protocol.
+ *     So, this PLIP can't communicate the PLIP of Linux v1.0.
+ */
+
+/*
+ *     To use with DOS box, please do (Turn on ARP switch):
+ *     # ifconfig plip[0-2] arp
+ */
+static const char version[] = "NET3 PLIP version 2.4-parport gniibe@mri.co.jp\n";
+
+/*
+  Sources:
+       Ideas and protocols came from Russ Nelson's <nelson@crynwr.com>
+       "parallel.asm" parallel port packet driver.
+
+  The "Crynwr" parallel port standard specifies the following protocol:
+    Trigger by sending nibble '0x8' (this causes interrupt on other end)
+    count-low octet
+    count-high octet
+    ... data octets
+    checksum octet
+  Each octet is sent as <wait for rx. '0x1?'> <send 0x10+(octet&0x0F)>
+                       <wait for rx. '0x0?'> <send 0x00+((octet>>4)&0x0F)>
+
+  The packet is encapsulated as if it were ethernet.
+
+  The cable used is a de facto standard parallel null cable -- sold as
+  a "LapLink" cable by various places.  You'll need a 12-conductor cable to
+  make one yourself.  The wiring is:
+    SLCTIN     17 - 17
+    GROUND     25 - 25
+    D0->ERROR  2 - 15          15 - 2
+    D1->SLCT   3 - 13          13 - 3
+    D2->PAPOUT 4 - 12          12 - 4
+    D3->ACK    5 - 10          10 - 5
+    D4->BUSY   6 - 11          11 - 6
+  Do not connect the other pins.  They are
+    D5,D6,D7 are 7,8,9
+    STROBE is 1, FEED is 14, INIT is 16
+    extra grounds are 18,19,20,21,22,23,24
+*/
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/fcntl.h>
+#include <linux/interrupt.h>
+#include <linux/string.h>
+#include <linux/slab.h>
+#include <linux/if_ether.h>
+#include <linux/in.h>
+#include <linux/errno.h>
+#include <linux/delay.h>
+#include <linux/init.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/inetdevice.h>
+#include <linux/skbuff.h>
+#include <linux/if_plip.h>
+#include <linux/workqueue.h>
+#include <linux/spinlock.h>
+#include <linux/completion.h>
+#include <linux/parport.h>
+#include <linux/bitops.h>
+
+#include <net/neighbour.h>
+
+#include <asm/system.h>
+#include <asm/irq.h>
+#include <asm/byteorder.h>
+
+/* Maximum number of devices to support. */
+#define PLIP_MAX  8
+
+/* Use 0 for production, 1 for verification, >2 for debug */
+#ifndef NET_DEBUG
+#define NET_DEBUG 1
+#endif
+static const unsigned int net_debug = NET_DEBUG;
+
+#define ENABLE(irq)  if (irq != -1) enable_irq(irq)
+#define DISABLE(irq) if (irq != -1) disable_irq(irq)
+
+/* In micro second */
+#define PLIP_DELAY_UNIT                   1
+
+/* Connection time out = PLIP_TRIGGER_WAIT * PLIP_DELAY_UNIT usec */
+#define PLIP_TRIGGER_WAIT       500
+
+/* Nibble time out = PLIP_NIBBLE_WAIT * PLIP_DELAY_UNIT usec */
+#define PLIP_NIBBLE_WAIT        3000
+
+/* Bottom halves */
+static void plip_kick_bh(struct work_struct *work);
+static void plip_bh(struct work_struct *work);
+static void plip_timer_bh(struct work_struct *work);
+
+/* Interrupt handler */
+static void plip_interrupt(void *dev_id);
+
+/* Functions for DEV methods */
+static int plip_tx_packet(struct sk_buff *skb, struct net_device *dev);
+static int plip_hard_header(struct sk_buff *skb, struct net_device *dev,
+                            unsigned short type, const void *daddr,
+                           const void *saddr, unsigned len);
+static int plip_hard_header_cache(const struct neighbour *neigh,
+                                  struct hh_cache *hh, __be16 type);
+static int plip_open(struct net_device *dev);
+static int plip_close(struct net_device *dev);
+static int plip_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd);
+static int plip_preempt(void *handle);
+static void plip_wakeup(void *handle);
+
+enum plip_connection_state {
+       PLIP_CN_NONE=0,
+       PLIP_CN_RECEIVE,
+       PLIP_CN_SEND,
+       PLIP_CN_CLOSING,
+       PLIP_CN_ERROR
+};
+
+enum plip_packet_state {
+       PLIP_PK_DONE=0,
+       PLIP_PK_TRIGGER,
+       PLIP_PK_LENGTH_LSB,
+       PLIP_PK_LENGTH_MSB,
+       PLIP_PK_DATA,
+       PLIP_PK_CHECKSUM
+};
+
+enum plip_nibble_state {
+       PLIP_NB_BEGIN,
+       PLIP_NB_1,
+       PLIP_NB_2,
+};
+
+struct plip_local {
+       enum plip_packet_state state;
+       enum plip_nibble_state nibble;
+       union {
+               struct {
+#if defined(__LITTLE_ENDIAN)
+                       unsigned char lsb;
+                       unsigned char msb;
+#elif defined(__BIG_ENDIAN)
+                       unsigned char msb;
+                       unsigned char lsb;
+#else
+#error "Please fix the endianness defines in <asm/byteorder.h>"
+#endif
+               } b;
+               unsigned short h;
+       } length;
+       unsigned short byte;
+       unsigned char  checksum;
+       unsigned char  data;
+       struct sk_buff *skb;
+};
+
+struct net_local {
+       struct net_device *dev;
+       struct work_struct immediate;
+       struct delayed_work deferred;
+       struct delayed_work timer;
+       struct plip_local snd_data;
+       struct plip_local rcv_data;
+       struct pardevice *pardev;
+       unsigned long  trigger;
+       unsigned long  nibble;
+       enum plip_connection_state connection;
+       unsigned short timeout_count;
+       int is_deferred;
+       int port_owner;
+       int should_relinquish;
+       spinlock_t lock;
+       atomic_t kill_timer;
+       struct completion killed_timer_cmp;
+};
+
+static inline void enable_parport_interrupts (struct net_device *dev)
+{
+       if (dev->irq != -1)
+       {
+               struct parport *port =
+                  ((struct net_local *)netdev_priv(dev))->pardev->port;
+               port->ops->enable_irq (port);
+       }
+}
+
+static inline void disable_parport_interrupts (struct net_device *dev)
+{
+       if (dev->irq != -1)
+       {
+               struct parport *port =
