net: powerpc: remove the legacy iSeries ethernet driver
authorStephen Rothwell <sfr@canb.auug.org.au>
Sat, 10 Mar 2012 04:41:26 +0000 (23:41 -0500)
committerDavid S. Miller <davem@davemloft.net>
Sat, 10 Mar 2012 04:41:26 +0000 (23:41 -0500)
This driver is specific to the PowerPC legcay iSeries platform which is
being removed.

Cc: David Miller <davem@davemloft.net>
Cc: <netdev@vger.kernel.org>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
drivers/net/ethernet/ibm/Kconfig
drivers/net/ethernet/ibm/Makefile
drivers/net/ethernet/ibm/iseries_veth.c [deleted file]

index 9e16f3fa97b2c7ec918347de09887f4d04ad18d3..b9773d2291927e4164197b20ca78a00694c6528d 100644 (file)
@@ -29,10 +29,6 @@ config IBMVETH
          To compile this driver as a module, choose M here. The module will
          be called ibmveth.
 
-config ISERIES_VETH
-       tristate "iSeries Virtual Ethernet driver support"
-       depends on PPC_ISERIES
-
 source "drivers/net/ethernet/ibm/emac/Kconfig"
 
 config EHEA
index 5a7d4e9ac8035b3e9e34c6f683362a7edaef5bb6..2f04e71a5926c788bef7fc0a96b3bb5239d07958 100644 (file)
@@ -3,6 +3,5 @@
 #
 
 obj-$(CONFIG_IBMVETH) += ibmveth.o
-obj-$(CONFIG_ISERIES_VETH) += iseries_veth.o
 obj-$(CONFIG_IBM_EMAC) += emac/
 obj-$(CONFIG_EHEA) += ehea/
diff --git a/drivers/net/ethernet/ibm/iseries_veth.c b/drivers/net/ethernet/ibm/iseries_veth.c
deleted file mode 100644 (file)
index 1cafa65..0000000
+++ /dev/null
@@ -1,1708 +0,0 @@
-/* File veth.c created by Kyle A. Lucke on Mon Aug  7 2000. */
-/*
- * IBM eServer iSeries Virtual Ethernet Device Driver
- * Copyright (C) 2001 Kyle A. Lucke (klucke@us.ibm.com), IBM Corp.
- * Substantially cleaned up by:
- * Copyright (C) 2003 David Gibson <dwg@au1.ibm.com>, IBM Corporation.
- * Copyright (C) 2004-2005 Michael Ellerman, IBM Corporation.
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License as
- * published by the Free Software Foundation; either version 2 of the
- * License, or (at your option) any later version.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
- * USA
- *
- *
- * This module implements the virtual ethernet device for iSeries LPAR
- * Linux.  It uses hypervisor message passing to implement an
- * ethernet-like network device communicating between partitions on
- * the iSeries.
- *
- * The iSeries LPAR hypervisor currently allows for up to 16 different
- * virtual ethernets.  These are all dynamically configurable on
- * OS/400 partitions, but dynamic configuration is not supported under
- * Linux yet.  An ethXX network device will be created for each
- * virtual ethernet this partition is connected to.
- *
- * - This driver is responsible for routing packets to and from other
- *   partitions.  The MAC addresses used by the virtual ethernets
- *   contains meaning and must not be modified.
- *
- * - Having 2 virtual ethernets to the same remote partition DOES NOT
- *   double the available bandwidth.  The 2 devices will share the
- *   available hypervisor bandwidth.
- *
- * - If you send a packet to your own mac address, it will just be
- *   dropped, you won't get it on the receive side.
- *
- * - Multicast is implemented by sending the frame frame to every
- *   other partition.  It is the responsibility of the receiving
- *   partition to filter the addresses desired.
- *
- * Tunable parameters:
- *
- * VETH_NUMBUFFERS: This compile time option defaults to 120.  It
- * controls how much memory Linux will allocate per remote partition
- * it is communicating with.  It can be thought of as the maximum
- * number of packets outstanding to a remote partition at a time.
- */
-
-#include <linux/module.h>
-#include <linux/types.h>
-#include <linux/errno.h>
-#include <linux/ioport.h>
-#include <linux/kernel.h>
-#include <linux/netdevice.h>
-#include <linux/etherdevice.h>
-#include <linux/skbuff.h>
-#include <linux/init.h>
-#include <linux/delay.h>
-#include <linux/mm.h>
-#include <linux/ethtool.h>
-#include <linux/if_ether.h>
-#include <linux/slab.h>
-
-#include <asm/abs_addr.h>
-#include <asm/iseries/mf.h>
-#include <asm/uaccess.h>
-#include <asm/firmware.h>
-#include <asm/iseries/hv_lp_config.h>
-#include <asm/iseries/hv_types.h>
-#include <asm/iseries/hv_lp_event.h>
-#include <asm/iommu.h>
-#include <asm/vio.h>
-
-#undef DEBUG
-
-MODULE_AUTHOR("Kyle Lucke <klucke@us.ibm.com>");
-MODULE_DESCRIPTION("iSeries Virtual ethernet driver");
-MODULE_LICENSE("GPL");
-
-#define VETH_EVENT_CAP (0)
-#define VETH_EVENT_FRAMES      (1)
-#define VETH_EVENT_MONITOR     (2)
-#define VETH_EVENT_FRAMES_ACK  (3)
-
-#define VETH_MAX_ACKS_PER_MSG  (20)
-#define VETH_MAX_FRAMES_PER_MSG        (6)
-
-struct veth_frames_data {
-       u32 addr[VETH_MAX_FRAMES_PER_MSG];
-       u16 len[VETH_MAX_FRAMES_PER_MSG];
-       u32 eofmask;
-};
-#define VETH_EOF_SHIFT         (32-VETH_MAX_FRAMES_PER_MSG)
-
-struct veth_frames_ack_data {
-       u16 token[VETH_MAX_ACKS_PER_MSG];
-};
-
-struct veth_cap_data {
-       u8 caps_version;
-       u8 rsvd1;
-       u16 num_buffers;
-       u16 ack_threshold;
-       u16 rsvd2;
-       u32 ack_timeout;
-       u32 rsvd3;
-       u64 rsvd4[3];
-};
-
-struct veth_lpevent {
-       struct HvLpEvent base_event;
-       union {
-               struct veth_cap_data caps_data;
-               struct veth_frames_data frames_data;
-               struct veth_frames_ack_data frames_ack_data;
-       } u;
-
-};
-
-#define DRV_NAME       "iseries_veth"
-#define DRV_VERSION    "2.0"
-
-#define VETH_NUMBUFFERS                (120)