+                  ((struct net_local *)netdev_priv(dev))->pardev->port;
+               port->ops->disable_irq (port);
+       }
+}
+
+static inline void write_data (struct net_device *dev, unsigned char data)
+{
+       struct parport *port =
+          ((struct net_local *)netdev_priv(dev))->pardev->port;
+
+       port->ops->write_data (port, data);
+}
+
+static inline unsigned char read_status (struct net_device *dev)
+{
+       struct parport *port =
+          ((struct net_local *)netdev_priv(dev))->pardev->port;
+
+       return port->ops->read_status (port);
+}
+
+static const struct header_ops plip_header_ops = {
+       .create = plip_hard_header,
+       .cache  = plip_hard_header_cache,
+};
+
+static const struct net_device_ops plip_netdev_ops = {
+       .ndo_open                = plip_open,
+       .ndo_stop                = plip_close,
+       .ndo_start_xmit          = plip_tx_packet,
+       .ndo_do_ioctl            = plip_ioctl,
+       .ndo_change_mtu          = eth_change_mtu,
+       .ndo_set_mac_address     = eth_mac_addr,
+       .ndo_validate_addr       = eth_validate_addr,
+};
+
+/* Entry point of PLIP driver.
+   Probe the hardware, and register/initialize the driver.
+
+   PLIP is rather weird, because of the way it interacts with the parport
+   system.  It is _not_ initialised from Space.c.  Instead, plip_init()
+   is called, and that function makes up a "struct net_device" for each port, and
+   then calls us here.
+
+   */
+static void
+plip_init_netdev(struct net_device *dev)
+{
+       struct net_local *nl = netdev_priv(dev);
+
+       /* Then, override parts of it */
+       dev->tx_queue_len        = 10;
+       dev->flags               = IFF_POINTOPOINT|IFF_NOARP;
+       memset(dev->dev_addr, 0xfc, ETH_ALEN);
+
+       dev->netdev_ops          = &plip_netdev_ops;
+       dev->header_ops          = &plip_header_ops;
+
+
+       nl->port_owner = 0;
+
+       /* Initialize constants */
+       nl->trigger     = PLIP_TRIGGER_WAIT;
+       nl->nibble      = PLIP_NIBBLE_WAIT;
+
+       /* Initialize task queue structures */
+       INIT_WORK(&nl->immediate, plip_bh);
+       INIT_DELAYED_WORK(&nl->deferred, plip_kick_bh);
+
+       if (dev->irq == -1)
+               INIT_DELAYED_WORK(&nl->timer, plip_timer_bh);
+
+       spin_lock_init(&nl->lock);
+}
+
+/* Bottom half handler for the delayed request.
+   This routine is kicked by do_timer().
+   Request `plip_bh' to be invoked. */
+static void
+plip_kick_bh(struct work_struct *work)
+{
+       struct net_local *nl =
+               container_of(work, struct net_local, deferred.work);
+
+       if (nl->is_deferred)
+               schedule_work(&nl->immediate);
+}
+
+/* Forward declarations of internal routines */
+static int plip_none(struct net_device *, struct net_local *,
+                    struct plip_local *, struct plip_local *);
+static int plip_receive_packet(struct net_device *, struct net_local *,
+                              struct plip_local *, struct plip_local *);
+static int plip_send_packet(struct net_device *, struct net_local *,
+                           struct plip_local *, struct plip_local *);
+static int plip_connection_close(struct net_device *, struct net_local *,
+                                struct plip_local *, struct plip_local *);
+static int plip_error(struct net_device *, struct net_local *,
+                     struct plip_local *, struct plip_local *);
+static int plip_bh_timeout_error(struct net_device *dev, struct net_local *nl,
+                                struct plip_local *snd,
+                                struct plip_local *rcv,
+                                int error);
+
+#define OK        0
+#define TIMEOUT   1
+#define ERROR     2
+#define HS_TIMEOUT     3
+
+typedef int (*plip_func)(struct net_device *dev, struct net_local *nl,
+                        struct plip_local *snd, struct plip_local *rcv);
+
+static const plip_func connection_state_table[] =
+{
+       plip_none,
+       plip_receive_packet,
+       plip_send_packet,
+       plip_connection_close,
+       plip_error
+};
+
+/* Bottom half handler of PLIP. */
+static void
+plip_bh(struct work_struct *work)
+{
+       struct net_local *nl = container_of(work, struct net_local, immediate);
+       struct plip_local *snd = &nl->snd_data;
+       struct plip_local *rcv = &nl->rcv_data;
+       plip_func f;
+       int r;
+
+       nl->is_deferred = 0;
+       f = connection_state_table[nl->connection];
+       if ((r = (*f)(nl->dev, nl, snd, rcv)) != OK &&
+           (r = plip_bh_timeout_error(nl->dev, nl, snd, rcv, r)) != OK) {
+               nl->is_deferred = 1;
+               schedule_delayed_work(&nl->deferred, 1);
+       }
+}
+
+static void
+plip_timer_bh(struct work_struct *work)
+{
+       struct net_local *nl =
+               container_of(work, struct net_local, timer.work);
+
+       if (!(atomic_read (&nl->kill_timer))) {
+               plip_interrupt (nl->dev);
+
+               schedule_delayed_work(&nl->timer, 1);
+       }
+       else {
+               complete(&nl->killed_timer_cmp);
+       }
+}
+
+static int
+plip_bh_timeout_error(struct net_device *dev, struct net_local *nl,
+                     struct plip_local *snd, struct plip_local *rcv,
+                     int error)
+{
+       unsigned char c0;
+       /*
+        * This is tricky. If we got here from the beginning of send (either
+        * with ERROR or HS_TIMEOUT) we have IRQ enabled. Otherwise it's
+        * already disabled. With the old variant of {enable,disable}_irq()
+        * extra disable_irq() was a no-op. Now it became mortal - it's
+        * unbalanced and thus we'll never re-enable IRQ (until rmmod plip,
+        * that is). So we have to treat HS_TIMEOUT and ERROR from send
+        * in a special way.