-#define VETH_ACKTIMEOUT        (1000000) /* microseconds */
-#define VETH_MAX_MCAST         (12)
-
-#define VETH_MAX_MTU           (9000)
-
-#if VETH_NUMBUFFERS < 10
-#define ACK_THRESHOLD          (1)
-#elif VETH_NUMBUFFERS < 20
-#define ACK_THRESHOLD          (4)
-#elif VETH_NUMBUFFERS < 40
-#define ACK_THRESHOLD          (10)
-#else
-#define ACK_THRESHOLD          (20)
-#endif
-
-#define        VETH_STATE_SHUTDOWN     (0x0001)
-#define VETH_STATE_OPEN                (0x0002)
-#define VETH_STATE_RESET       (0x0004)
-#define VETH_STATE_SENTMON     (0x0008)
-#define VETH_STATE_SENTCAPS    (0x0010)
-#define VETH_STATE_GOTCAPACK   (0x0020)
-#define VETH_STATE_GOTCAPS     (0x0040)
-#define VETH_STATE_SENTCAPACK  (0x0080)
-#define VETH_STATE_READY       (0x0100)
-
-struct veth_msg {
-       struct veth_msg *next;
-       struct veth_frames_data data;
-       int token;
-       int in_use;
-       struct sk_buff *skb;
-       struct device *dev;
-};
-
-struct veth_lpar_connection {
-       HvLpIndex remote_lp;
-       struct delayed_work statemachine_wq;
-       struct veth_msg *msgs;
-       int num_events;
-       struct veth_cap_data local_caps;
-
-       struct kobject kobject;
-       struct timer_list ack_timer;
-
-       struct timer_list reset_timer;
-       unsigned int reset_timeout;
-       unsigned long last_contact;
-       int outstanding_tx;
-
-       spinlock_t lock;
-       unsigned long state;
-       HvLpInstanceId src_inst;
-       HvLpInstanceId dst_inst;
-       struct veth_lpevent cap_event, cap_ack_event;
-       u16 pending_acks[VETH_MAX_ACKS_PER_MSG];
-       u32 num_pending_acks;
-
-       int num_ack_events;
-       struct veth_cap_data remote_caps;
-       u32 ack_timeout;
-
-       struct veth_msg *msg_stack_head;
-};
-
-struct veth_port {
-       struct device *dev;
-       u64 mac_addr;
-       HvLpIndexMap lpar_map;
-
-       /* queue_lock protects the stopped_map and dev's queue. */
-       spinlock_t queue_lock;
-       HvLpIndexMap stopped_map;
-
-       /* mcast_gate protects promiscuous, num_mcast & mcast_addr. */
-       rwlock_t mcast_gate;
-       int promiscuous;
-       int num_mcast;
-       u64 mcast_addr[VETH_MAX_MCAST];
-
-       struct kobject kobject;
-};
-
-static HvLpIndex this_lp;
-static struct veth_lpar_connection *veth_cnx[HVMAXARCHITECTEDLPS]; /* = 0 */
-static struct net_device *veth_dev[HVMAXARCHITECTEDVIRTUALLANS]; /* = 0 */
-
-static int veth_start_xmit(struct sk_buff *skb, struct net_device *dev);
-static void veth_recycle_msg(struct veth_lpar_connection *, struct veth_msg *);
-static void veth_wake_queues(struct veth_lpar_connection *cnx);
-static void veth_stop_queues(struct veth_lpar_connection *cnx);
-static void veth_receive(struct veth_lpar_connection *, struct veth_lpevent *);
-static void veth_release_connection(struct kobject *kobject);
-static void veth_timed_ack(unsigned long ptr);
-static void veth_timed_reset(unsigned long ptr);
-
-/*
- * Utility functions
- */
-
-#define veth_info(fmt, args...) \
-       printk(KERN_INFO DRV_NAME ": " fmt, ## args)
-
-#define veth_error(fmt, args...) \
-       printk(KERN_ERR DRV_NAME ": Error: " fmt, ## args)
-
-#ifdef DEBUG
-#define veth_debug(fmt, args...) \
-       printk(KERN_DEBUG DRV_NAME ": " fmt, ## args)
-#else
-#define veth_debug(fmt, args...) do {} while (0)
-#endif
-
-/* You must hold the connection's lock when you call this function. */
-static inline void veth_stack_push(struct veth_lpar_connection *cnx,
-                                  struct veth_msg *msg)
-{
-       msg->next = cnx->msg_stack_head;
-       cnx->msg_stack_head = msg;
-}
-
-/* You must hold the connection's lock when you call this function. */
-static inline struct veth_msg *veth_stack_pop(struct veth_lpar_connection *cnx)
-{
-       struct veth_msg *msg;
-
-       msg = cnx->msg_stack_head;
-       if (msg)
-               cnx->msg_stack_head = cnx->msg_stack_head->next;
-
-       return msg;
-}
-
-/* You must hold the connection's lock when you call this function. */
-static inline int veth_stack_is_empty(struct veth_lpar_connection *cnx)
-{
-       return cnx->msg_stack_head == NULL;
-}
-
-static inline HvLpEvent_Rc
-veth_signalevent(struct veth_lpar_connection *cnx, u16 subtype,
-                HvLpEvent_AckInd ackind, HvLpEvent_AckType acktype,
-                u64 token,
-                u64 data1, u64 data2, u64 data3, u64 data4, u64 data5)
-{
-       return HvCallEvent_signalLpEventFast(cnx->remote_lp,
-                                            HvLpEvent_Type_VirtualLan,
-                                            subtype, ackind, acktype,
-                                            cnx->src_inst,
-                                            cnx->dst_inst,
-                                            token, data1, data2, data3,
-                                            data4, data5);
-}
-
-static inline HvLpEvent_Rc veth_signaldata(struct veth_lpar_connection *cnx,
-                                          u16 subtype, u64 token, void *data)
-{
-       u64 *p = (u64 *) data;
-
-       return veth_signalevent(cnx, subtype, HvLpEvent_AckInd_NoAck,
-                               HvLpEvent_AckType_ImmediateAck,
-                               token, p[0], p[1], p[2], p[3], p[4]);
-}
-
-struct veth_allocation {
-       struct completion c;
-       int num;
-};
-
-static void veth_complete_allocation(void *parm, int number)
-{
-       struct veth_allocation *vc = (struct veth_allocation *)parm;
-
-       vc->num = number;
-       complete(&vc->c);
-}
-
-static int veth_allocate_events(HvLpIndex rlp, int number)
-{
-       struct veth_allocation vc =
-               { COMPLETION_INITIALIZER_ONSTACK(vc.c), 0 };
-
-       mf_allocate_lp_events(rlp, HvLpEvent_Type_VirtualLan,
-                           sizeof(struct veth_lpevent), number,
-                           &veth_complete_allocation, &vc);
-       wait_for_completion(&vc.c);
-
-       return vc.num;
-}
-
-/*
- * sysfs support
- */
-
-struct veth_cnx_attribute {
-       struct attribute attr;
-       ssize_t (*show)(struct veth_lpar_connection *, char *buf);
-       ssize_t (*store)(struct veth_lpar_connection *, const char *buf);
-};
-
-static ssize_t veth_cnx_attribute_show(struct kobject *kobj,
-               struct attribute *attr, char *buf)
-{
-       struct veth_cnx_attribute *cnx_attr;
-       struct veth_lpar_connection *cnx;
-
-       cnx_attr = container_of(attr, struct veth_cnx_attribute, attr);
-       cnx = container_of(kobj, struct veth_lpar_connection, kobject);
-
-       if (!cnx_attr->show)
-               return -EIO;
-
-       return cnx_attr->show(cnx, buf);
-}
-
-#define CUSTOM_CNX_ATTR(_name, _format, _expression)                   \
-static ssize_t _name##_show(struct veth_lpar_connection *cnx, char *buf)\
-{                                                                      \
-       return sprintf(buf, _format, _expression);                      \
-}                                                                      \
-struct veth_cnx_attribute veth_cnx_attr_##_name = __ATTR_RO(_name)
-
-#define SIMPLE_CNX_ATTR(_name) \
-       CUSTOM_CNX_ATTR(_name, "%lu\n", (unsigned long)cnx->_name)
-
-SIMPLE_CNX_ATTR(outstanding_tx);
-SIMPLE_CNX_ATTR(remote_lp);
-SIMPLE_CNX_ATTR(num_events);
-SIMPLE_CNX_ATTR(src_inst);
-SIMPLE_CNX_ATTR(dst_inst);
-SIMPLE_CNX_ATTR(num_pending_acks);
-SIMPLE_CNX_ATTR(num_ack_events);
-CUSTOM_CNX_ATTR(ack_timeout, "%d\n", jiffies_to_msecs(cnx->ack_timeout));
-CUSTOM_CNX_ATTR(reset_timeout, "%d\n", jiffies_to_msecs(cnx->reset_timeout));
-CUSTOM_CNX_ATTR(state, "0x%.4lX\n", cnx->state);
-CUSTOM_CNX_ATTR(last_contact, "%d\n", cnx->last_contact ?
-               jiffies_to_msecs(jiffies - cnx->last_contact) : 0);
-
-#define GET_CNX_ATTR(_name)    (&veth_cnx_attr_##_name.attr)
-
-static struct attribute *veth_cnx_default_attrs[] = {
-       GET_CNX_ATTR(outstanding_tx),
-       GET_CNX_ATTR(remote_lp),
-       GET_CNX_ATTR(num_events),
-       GET_CNX_ATTR(reset_timeout),
-       GET_CNX_ATTR(last_contact),
-       GET_CNX_ATTR(state),
-       GET_CNX_ATTR(src_inst),
-       GET_CNX_ATTR(dst_inst),
-       GET_CNX_ATTR(num_pending_acks),
-       GET_CNX_ATTR(num_ack_events),
-       GET_CNX_ATTR(ack_timeout),
-       NULL
-};
-
-static const struct sysfs_ops veth_cnx_sysfs_ops = {
-               .show = veth_cnx_attribute_show
-};
-
-static struct kobj_type veth_lpar_connection_ktype = {
-       .release        = veth_release_connection,
-       .sysfs_ops      = &veth_cnx_sysfs_ops,
-       .default_attrs  = veth_cnx_default_attrs
-};
-
-struct veth_port_attribute {
-       struct attribute attr;
-       ssize_t (*show)(struct veth_port *, char *buf);
-       ssize_t (*store)(struct veth_port *, const char *buf);
-};
-
-static ssize_t veth_port_attribute_show(struct kobject *kobj,
-               struct attribute *attr, char *buf)
-{
-       struct veth_port_attribute *port_attr;
-       struct veth_port *port;
-
-       port_attr = container_of(attr, struct veth_port_attribute, attr);
-       port = container_of(kobj, struct veth_port, kobject);
-
-       if (!port_attr->show)
-               return -EIO;
-
-       return port_attr->show(port, buf);
-}
-
-#define CUSTOM_PORT_ATTR(_name, _format, _expression)                  \
-static ssize_t _name##_show(struct veth_port *port, char *buf)         \
-{                                                                      \
-       return sprintf(buf, _format, _expression);                      \
-}                                                                      \
-struct veth_port_attribute veth_port_attr_##_name = __ATTR_RO(_name)
-
-#define SIMPLE_PORT_ATTR(_name)        \
-       CUSTOM_PORT_ATTR(_name, "%lu\n", (unsigned long)port->_name)
-
-SIMPLE_PORT_ATTR(promiscuous);
-SIMPLE_PORT_ATTR(num_mcast);
-CUSTOM_PORT_ATTR(lpar_map, "0x%X\n", port->lpar_map);
-CUSTOM_PORT_ATTR(stopped_map, "0x%X\n", port->stopped_map);
-CUSTOM_PORT_ATTR(mac_addr, "0x%llX\n", port->mac_addr);
-
-#define GET_PORT_ATTR(_name)   (&veth_port_attr_##_name.attr)
-static struct attribute *veth_port_default_attrs[] = {
-       GET_PORT_ATTR(mac_addr),
-       GET_PORT_ATTR(lpar_map),
-       GET_PORT_ATTR(stopped_map),
-       GET_PORT_ATTR(promiscuous),
-       GET_PORT_ATTR(num_mcast),
-       NULL
-};
-
-static const struct sysfs_ops veth_port_sysfs_ops = {
-       .show = veth_port_attribute_show
-};
-
-static struct kobj_type veth_port_ktype = {
-       .sysfs_ops      = &veth_port_sysfs_ops,
-       .default_attrs  = veth_port_default_attrs
-};
-
-/*
- * LPAR connection code
- */
-
-static inline void veth_kick_statemachine(struct veth_lpar_connection *cnx)
-{
-       schedule_delayed_work(&cnx->statemachine_wq, 0);
-}
-
-static void veth_take_cap(struct veth_lpar_connection *cnx,
-                         struct veth_lpevent *event)
-{
-       unsigned long flags;
-
-       spin_lock_irqsave(&cnx->lock, flags);
-       /* Receiving caps may mean the other end has just come up, so