+        */
+
+       spin_lock_irq(&nl->lock);
+       if (nl->connection == PLIP_CN_SEND) {
+
+               if (error != ERROR) { /* Timeout */
+                       nl->timeout_count++;
+                       if ((error == HS_TIMEOUT && nl->timeout_count <= 10) ||
+                           nl->timeout_count <= 3) {
+                               spin_unlock_irq(&nl->lock);
+                               /* Try again later */
+                               return TIMEOUT;
+                       }
+                       c0 = read_status(dev);
+                       printk(KERN_WARNING "%s: transmit timeout(%d,%02x)\n",
+                              dev->name, snd->state, c0);
+               } else
+                       error = HS_TIMEOUT;
+               dev->stats.tx_errors++;
+               dev->stats.tx_aborted_errors++;
+       } else if (nl->connection == PLIP_CN_RECEIVE) {
+               if (rcv->state == PLIP_PK_TRIGGER) {
+                       /* Transmission was interrupted. */
+                       spin_unlock_irq(&nl->lock);
+                       return OK;
+               }
+               if (error != ERROR) { /* Timeout */
+                       if (++nl->timeout_count <= 3) {
+                               spin_unlock_irq(&nl->lock);
+                               /* Try again later */
+                               return TIMEOUT;
+                       }
+                       c0 = read_status(dev);
+                       printk(KERN_WARNING "%s: receive timeout(%d,%02x)\n",
+                              dev->name, rcv->state, c0);
+               }
+               dev->stats.rx_dropped++;
+       }
+       rcv->state = PLIP_PK_DONE;
+       if (rcv->skb) {
+               kfree_skb(rcv->skb);
+               rcv->skb = NULL;
+       }
+       snd->state = PLIP_PK_DONE;
+       if (snd->skb) {
+               dev_kfree_skb(snd->skb);
+               snd->skb = NULL;
+       }
+       spin_unlock_irq(&nl->lock);
+       if (error == HS_TIMEOUT) {
+               DISABLE(dev->irq);
+               synchronize_irq(dev->irq);
+       }
+       disable_parport_interrupts (dev);
+       netif_stop_queue (dev);
+       nl->connection = PLIP_CN_ERROR;
+       write_data (dev, 0x00);
+
+       return TIMEOUT;
+}
+
+static int
+plip_none(struct net_device *dev, struct net_local *nl,
+         struct plip_local *snd, struct plip_local *rcv)
+{
+       return OK;
+}
+
+/* PLIP_RECEIVE --- receive a byte(two nibbles)
+   Returns OK on success, TIMEOUT on timeout */
+static inline int
+plip_receive(unsigned short nibble_timeout, struct net_device *dev,
+            enum plip_nibble_state *ns_p, unsigned char *data_p)
+{
+       unsigned char c0, c1;
+       unsigned int cx;
+
+       switch (*ns_p) {
+       case PLIP_NB_BEGIN:
+               cx = nibble_timeout;
+               while (1) {
+                       c0 = read_status(dev);
+                       udelay(PLIP_DELAY_UNIT);
+                       if ((c0 & 0x80) == 0) {
+                               c1 = read_status(dev);
+                               if (c0 == c1)
+                                       break;
+                       }
+                       if (--cx == 0)
+                               return TIMEOUT;
+               }
+               *data_p = (c0 >> 3) & 0x0f;
+               write_data (dev, 0x10); /* send ACK */
+               *ns_p = PLIP_NB_1;
+
+       case PLIP_NB_1:
+               cx = nibble_timeout;
+               while (1) {
+                       c0 = read_status(dev);
+                       udelay(PLIP_DELAY_UNIT);
+                       if (c0 & 0x80) {
+                               c1 = read_status(dev);
+                               if (c0 == c1)
+                                       break;
+                       }
+                       if (--cx == 0)
+                               return TIMEOUT;
+               }
+               *data_p |= (c0 << 1) & 0xf0;
+               write_data (dev, 0x00); /* send ACK */
+               *ns_p = PLIP_NB_BEGIN;
+       case PLIP_NB_2:
+               break;
+       }
+       return OK;
+}
+
+/*
+ *     Determine the packet's protocol ID. The rule here is that we
+ *     assume 802.3 if the type field is short enough to be a length.
+ *     This is normal practice and works for any 'now in use' protocol.
+ *
+ *     PLIP is ethernet ish but the daddr might not be valid if unicast.
+ *     PLIP fortunately has no bus architecture (its Point-to-point).
+ *
+ *     We can't fix the daddr thing as that quirk (more bug) is embedded
+ *     in far too many old systems not all even running Linux.
+ */
+
+static __be16 plip_type_trans(struct sk_buff *skb, struct net_device *dev)
+{
+       struct ethhdr *eth;
+       unsigned char *rawp;
+
+       skb_reset_mac_header(skb);
+       skb_pull(skb,dev->hard_header_len);
+       eth = eth_hdr(skb);
+
+       if(*eth->h_dest&1)
+       {
+               if(memcmp(eth->h_dest,dev->broadcast, ETH_ALEN)==0)
+                       skb->pkt_type=PACKET_BROADCAST;
+               else
+                       skb->pkt_type=PACKET_MULTICAST;
+       }
+
+       /*
+        *      This ALLMULTI check should be redundant by 1.4
+        *      so don't forget to remove it.