-        * we need to reload the instance ID of the far end */
-       cnx->dst_inst =
-               HvCallEvent_getTargetLpInstanceId(cnx->remote_lp,
-                                                 HvLpEvent_Type_VirtualLan);
-
-       if (cnx->state & VETH_STATE_GOTCAPS) {
-               veth_error("Received a second capabilities from LPAR %d.\n",
-                          cnx->remote_lp);
-               event->base_event.xRc = HvLpEvent_Rc_BufferNotAvailable;
-               HvCallEvent_ackLpEvent((struct HvLpEvent *) event);
-       } else {
-               memcpy(&cnx->cap_event, event, sizeof(cnx->cap_event));
-               cnx->state |= VETH_STATE_GOTCAPS;
-               veth_kick_statemachine(cnx);
-       }
-       spin_unlock_irqrestore(&cnx->lock, flags);
-}
-
-static void veth_take_cap_ack(struct veth_lpar_connection *cnx,
-                             struct veth_lpevent *event)
-{
-       unsigned long flags;
-
-       spin_lock_irqsave(&cnx->lock, flags);
-       if (cnx->state & VETH_STATE_GOTCAPACK) {
-               veth_error("Received a second capabilities ack from LPAR %d.\n",
-                          cnx->remote_lp);
-       } else {
-               memcpy(&cnx->cap_ack_event, event,
-                      sizeof(cnx->cap_ack_event));
-               cnx->state |= VETH_STATE_GOTCAPACK;
-               veth_kick_statemachine(cnx);
-       }
-       spin_unlock_irqrestore(&cnx->lock, flags);
-}
-
-static void veth_take_monitor_ack(struct veth_lpar_connection *cnx,
-                                 struct veth_lpevent *event)
-{
-       unsigned long flags;
-
-       spin_lock_irqsave(&cnx->lock, flags);
-       veth_debug("cnx %d: lost connection.\n", cnx->remote_lp);
-
-       /* Avoid kicking the statemachine once we're shutdown.
-        * It's unnecessary and it could break veth_stop_connection(). */
-
-       if (! (cnx->state & VETH_STATE_SHUTDOWN)) {
-               cnx->state |= VETH_STATE_RESET;
-               veth_kick_statemachine(cnx);
-       }
-       spin_unlock_irqrestore(&cnx->lock, flags);
-}
-
-static void veth_handle_ack(struct veth_lpevent *event)
-{
-       HvLpIndex rlp = event->base_event.xTargetLp;
-       struct veth_lpar_connection *cnx = veth_cnx[rlp];
-
-       BUG_ON(! cnx);
-
-       switch (event->base_event.xSubtype) {
-       case VETH_EVENT_CAP:
-               veth_take_cap_ack(cnx, event);
-               break;
-       case VETH_EVENT_MONITOR:
-               veth_take_monitor_ack(cnx, event);
-               break;
-       default:
-               veth_error("Unknown ack type %d from LPAR %d.\n",
-                               event->base_event.xSubtype, rlp);
-       }
-}
-
-static void veth_handle_int(struct veth_lpevent *event)
-{
-       HvLpIndex rlp = event->base_event.xSourceLp;
-       struct veth_lpar_connection *cnx = veth_cnx[rlp];
-       unsigned long flags;
-       int i, acked = 0;
-
-       BUG_ON(! cnx);
-
-       switch (event->base_event.xSubtype) {
-       case VETH_EVENT_CAP:
-               veth_take_cap(cnx, event);
-               break;
-       case VETH_EVENT_MONITOR:
-               /* do nothing... this'll hang out here til we're dead,
-                * and the hypervisor will return it for us. */
-               break;
-       case VETH_EVENT_FRAMES_ACK:
-               spin_lock_irqsave(&cnx->lock, flags);
-
-               for (i = 0; i < VETH_MAX_ACKS_PER_MSG; ++i) {
-                       u16 msgnum = event->u.frames_ack_data.token[i];
-
-                       if (msgnum < VETH_NUMBUFFERS) {
-                               veth_recycle_msg(cnx, cnx->msgs + msgnum);
-                               cnx->outstanding_tx--;
-                               acked++;
-                       }
-               }
-
-               if (acked > 0) {
-                       cnx->last_contact = jiffies;
-                       veth_wake_queues(cnx);
-               }
-
-               spin_unlock_irqrestore(&cnx->lock, flags);
-               break;
-       case VETH_EVENT_FRAMES:
-               veth_receive(cnx, event);
-               break;
-       default:
-               veth_error("Unknown interrupt type %d from LPAR %d.\n",
-                               event->base_event.xSubtype, rlp);
-       }
-}
-
-static void veth_handle_event(struct HvLpEvent *event)
-{
-       struct veth_lpevent *veth_event = (struct veth_lpevent *)event;
-
-       if (hvlpevent_is_ack(event))
-               veth_handle_ack(veth_event);
-       else
-               veth_handle_int(veth_event);
-}
-
-static int veth_process_caps(struct veth_lpar_connection *cnx)
-{
-       struct veth_cap_data *remote_caps = &cnx->remote_caps;
-       int num_acks_needed;
-
-       /* Convert timer to jiffies */
-       cnx->ack_timeout = remote_caps->ack_timeout * HZ / 1000000;
-
-       if ( (remote_caps->num_buffers == 0) ||
-            (remote_caps->ack_threshold > VETH_MAX_ACKS_PER_MSG) ||
-            (remote_caps->ack_threshold == 0) ||
-            (cnx->ack_timeout == 0) ) {
-               veth_error("Received incompatible capabilities from LPAR %d.\n",
-                               cnx->remote_lp);
-               return HvLpEvent_Rc_InvalidSubtypeData;
-       }
-
-       num_acks_needed = (remote_caps->num_buffers
-                          / remote_caps->ack_threshold) + 1;
-
-       /* FIXME: locking on num_ack_events? */
-       if (cnx->num_ack_events < num_acks_needed) {
-               int num;
-
-               num = veth_allocate_events(cnx->remote_lp,
-                                          num_acks_needed-cnx->num_ack_events);
-               if (num > 0)
-                       cnx->num_ack_events += num;
-
-               if (cnx->num_ack_events < num_acks_needed) {
-                       veth_error("Couldn't allocate enough ack events "
-                                       "for LPAR %d.\n", cnx->remote_lp);
-
-                       return HvLpEvent_Rc_BufferNotAvailable;
-               }
-       }
-
-
-       return HvLpEvent_Rc_Good;
-}
-
-/* FIXME: The gotos here are a bit dubious */
-static void veth_statemachine(struct work_struct *work)
-{
-       struct veth_lpar_connection *cnx =
-               container_of(work, struct veth_lpar_connection,
-                            statemachine_wq.work);
-       int rlp = cnx->remote_lp;
-       int rc;
-
-       spin_lock_irq(&cnx->lock);
-
- restart:
-       if (cnx->state & VETH_STATE_RESET) {
-               if (cnx->state & VETH_STATE_OPEN)
-                       HvCallEvent_closeLpEventPath(cnx->remote_lp,
-                                                    HvLpEvent_Type_VirtualLan);
-
-               /*
-                * Reset ack data. This prevents the ack_timer actually
-                * doing anything, even if it runs one more time when
-                * we drop the lock below.
-                */
-               memset(&cnx->pending_acks, 0xff, sizeof (cnx->pending_acks));
-               cnx->num_pending_acks = 0;
-
-               cnx->state &= ~(VETH_STATE_RESET | VETH_STATE_SENTMON
-                               | VETH_STATE_OPEN | VETH_STATE_SENTCAPS
-                               | VETH_STATE_GOTCAPACK | VETH_STATE_GOTCAPS
-                               | VETH_STATE_SENTCAPACK | VETH_STATE_READY);
-
-               /* Clean up any leftover messages */
-               if (cnx->msgs) {
-                       int i;
-                       for (i = 0; i < VETH_NUMBUFFERS; ++i)
-                               veth_recycle_msg(cnx, cnx->msgs + i);
-               }
-
-               cnx->outstanding_tx = 0;
-               veth_wake_queues(cnx);
-
-               /* Drop the lock so we can do stuff that might sleep or
-                * take other locks. */
-               spin_unlock_irq(&cnx->lock);
-
-               del_timer_sync(&cnx->ack_timer);
-               del_timer_sync(&cnx->reset_timer);
-
-               spin_lock_irq(&cnx->lock);
-
-               if (cnx->state & VETH_STATE_RESET)
-                       goto restart;
-
-               /* Hack, wait for the other end to reset itself. */
-               if (! (cnx->state & VETH_STATE_SHUTDOWN)) {
-                       schedule_delayed_work(&cnx->statemachine_wq, 5 * HZ);
-                       goto out;
-               }
-       }
-
-       if (cnx->state & VETH_STATE_SHUTDOWN)
-               /* It's all over, do nothing */
-               goto out;
-
-       if ( !(cnx->state & VETH_STATE_OPEN) ) {
-               if (! cnx->msgs || (cnx->num_events < (2 + VETH_NUMBUFFERS)) )
-                       goto cant_cope;
-
-               HvCallEvent_openLpEventPath(rlp, HvLpEvent_Type_VirtualLan);
-               cnx->src_inst =
-                       HvCallEvent_getSourceLpInstanceId(rlp,
-                                                         HvLpEvent_Type_VirtualLan);
-               cnx->dst_inst =
-                       HvCallEvent_getTargetLpInstanceId(rlp,
-                                                         HvLpEvent_Type_VirtualLan);
-               cnx->state |= VETH_STATE_OPEN;
-       }
-
-       if ( (cnx->state & VETH_STATE_OPEN) &&
-            !(cnx->state & VETH_STATE_SENTMON) ) {
-               rc = veth_signalevent(cnx, VETH_EVENT_MONITOR,
-                                     HvLpEvent_AckInd_DoAck,
-                                     HvLpEvent_AckType_DeferredAck,
-                                     0, 0, 0, 0, 0, 0);
-
-               if (rc == HvLpEvent_Rc_Good) {
-                       cnx->state |= VETH_STATE_SENTMON;
-               } else {
-                       if ( (rc != HvLpEvent_Rc_PartitionDead) &&
-                            (rc != HvLpEvent_Rc_PathClosed) )
-                               veth_error("Error sending monitor to LPAR %d, "
-                                               "rc = %d\n", rlp, rc);
-
-                       /* Oh well, hope we get a cap from the other
-                        * end and do better when that kicks us */
-                       goto out;
-               }
-       }
-
-       if ( (cnx->state & VETH_STATE_OPEN) &&
-            !(cnx->state & VETH_STATE_SENTCAPS)) {
-               u64 *rawcap = (u64 *)&cnx->local_caps;
-
-               rc = veth_signalevent(cnx, VETH_EVENT_CAP,
-                                     HvLpEvent_AckInd_DoAck,
-                                     HvLpEvent_AckType_ImmediateAck,
-                                     0, rawcap[0], rawcap[1], rawcap[2],
-                                     rawcap[3], rawcap[4]);
-
-               if (rc == HvLpEvent_Rc_Good) {
-                       cnx->state |= VETH_STATE_SENTCAPS;
-               } else {
-                       if ( (rc != HvLpEvent_Rc_PartitionDead) &&
-                            (rc != HvLpEvent_Rc_PathClosed) )
-                               veth_error("Error sending caps to LPAR %d, "
-                                               "rc = %d\n", rlp, rc);