+        */
+
+       if (ntohs(eth->h_proto) >= 1536)
+               return eth->h_proto;
+
+       rawp = skb->data;
+
+       /*
+        *      This is a magic hack to spot IPX packets. Older Novell breaks
+        *      the protocol design and runs IPX over 802.3 without an 802.2 LLC
+        *      layer. We look for FFFF which isn't a used 802.2 SSAP/DSAP. This
+        *      won't work for fault tolerant netware but does for the rest.
+        */
+       if (*(unsigned short *)rawp == 0xFFFF)
+               return htons(ETH_P_802_3);
+
+       /*
+        *      Real 802.2 LLC
+        */
+       return htons(ETH_P_802_2);
+}
+
+/* PLIP_RECEIVE_PACKET --- receive a packet */
+static int
+plip_receive_packet(struct net_device *dev, struct net_local *nl,
+                   struct plip_local *snd, struct plip_local *rcv)
+{
+       unsigned short nibble_timeout = nl->nibble;
+       unsigned char *lbuf;
+
+       switch (rcv->state) {
+       case PLIP_PK_TRIGGER:
+               DISABLE(dev->irq);
+               /* Don't need to synchronize irq, as we can safely ignore it */
+               disable_parport_interrupts (dev);
+               write_data (dev, 0x01); /* send ACK */
+               if (net_debug > 2)
+                       printk(KERN_DEBUG "%s: receive start\n", dev->name);
+               rcv->state = PLIP_PK_LENGTH_LSB;
+               rcv->nibble = PLIP_NB_BEGIN;
+
+       case PLIP_PK_LENGTH_LSB:
+               if (snd->state != PLIP_PK_DONE) {
+                       if (plip_receive(nl->trigger, dev,
+                                        &rcv->nibble, &rcv->length.b.lsb)) {
+                               /* collision, here dev->tbusy == 1 */
+                               rcv->state = PLIP_PK_DONE;
+                               nl->is_deferred = 1;
+                               nl->connection = PLIP_CN_SEND;
+                               schedule_delayed_work(&nl->deferred, 1);
+                               enable_parport_interrupts (dev);
+                               ENABLE(dev->irq);
+                               return OK;
+                       }
+               } else {
+                       if (plip_receive(nibble_timeout, dev,
+                                        &rcv->nibble, &rcv->length.b.lsb))
+                               return TIMEOUT;
+               }
+               rcv->state = PLIP_PK_LENGTH_MSB;
+
+       case PLIP_PK_LENGTH_MSB:
+               if (plip_receive(nibble_timeout, dev,
+                                &rcv->nibble, &rcv->length.b.msb))
+                       return TIMEOUT;
+               if (rcv->length.h > dev->mtu + dev->hard_header_len ||
+                   rcv->length.h < 8) {
+                       printk(KERN_WARNING "%s: bogus packet size %d.\n", dev->name, rcv->length.h);
+                       return ERROR;
+               }
+               /* Malloc up new buffer. */
+               rcv->skb = dev_alloc_skb(rcv->length.h + 2);
+               if (rcv->skb == NULL) {
+                       printk(KERN_ERR "%s: Memory squeeze.\n", dev->name);
+                       return ERROR;
+               }
+               skb_reserve(rcv->skb, 2);       /* Align IP on 16 byte boundaries */
+               skb_put(rcv->skb,rcv->length.h);
+               rcv->skb->dev = dev;
+               rcv->state = PLIP_PK_DATA;
+               rcv->byte = 0;
+               rcv->checksum = 0;
+
+       case PLIP_PK_DATA:
+               lbuf = rcv->skb->data;
+               do {
+                       if (plip_receive(nibble_timeout, dev,
+                                        &rcv->nibble, &lbuf[rcv->byte]))
+                               return TIMEOUT;
+               } while (++rcv->byte < rcv->length.h);
+               do {
+                       rcv->checksum += lbuf[--rcv->byte];
+               } while (rcv->byte);
+               rcv->state = PLIP_PK_CHECKSUM;
+
+       case PLIP_PK_CHECKSUM:
+               if (plip_receive(nibble_timeout, dev,
+                                &rcv->nibble, &rcv->data))
+                       return TIMEOUT;
+               if (rcv->data != rcv->checksum) {
+                       dev->stats.rx_crc_errors++;
+                       if (net_debug)
+                               printk(KERN_DEBUG "%s: checksum error\n", dev->name);
+                       return ERROR;
+               }
+               rcv->state = PLIP_PK_DONE;
+
+       case PLIP_PK_DONE:
+               /* Inform the upper layer for the arrival of a packet. */
+               rcv->skb->protocol=plip_type_trans(rcv->skb, dev);
+               netif_rx_ni(rcv->skb);
+               dev->stats.rx_bytes += rcv->length.h;
+               dev->stats.rx_packets++;
+               rcv->skb = NULL;
+               if (net_debug > 2)
+                       printk(KERN_DEBUG "%s: receive end\n", dev->name);
+
+               /* Close the connection. */
+               write_data (dev, 0x00);
+               spin_lock_irq(&nl->lock);
+               if (snd->state != PLIP_PK_DONE) {
+                       nl->connection = PLIP_CN_SEND;
+                       spin_unlock_irq(&nl->lock);
+                       schedule_work(&nl->immediate);
+                       enable_parport_interrupts (dev);
+                       ENABLE(dev->irq);
+                       return OK;
+               } else {
+                       nl->connection = PLIP_CN_NONE;
+                       spin_unlock_irq(&nl->lock);
+                       enable_parport_interrupts (dev);
+                       ENABLE(dev->irq);
+                       return OK;
+               }
+       }
+       return OK;
+}
+
+/* PLIP_SEND --- send a byte (two nibbles)
+   Returns OK on success, TIMEOUT when timeout    */
+static inline int
+plip_send(unsigned short nibble_timeout, struct net_device *dev,
+         enum plip_nibble_state *ns_p, unsigned char data)