-
-                       /* Oh well, hope we get a cap from the other
-                        * end and do better when that kicks us */
-                       goto out;
-               }
-       }
-
-       if ((cnx->state & VETH_STATE_GOTCAPS) &&
-           !(cnx->state & VETH_STATE_SENTCAPACK)) {
-               struct veth_cap_data *remote_caps = &cnx->remote_caps;
-
-               memcpy(remote_caps, &cnx->cap_event.u.caps_data,
-                      sizeof(*remote_caps));
-
-               spin_unlock_irq(&cnx->lock);
-               rc = veth_process_caps(cnx);
-               spin_lock_irq(&cnx->lock);
-
-               /* We dropped the lock, so recheck for anything which
-                * might mess us up */
-               if (cnx->state & (VETH_STATE_RESET|VETH_STATE_SHUTDOWN))
-                       goto restart;
-
-               cnx->cap_event.base_event.xRc = rc;
-               HvCallEvent_ackLpEvent((struct HvLpEvent *)&cnx->cap_event);
-               if (rc == HvLpEvent_Rc_Good)
-                       cnx->state |= VETH_STATE_SENTCAPACK;
-               else
-                       goto cant_cope;
-       }
-
-       if ((cnx->state & VETH_STATE_GOTCAPACK) &&
-           (cnx->state & VETH_STATE_GOTCAPS) &&
-           !(cnx->state & VETH_STATE_READY)) {
-               if (cnx->cap_ack_event.base_event.xRc == HvLpEvent_Rc_Good) {
-                       /* Start the ACK timer */
-                       cnx->ack_timer.expires = jiffies + cnx->ack_timeout;
-                       add_timer(&cnx->ack_timer);
-                       cnx->state |= VETH_STATE_READY;
-               } else {
-                       veth_error("Caps rejected by LPAR %d, rc = %d\n",
-                                       rlp, cnx->cap_ack_event.base_event.xRc);
-                       goto cant_cope;
-               }
-       }
-
- out:
-       spin_unlock_irq(&cnx->lock);
-       return;
-
- cant_cope:
-       /* FIXME: we get here if something happens we really can't
-        * cope with.  The link will never work once we get here, and
-        * all we can do is not lock the rest of the system up */
-       veth_error("Unrecoverable error on connection to LPAR %d, shutting down"
-                       " (state = 0x%04lx)\n", rlp, cnx->state);
-       cnx->state |= VETH_STATE_SHUTDOWN;
-       spin_unlock_irq(&cnx->lock);
-}
-
-static int veth_init_connection(u8 rlp)
-{
-       struct veth_lpar_connection *cnx;
-       struct veth_msg *msgs;
-       int i;
-
-       if ( (rlp == this_lp) ||
-            ! HvLpConfig_doLpsCommunicateOnVirtualLan(this_lp, rlp) )
-               return 0;
-
-       cnx = kzalloc(sizeof(*cnx), GFP_KERNEL);
-       if (! cnx)
-               return -ENOMEM;
-
-       cnx->remote_lp = rlp;
-       spin_lock_init(&cnx->lock);
-       INIT_DELAYED_WORK(&cnx->statemachine_wq, veth_statemachine);
-
-       init_timer(&cnx->ack_timer);
-       cnx->ack_timer.function = veth_timed_ack;
-       cnx->ack_timer.data = (unsigned long) cnx;
-
-       init_timer(&cnx->reset_timer);
-       cnx->reset_timer.function = veth_timed_reset;
-       cnx->reset_timer.data = (unsigned long) cnx;
-       cnx->reset_timeout = 5 * HZ * (VETH_ACKTIMEOUT / 1000000);
-
-       memset(&cnx->pending_acks, 0xff, sizeof (cnx->pending_acks));
-
-       veth_cnx[rlp] = cnx;
-
-       /* This gets us 1 reference, which is held on behalf of the driver
-        * infrastructure. It's released at module unload. */
-       kobject_init(&cnx->kobject, &veth_lpar_connection_ktype);
-
-       msgs = kcalloc(VETH_NUMBUFFERS, sizeof(struct veth_msg), GFP_KERNEL);
-       if (! msgs) {
-               veth_error("Can't allocate buffers for LPAR %d.\n", rlp);
-               return -ENOMEM;
-       }
-
-       cnx->msgs = msgs;
-
-       for (i = 0; i < VETH_NUMBUFFERS; i++) {
-               msgs[i].token = i;
-               veth_stack_push(cnx, msgs + i);
-       }
-
-       cnx->num_events = veth_allocate_events(rlp, 2 + VETH_NUMBUFFERS);
-
-       if (cnx->num_events < (2 + VETH_NUMBUFFERS)) {
-               veth_error("Can't allocate enough events for LPAR %d.\n", rlp);
-               return -ENOMEM;
-       }
-
-       cnx->local_caps.num_buffers = VETH_NUMBUFFERS;
-       cnx->local_caps.ack_threshold = ACK_THRESHOLD;
-       cnx->local_caps.ack_timeout = VETH_ACKTIMEOUT;
-
-       return 0;
-}
-
-static void veth_stop_connection(struct veth_lpar_connection *cnx)
-{
-       if (!cnx)
-               return;
-
-       spin_lock_irq(&cnx->lock);
-       cnx->state |= VETH_STATE_RESET | VETH_STATE_SHUTDOWN;
-       veth_kick_statemachine(cnx);
-       spin_unlock_irq(&cnx->lock);
-
-       /* ensure the statemachine runs now and waits for its completion */
-       flush_delayed_work_sync(&cnx->statemachine_wq);
-}
-
-static void veth_destroy_connection(struct veth_lpar_connection *cnx)
-{
-       if (!cnx)
-               return;
-
-       if (cnx->num_events > 0)
-               mf_deallocate_lp_events(cnx->remote_lp,
-                                     HvLpEvent_Type_VirtualLan,
-                                     cnx->num_events,
-                                     NULL, NULL);
-       if (cnx->num_ack_events > 0)
-               mf_deallocate_lp_events(cnx->remote_lp,
-                                     HvLpEvent_Type_VirtualLan,
-                                     cnx->num_ack_events,
-                                     NULL, NULL);
-
-       kfree(cnx->msgs);
-       veth_cnx[cnx->remote_lp] = NULL;
-       kfree(cnx);
-}
-
-static void veth_release_connection(struct kobject *kobj)
-{
-       struct veth_lpar_connection *cnx;
-       cnx = container_of(kobj, struct veth_lpar_connection, kobject);
-       veth_stop_connection(cnx);
-       veth_destroy_connection(cnx);
-}
-
-/*
- * net_device code
- */
-
-static int veth_open(struct net_device *dev)
-{
-       netif_start_queue(dev);
-       return 0;
-}
-
-static int veth_close(struct net_device *dev)
-{
-       netif_stop_queue(dev);
-       return 0;
-}
-
-static int veth_change_mtu(struct net_device *dev, int new_mtu)
-{
-       if ((new_mtu < 68) || (new_mtu > VETH_MAX_MTU))
-               return -EINVAL;
-       dev->mtu = new_mtu;
-       return 0;
-}
-
-static void veth_set_multicast_list(struct net_device *dev)
-{
-       struct veth_port *port = netdev_priv(dev);
-       unsigned long flags;
-
-       write_lock_irqsave(&port->mcast_gate, flags);
-
-       if ((dev->flags & IFF_PROMISC) || (dev->flags & IFF_ALLMULTI) ||
-                       (netdev_mc_count(dev) > VETH_MAX_MCAST)) {
-               port->promiscuous = 1;
-       } else {
-               struct netdev_hw_addr *ha;
-
-               port->promiscuous = 0;
-
-               /* Update table */
-               port->num_mcast = 0;
-
-               netdev_for_each_mc_addr(ha, dev) {
-                       u8 *addr = ha->addr;
-                       u64 xaddr = 0;
-
-                       memcpy(&xaddr, addr, ETH_ALEN);
-                       port->mcast_addr[port->num_mcast] = xaddr;
-                       port->num_mcast++;
-               }
-       }
-
-       write_unlock_irqrestore(&port->mcast_gate, flags);
-}
-
-static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
-{
-       strncpy(info->driver, DRV_NAME, sizeof(info->driver) - 1);
-       info->driver[sizeof(info->driver) - 1] = '\0';
-       strncpy(info->version, DRV_VERSION, sizeof(info->version) - 1);
-       info->version[sizeof(info->version) - 1] = '\0';
-}
-
-static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *ecmd)
-{
-       ecmd->supported = (SUPPORTED_1000baseT_Full
-                         | SUPPORTED_Autoneg | SUPPORTED_FIBRE);
-       ecmd->advertising = (SUPPORTED_1000baseT_Full
-                           | SUPPORTED_Autoneg | SUPPORTED_FIBRE);
-       ecmd->port = PORT_FIBRE;
-       ecmd->transceiver = XCVR_INTERNAL;
-       ecmd->phy_address = 0;
-       ecmd->speed = SPEED_1000;
-       ecmd->duplex = DUPLEX_FULL;
-       ecmd->autoneg = AUTONEG_ENABLE;
-       ecmd->maxtxpkt = 120;
-       ecmd->maxrxpkt = 120;
-       return 0;
-}
-
-static const struct ethtool_ops ops = {
-       .get_drvinfo = veth_get_drvinfo,
-       .get_settings = veth_get_settings,
-       .get_link = ethtool_op_get_link,
-};
-
-static const struct net_device_ops veth_netdev_ops = {
-       .ndo_open               = veth_open,
-       .ndo_stop               = veth_close,
-       .ndo_start_xmit         = veth_start_xmit,
-       .ndo_change_mtu         = veth_change_mtu,
-       .ndo_set_rx_mode        = veth_set_multicast_list,
-       .ndo_set_mac_address    = NULL,
-       .ndo_validate_addr      = eth_validate_addr,
-};
-
-static struct net_device *veth_probe_one(int vlan,
-               struct vio_dev *vio_dev)
-{
-       struct net_device *dev;
-       struct veth_port *port;
-       struct device *vdev = &vio_dev->dev;
-       int i, rc;
-       const unsigned char *mac_addr;
-
-       mac_addr = vio_get_attribute(vio_dev, "local-mac-address", NULL);
-       if (mac_addr == NULL)
-               mac_addr = vio_get_attribute(vio_dev, "mac-address", NULL);
-       if (mac_addr == NULL) {
-               veth_error("Unable to fetch MAC address from device tree.\n");
-               return NULL;
-       }
-
-       dev = alloc_etherdev(sizeof (struct veth_port));
-       if (!dev)
-               return NULL;
-
-       port = netdev_priv(dev);
-
-       spin_lock_init(&port->queue_lock);
-       rwlock_init(&port->mcast_gate);
-       port->stopped_map = 0;
-
-       for (i = 0; i < HVMAXARCHITECTEDLPS; i++) {
-               HvLpVirtualLanIndexMap map;
-
-               if (i == this_lp)
-                       continue;
-               map = HvLpConfig_getVirtualLanIndexMapForLp(i);
-               if (map & (0x8000 >> vlan))
-                       port->lpar_map |= (1 << i);
-       }
-       port->dev = vdev;
-
-       memcpy(dev->dev_addr, mac_addr, ETH_ALEN);
-
-       dev->mtu = VETH_MAX_MTU;
-
-       memcpy(&port->mac_addr, mac_addr, ETH_ALEN);
-
-       dev->netdev_ops = &veth_netdev_ops;
-       SET_ETHTOOL_OPS(dev, &ops);
-
-       SET_NETDEV_DEV(dev, vdev);
-
-       rc = register_netdev(dev);
-       if (rc != 0) {
-               veth_error("Failed registering net device for vlan%d.\n", vlan);
-               free_netdev(dev);
-               return NULL;
-       }
-
-       kobject_init(&port->kobject, &veth_port_ktype);
-       if (0 != kobject_add(&port->kobject, &dev->dev.kobj, "veth_port"))
-               veth_error("Failed adding port for %s to sysfs.\n", dev->name);
-
-       veth_info("%s attached to iSeries vlan %d (LPAR map = 0x%.4X)\n",
-                       dev->name, vlan, port->lpar_map);
-
-       return dev;
-}
-
-/*
- * Tx path
- */
-
-static int veth_transmit_to_one(struct sk_buff *skb, HvLpIndex rlp,
-                               struct net_device *dev)
-{
-       struct veth_lpar_connection *cnx = veth_cnx[rlp];