+{
+       unsigned char c0;
+       unsigned int cx;
+
+       switch (*ns_p) {
+       case PLIP_NB_BEGIN:
+               write_data (dev, data & 0x0f);
+               *ns_p = PLIP_NB_1;
+
+       case PLIP_NB_1:
+               write_data (dev, 0x10 | (data & 0x0f));
+               cx = nibble_timeout;
+               while (1) {
+                       c0 = read_status(dev);
+                       if ((c0 & 0x80) == 0)
+                               break;
+                       if (--cx == 0)
+                               return TIMEOUT;
+                       udelay(PLIP_DELAY_UNIT);
+               }
+               write_data (dev, 0x10 | (data >> 4));
+               *ns_p = PLIP_NB_2;
+
+       case PLIP_NB_2:
+               write_data (dev, (data >> 4));
+               cx = nibble_timeout;
+               while (1) {
+                       c0 = read_status(dev);
+                       if (c0 & 0x80)
+                               break;
+                       if (--cx == 0)
+                               return TIMEOUT;
+                       udelay(PLIP_DELAY_UNIT);
+               }
+               *ns_p = PLIP_NB_BEGIN;
+               return OK;
+       }
+       return OK;
+}
+
+/* PLIP_SEND_PACKET --- send a packet */
+static int
+plip_send_packet(struct net_device *dev, struct net_local *nl,
+                struct plip_local *snd, struct plip_local *rcv)
+{
+       unsigned short nibble_timeout = nl->nibble;
+       unsigned char *lbuf;
+       unsigned char c0;
+       unsigned int cx;
+
+       if (snd->skb == NULL || (lbuf = snd->skb->data) == NULL) {
+               printk(KERN_DEBUG "%s: send skb lost\n", dev->name);
+               snd->state = PLIP_PK_DONE;
+               snd->skb = NULL;
+               return ERROR;
+       }
+
+       switch (snd->state) {
+       case PLIP_PK_TRIGGER:
+               if ((read_status(dev) & 0xf8) != 0x80)
+                       return HS_TIMEOUT;
+
+               /* Trigger remote rx interrupt. */
+               write_data (dev, 0x08);
+               cx = nl->trigger;
+               while (1) {
+                       udelay(PLIP_DELAY_UNIT);
+                       spin_lock_irq(&nl->lock);
+                       if (nl->connection == PLIP_CN_RECEIVE) {
+                               spin_unlock_irq(&nl->lock);
+                               /* Interrupted. */
+                               dev->stats.collisions++;
+                               return OK;
+                       }
+                       c0 = read_status(dev);
+                       if (c0 & 0x08) {
+                               spin_unlock_irq(&nl->lock);
+                               DISABLE(dev->irq);
+                               synchronize_irq(dev->irq);
+                               if (nl->connection == PLIP_CN_RECEIVE) {
+                                       /* Interrupted.
+                                          We don't need to enable irq,
+                                          as it is soon disabled.    */
+                                       /* Yes, we do. New variant of
+                                          {enable,disable}_irq *counts*
+                                          them.  -- AV  */
+                                       ENABLE(dev->irq);
+                                       dev->stats.collisions++;
+                                       return OK;
+                               }
+                               disable_parport_interrupts (dev);
+                               if (net_debug > 2)
+                                       printk(KERN_DEBUG "%s: send start\n", dev->name);
+                               snd->state = PLIP_PK_LENGTH_LSB;
+                               snd->nibble = PLIP_NB_BEGIN;
+                               nl->timeout_count = 0;
+                               break;
+                       }
+                       spin_unlock_irq(&nl->lock);
+                       if (--cx == 0) {
+                               write_data (dev, 0x00);
+                               return HS_TIMEOUT;
+                       }
+               }
+
+       case PLIP_PK_LENGTH_LSB:
+               if (plip_send(nibble_timeout, dev,
+                             &snd->nibble, snd->length.b.lsb))
+                       return TIMEOUT;
+               snd->state = PLIP_PK_LENGTH_MSB;
+
+       case PLIP_PK_LENGTH_MSB:
+               if (plip_send(nibble_timeout, dev,
+                             &snd->nibble, snd->length.b.msb))
+                       return TIMEOUT;
+               snd->state = PLIP_PK_DATA;
+               snd->byte = 0;
+               snd->checksum = 0;
+
+       case PLIP_PK_DATA:
+               do {
+                       if (plip_send(nibble_timeout, dev,
+                                     &snd->nibble, lbuf[snd->byte]))
+                               return TIMEOUT;
+               } while (++snd->byte < snd->length.h);
+               do {
+                       snd->checksum += lbuf[--snd->byte];
+               } while (snd->byte);
+               snd->state = PLIP_PK_CHECKSUM;
+
+       case PLIP_PK_CHECKSUM:
+               if (plip_send(nibble_timeout, dev,
+                             &snd->nibble, snd->checksum))
+                       return TIMEOUT;
+
+               dev->stats.tx_bytes += snd->skb->len;
+               dev_kfree_skb(snd->skb);
+               dev->stats.tx_packets++;
+               snd->state = PLIP_PK_DONE;
+
+       case PLIP_PK_DONE:
+               /* Close the connection */
+               write_data (dev, 0x00);
+               snd->skb = NULL;
+               if (net_debug > 2)
+                       printk(KERN_DEBUG "%s: send end\n", dev->name);
+               nl->connection = PLIP_CN_CLOSING;
+               nl->is_deferred = 1;
+               schedule_delayed_work(&nl->deferred, 1);
+               enable_parport_interrupts (dev);
+               ENABLE(dev->irq);
+               return OK;
+       }
+       return OK;
+}
+
+static int
+plip_connection_close(struct net_device *dev, struct net_local *nl,
+                     struct plip_local *snd, struct plip_local *rcv)
+{
+       spin_lock_irq(&nl->lock);
+       if (nl->connection == PLIP_CN_CLOSING) {
+               nl->connection = PLIP_CN_NONE;
+               netif_wake_queue (dev);
+       }
+       spin_unlock_irq(&nl->lock);
+       if (nl->should_relinquish) {
+               nl->should_relinquish = nl->port_owner = 0;
+               parport_release(nl->pardev);
+       }
+       return OK;
+}
+
+/* PLIP_ERROR --- wait till other end settled */
+static int
+plip_error(struct net_device *dev, struct net_local *nl,
+          struct plip_local *snd, struct plip_local *rcv)
+{
+       unsigned char status;
+
+       status = read_status(dev);
+       if ((status & 0xf8) == 0x80) {
+               if (net_debug > 2)
+                       printk(KERN_DEBUG "%s: reset interface.\n", dev->name);
+               nl->connection = PLIP_CN_NONE;
+               nl->should_relinquish = 0;
+               netif_start_queue (dev);
+               enable_parport_interrupts (dev);
+               ENABLE(dev->irq);
+               netif_wake_queue (dev);
+       } else {
+               nl->is_deferred = 1;
+               schedule_delayed_work(&nl->deferred, 1);
+       }
+
+       return OK;
+}
+
+/* Handle the parallel port interrupts. */
+static void
+plip_interrupt(void *dev_id)
+{
+       struct net_device *dev = dev_id;
+       struct net_local *nl;
+       struct plip_local *rcv;
+       unsigned char c0;
+       unsigned long flags;
+
+       nl = netdev_priv(dev);
+       rcv = &nl->rcv_data;
+
+       spin_lock_irqsave (&nl->lock, flags);
+
+       c0 = read_status(dev);
+       if ((c0 & 0xf8) != 0xc0) {
+               if ((dev->irq != -1) && (net_debug > 1))
+                       printk(KERN_DEBUG "%s: spurious interrupt\n", dev->name);
+               spin_unlock_irqrestore (&nl->lock, flags);
+               return;
+       }
+
+       if (net_debug > 3)
+               printk(KERN_DEBUG "%s: interrupt.\n", dev->name);
+
+       switch (nl->connection) {
+       case PLIP_CN_CLOSING:
+               netif_wake_queue (dev);
+       case PLIP_CN_NONE:
+       case PLIP_CN_SEND:
+               rcv->state = PLIP_PK_TRIGGER;
+               nl->connection = PLIP_CN_RECEIVE;
+               nl->timeout_count = 0;
+               schedule_work(&nl->immediate);
+               break;
+
+       case PLIP_CN_RECEIVE:
+               /* May occur because there is race condition
+                  around test and set of dev->interrupt.
+                  Ignore this interrupt. */
+               break;
+
+       case PLIP_CN_ERROR:
+               printk(KERN_ERR "%s: receive interrupt in error state\n", dev->name);
+               break;
+       }
+
+       spin_unlock_irqrestore(&nl->lock, flags);
+}
+
+static int
+plip_tx_packet(struct sk_buff *skb, struct net_device *dev)
+{
+       struct net_local *nl = netdev_priv(dev);
+       struct plip_local *snd = &nl->snd_data;
+
+       if (netif_queue_stopped(dev))
+               return NETDEV_TX_BUSY;
+
+       /* We may need to grab the bus */
+       if (!nl->port_owner) {
+               if (parport_claim(nl->pardev))
+                       return NETDEV_TX_BUSY;
+               nl->port_owner = 1;
+       }
+
+       netif_stop_queue (dev);
+
+       if (skb->len > dev->mtu + dev->hard_header_len) {
+               printk(KERN_WARNING "%s: packet too big, %d.\n", dev->name, (int)skb->len);
+               netif_start_queue (dev);
+               return NETDEV_TX_BUSY;
+       }
+
+       if (net_debug > 2)
+               printk(KERN_DEBUG "%s: send request\n", dev->name);
+
+       spin_lock_irq(&nl->lock);
+       snd->skb = skb;
+       snd->length.h = skb->len;
+       snd->state = PLIP_PK_TRIGGER;
+       if (nl->connection == PLIP_CN_NONE) {
+               nl->connection = PLIP_CN_SEND;
+               nl->timeout_count = 0;
+       }
+       schedule_work(&nl->immediate);
+       spin_unlock_irq(&nl->lock);
+
+       return NETDEV_TX_OK;
+}
+
+static void
+plip_rewrite_address(const struct net_device *dev, struct ethhdr *eth)
+{
+       const struct in_device *in_dev;
+
+       rcu_read_lock();
+       in_dev = __in_dev_get_rcu(dev);
+       if (in_dev) {
+               /* Any address will do - we take the first */
+               const struct in_ifaddr *ifa = in_dev->ifa_list;
+               if (ifa) {
+                       memcpy(eth->h_source, dev->dev_addr, 6);
+                       memset(eth->h_dest, 0xfc, 2);
+                       memcpy(eth->h_dest+2, &ifa->ifa_address, 4);
+               }
+       }
+       rcu_read_unlock();
+}
+
+static int
+plip_hard_header(struct sk_buff *skb, struct net_device *dev,
+                unsigned short type, const void *daddr,
+                const void *saddr, unsigned len)
+{
+       int ret;
+
+       ret = eth_header(skb, dev, type, daddr, saddr, len);
+       if (ret >= 0)
+               plip_rewrite_address (dev, (struct ethhdr *)skb->data);
+
+       return ret;
+}
+
+static int plip_hard_header_cache(const struct neighbour *neigh,
+                                 struct hh_cache *hh, __be16 type)
+{
+       int ret;
+
+       ret = eth_header_cache(neigh, hh, type);
+       if (ret == 0) {
+               struct ethhdr *eth;
+
+               eth = (struct ethhdr*)(((u8*)hh->hh_data) +
+                                      HH_DATA_OFF(sizeof(*eth)));
+               plip_rewrite_address (neigh->dev, eth);
+       }
+
+       return ret;
+}
+
+/* Open/initialize the board.  This is called (in the current kernel)
+   sometime after booting when the 'ifconfig' program is run.
+
+   This routine gets exclusive access to the parallel port by allocating
+   its IRQ line.
+ */
+static int
+plip_open(struct net_device *dev)
+{
+       struct net_local *nl = netdev_priv(dev);
+       struct in_device *in_dev;
+
+       /* Grab the port */
+       if (!nl->port_owner) {
+               if (parport_claim(nl->pardev)) return -EAGAIN;
+               nl->port_owner = 1;
+       }
+
+       nl->should_relinquish = 0;
+
+       /* Clear the data port. */
+       write_data (dev, 0x00);
+
+       /* Enable rx interrupt. */
+       enable_parport_interrupts (dev);
+       if (dev->irq == -1)
+       {
+               atomic_set (&nl->kill_timer, 0);
+               schedule_delayed_work(&nl->timer, 1);
+       }
+
+       /* Initialize the state machine. */
+       nl->rcv_data.state = nl->snd_data.state = PLIP_PK_DONE;
+       nl->rcv_data.skb = nl->snd_data.skb = NULL;
+       nl->connection = PLIP_CN_NONE;
+       nl->is_deferred = 0;
+
+       /* Fill in the MAC-level header.
+          We used to abuse dev->broadcast to store the point-to-point
+          MAC address, but we no longer do it. Instead, we fetch the
+          interface address whenever it is needed, which is cheap enough
+          because we use the hh_cache. Actually, abusing dev->broadcast
+          didn't work, because when using plip_open the point-to-point
+          address isn't yet known.
+          PLIP doesn't have a real MAC address, but we need it to be
+          DOS compatible, and to properly support taps (otherwise,
+          when the device address isn't identical to the address of a
+          received frame, the kernel incorrectly drops it).             */
+
+       in_dev=__in_dev_get_rtnl(dev);
+       if (in_dev) {
+               /* Any address will do - we take the first. We already
+                  have the first two bytes filled with 0xfc, from
+                  plip_init_dev(). */
+               struct in_ifaddr *ifa=in_dev->ifa_list;
+               if (ifa != NULL) {
+                       memcpy(dev->dev_addr+2, &ifa->ifa_local, 4);
+               }
+       }
+
+       netif_start_queue (dev);
+
+       return 0;
+}
+
+/* The inverse routine to plip_open (). */
+static int
+plip_close(struct net_device *dev)
+{
+       struct net_local *nl = netdev_priv(dev);
+       struct plip_local *snd = &nl->snd_data;
+       struct plip_local *rcv = &nl->rcv_data;
+
+       netif_stop_queue (dev);
+       DISABLE(dev->irq);
+       synchronize_irq(dev->irq);
+
+       if (dev->irq == -1)
+       {
+               init_completion(&nl->killed_timer_cmp);
+               atomic_set (&nl->kill_timer, 1);
+               wait_for_completion(&nl->killed_timer_cmp);
+       }
+
+#ifdef NOTDEF
+       outb(0x00, PAR_DATA(dev));
+#endif
+       nl->is_deferred = 0;
+       nl->connection = PLIP_CN_NONE;
+       if (nl->port_owner) {
+               parport_release(nl->pardev);
+               nl->port_owner = 0;
+       }
+
+       snd->state = PLIP_PK_DONE;
+       if (snd->skb) {
+               dev_kfree_skb(snd->skb);
+               snd->skb = NULL;
+       }
+       rcv->state = PLIP_PK_DONE;
+       if (rcv->skb) {
+               kfree_skb(rcv->skb);
+               rcv->skb = NULL;
+       }
+
+#ifdef NOTDEF
+       /* Reset. */
+       outb(0x00, PAR_CONTROL(dev));
+#endif
+       return 0;
+}
+
+static int
+plip_preempt(void *handle)
+{
+       struct net_device *dev = (struct net_device *)handle;
+       struct net_local *nl = netdev_priv(dev);
+
+       /* Stand our ground if a datagram is on the wire */
+       if (nl->connection != PLIP_CN_NONE) {
+               nl->should_relinquish = 1;
+               return 1;
+       }
+
+       nl->port_owner = 0;     /* Remember that we released the bus */
+       return 0;
+}
+
+static void
+plip_wakeup(void *handle)
+{
+       struct net_device *dev = (struct net_device *)handle;
+       struct net_local *nl = netdev_priv(dev);
+
+       if (nl->port_owner) {
+               /* Why are we being woken up? */
+               printk(KERN_DEBUG "%s: why am I being woken up?\n", dev->name);
+               if (!parport_claim(nl->pardev))
+                       /* bus_owner is already set (but why?) */
+                       printk(KERN_DEBUG "%s: I'm broken.\n", dev->name);
+               else
+                       return;
+       }
+
+       if (!(dev->flags & IFF_UP))
+               /* Don't need the port when the interface is down */
+               return;
+
+       if (!parport_claim(nl->pardev)) {
+               nl->port_owner = 1;
+               /* Clear the data port. */
+               write_data (dev, 0x00);
+       }
+}
+
+static int
+plip_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
+{
+       struct net_local *nl = netdev_priv(dev);
+       struct plipconf *pc = (struct plipconf *) &rq->ifr_ifru;
+
+       if (cmd != SIOCDEVPLIP)
+               return -EOPNOTSUPP;
+
+       switch(pc->pcmd) {
+       case PLIP_GET_TIMEOUT:
+               pc->trigger = nl->trigger;
+               pc->nibble  = nl->nibble;
+               break;
+       case PLIP_SET_TIMEOUT:
+               if(!capable(CAP_NET_ADMIN))
+                       return -EPERM;
+               nl->trigger = pc->trigger;
+               nl->nibble  = pc->nibble;
+               break;
+       default:
+               return -EOPNOTSUPP;
+       }
+       return 0;
+}
+
+static int parport[PLIP_MAX] = { [0 ... PLIP_MAX-1] = -1 };
+static int timid;
+
+module_param_array(parport, int, NULL, 0);
+module_param(timid, int, 0);
+MODULE_PARM_DESC(parport, "List of parport device numbers to use by plip");
+
+static struct net_device *dev_plip[PLIP_MAX] = { NULL, };
+
+static inline int
+plip_searchfor(int list[], int a)
+{
+       int i;
+       for (i = 0; i < PLIP_MAX && list[i] != -1; i++) {
+               if (list[i] == a) return 1;
+       }
+       return 0;
+}
+
+/* plip_attach() is called (by the parport code) when a port is
+ * available to use. */
+static void plip_attach (struct parport *port)
+{
+       static int unit;
+       struct net_device *dev;
+       struct net_local *nl;
+       char name[IFNAMSIZ];
+
+       if ((parport[0] == -1 && (!timid || !port->devices)) ||
+           plip_searchfor(parport, port->number)) {
+               if (unit == PLIP_MAX) {
+                       printk(KERN_ERR "plip: too many devices\n");
+                       return;
+               }
+
+               sprintf(name, "plip%d", unit);
+               dev = alloc_etherdev(sizeof(struct net_local));
+               if (!dev) {
+                       printk(KERN_ERR "plip: memory squeeze\n");
+                       return;
+               }
+
+               strcpy(dev->name, name);
+
+               dev->irq = port->irq;
+               dev->base_addr = port->base;
+               if (port->irq == -1) {
+                       printk(KERN_INFO "plip: %s has no IRQ. Using IRQ-less mode,"
+                                "which is fairly inefficient!\n", port->name);
+               }
+
+               nl = netdev_priv(dev);
+               nl->dev = dev;
+               nl->pardev = parport_register_device(port, dev->name, plip_preempt,
+                                                plip_wakeup, plip_interrupt,
+                                                0, dev);
+
+               if (!nl->pardev) {
+                       printk(KERN_ERR "%s: parport_register failed\n", name);
+                       goto err_free_dev;
+               }
+
+               plip_init_netdev(dev);
+
+               if (register_netdev(dev)) {
+                       printk(KERN_ERR "%s: network register failed\n", name);
+                       goto err_parport_unregister;
+               }
+
+               printk(KERN_INFO "%s", version);
+               if (dev->irq != -1)
+                       printk(KERN_INFO "%s: Parallel port at %#3lx, "
+                                        "using IRQ %d.\n",
+                                        dev->name, dev->base_addr, dev->irq);
+               else
+                       printk(KERN_INFO "%s: Parallel port at %#3lx, "
+                                        "not using IRQ.\n",
+                                        dev->name, dev->base_addr);
+               dev_plip[unit++] = dev;
+       }
+       return;
+
+err_parport_unregister:
+       parport_unregister_device(nl->pardev);
+err_free_dev:
+       free_netdev(dev);
+}
+
+/* plip_detach() is called (by the parport code) when a port is
+ * no longer available to use. */
+static void plip_detach (struct parport *port)
+{
+       /* Nothing to do */
+}
+
+static struct parport_driver plip_driver = {
+       .name   = "plip",
+       .attach = plip_attach,
+       .detach = plip_detach
+};
+
+static void __exit plip_cleanup_module (void)
+{
+       struct net_device *dev;
+       int i;
+
+       parport_unregister_driver (&plip_driver);
+
+       for (i=0; i < PLIP_MAX; i++) {
+               if ((dev = dev_plip[i])) {
+                       struct net_local *nl = netdev_priv(dev);
+                       unregister_netdev(dev);
+                       if (nl->port_owner)
+                               parport_release(nl->pardev);
+                       parport_unregister_device(nl->pardev);
+                       free_netdev(dev);
+                       dev_plip[i] = NULL;
+               }
+       }
+}
+
+#ifndef MODULE
+
+static int parport_ptr;
+
+static int __init plip_setup(char *str)
+{
+       int ints[4];
+
+       str = get_options(str, ARRAY_SIZE(ints), ints);
+
+       /* Ugh. */
+       if (!strncmp(str, "parport", 7)) {
+               int n = simple_strtoul(str+7, NULL, 10);
+               if (parport_ptr < PLIP_MAX)
+                       parport[parport_ptr++] = n;
+               else
+                       printk(KERN_INFO "plip: too many ports, %s ignored.\n",
+                              str);
+       } else if (!strcmp(str, "timid")) {
+               timid = 1;
+       } else {
+               if (ints[0] == 0 || ints[1] == 0) {
+                       /* disable driver on "plip=" or "plip=0" */
+                       parport[0] = -2;
+               } else {
+                       printk(KERN_WARNING "warning: 'plip=0x%x' ignored\n",
+                              ints[1]);
+               }
+       }
+       return 1;
+}
+
+__setup("plip=", plip_setup);
+
+#endif /* !MODULE */
+
+static int __init plip_init (void)
+{
+       if (parport[0] == -2)
+               return 0;
+
+       if (parport[0] != -1 && timid) {
+               printk(KERN_WARNING "plip: warning, ignoring `timid' since specific ports given.\n");
+               timid = 0;
+       }
+
+       if (parport_register_driver (&plip_driver)) {
+               printk (KERN_WARNING "plip: couldn't register driver\n");
+               return 1;
+       }
+
+       return 0;
+}
+
+module_init(plip_init);
+module_exit(plip_cleanup_module);
+MODULE_LICENSE("GPL");