-       struct veth_port *port = netdev_priv(dev);
-       HvLpEvent_Rc rc;
-       struct veth_msg *msg = NULL;
-       unsigned long flags;
-
-       if (! cnx)
-               return 0;
-
-       spin_lock_irqsave(&cnx->lock, flags);
-
-       if (! (cnx->state & VETH_STATE_READY))
-               goto no_error;
-
-       if ((skb->len - ETH_HLEN) > VETH_MAX_MTU)
-               goto drop;
-
-       msg = veth_stack_pop(cnx);
-       if (! msg)
-               goto drop;
-
-       msg->in_use = 1;
-       msg->skb = skb_get(skb);
-
-       msg->data.addr[0] = dma_map_single(port->dev, skb->data,
-                               skb->len, DMA_TO_DEVICE);
-
-       if (dma_mapping_error(port->dev, msg->data.addr[0]))
-               goto recycle_and_drop;
-
-       msg->dev = port->dev;
-       msg->data.len[0] = skb->len;
-       msg->data.eofmask = 1 << VETH_EOF_SHIFT;
-
-       rc = veth_signaldata(cnx, VETH_EVENT_FRAMES, msg->token, &msg->data);
-
-       if (rc != HvLpEvent_Rc_Good)
-               goto recycle_and_drop;
-
-       /* If the timer's not already running, start it now. */
-       if (0 == cnx->outstanding_tx)
-               mod_timer(&cnx->reset_timer, jiffies + cnx->reset_timeout);
-
-       cnx->last_contact = jiffies;
-       cnx->outstanding_tx++;
-
-       if (veth_stack_is_empty(cnx))
-               veth_stop_queues(cnx);
-
- no_error:
-       spin_unlock_irqrestore(&cnx->lock, flags);
-       return 0;
-
- recycle_and_drop:
-       veth_recycle_msg(cnx, msg);
- drop:
-       spin_unlock_irqrestore(&cnx->lock, flags);
-       return 1;
-}
-
-static void veth_transmit_to_many(struct sk_buff *skb,
-                                         HvLpIndexMap lpmask,
-                                         struct net_device *dev)
-{
-       int i, success, error;
-
-       success = error = 0;
-
-       for (i = 0; i < HVMAXARCHITECTEDLPS; i++) {
-               if ((lpmask & (1 << i)) == 0)
-                       continue;
-
-               if (veth_transmit_to_one(skb, i, dev))
-                       error = 1;
-               else
-                       success = 1;
-       }
-
-       if (error)
-               dev->stats.tx_errors++;
-
-       if (success) {
-               dev->stats.tx_packets++;
-               dev->stats.tx_bytes += skb->len;
-       }
-}
-
-static int veth_start_xmit(struct sk_buff *skb, struct net_device *dev)
-{
-       unsigned char *frame = skb->data;
-       struct veth_port *port = netdev_priv(dev);
-       HvLpIndexMap lpmask;
-
-       if (is_unicast_ether_addr(frame)) {
-               /* unicast packet */
-               HvLpIndex rlp = frame[5];
-
-               if ( ! ((1 << rlp) & port->lpar_map) ) {
-                       dev_kfree_skb(skb);
-                       return NETDEV_TX_OK;
-               }
-
-               lpmask = 1 << rlp;
-       } else {
-               lpmask = port->lpar_map;
-       }
-
-       veth_transmit_to_many(skb, lpmask, dev);
-
-       dev_kfree_skb(skb);
-
-       return NETDEV_TX_OK;
-}
-
-/* You must hold the connection's lock when you call this function. */
-static void veth_recycle_msg(struct veth_lpar_connection *cnx,
-                            struct veth_msg *msg)
-{
-       u32 dma_address, dma_length;
-
-       if (msg->in_use) {
-               msg->in_use = 0;
-               dma_address = msg->data.addr[0];
-               dma_length = msg->data.len[0];
-
-               if (!dma_mapping_error(msg->dev, dma_address))
-                       dma_unmap_single(msg->dev, dma_address, dma_length,
-                                       DMA_TO_DEVICE);
-
-               if (msg->skb) {
-                       dev_kfree_skb_any(msg->skb);
-                       msg->skb = NULL;
-               }
-
-               memset(&msg->data, 0, sizeof(msg->data));
-               veth_stack_push(cnx, msg);
-       } else if (cnx->state & VETH_STATE_OPEN) {
-               veth_error("Non-pending frame (# %d) acked by LPAR %d.\n",
-                               cnx->remote_lp, msg->token);
-       }
-}
-
-static void veth_wake_queues(struct veth_lpar_connection *cnx)
-{
-       int i;
-
-       for (i = 0; i < HVMAXARCHITECTEDVIRTUALLANS; i++) {
-               struct net_device *dev = veth_dev[i];
-               struct veth_port *port;
-               unsigned long flags;
-
-               if (! dev)
-                       continue;
-
-               port = netdev_priv(dev);
-
-               if (! (port->lpar_map & (1<<cnx->remote_lp)))
-                       continue;
-
-               spin_lock_irqsave(&port->queue_lock, flags);
-
-               port->stopped_map &= ~(1 << cnx->remote_lp);
-
-               if (0 == port->stopped_map && netif_queue_stopped(dev)) {
-                       veth_debug("cnx %d: woke queue for %s.\n",
-                                       cnx->remote_lp, dev->name);
-                       netif_wake_queue(dev);
-               }
-               spin_unlock_irqrestore(&port->queue_lock, flags);
-       }
-}
-
-static void veth_stop_queues(struct veth_lpar_connection *cnx)
-{
-       int i;
-
-       for (i = 0; i < HVMAXARCHITECTEDVIRTUALLANS; i++) {
-               struct net_device *dev = veth_dev[i];
-               struct veth_port *port;
-
-               if (! dev)
-                       continue;
-
-               port = netdev_priv(dev);
-
-               /* If this cnx is not on the vlan for this port, continue */
-               if (! (port->lpar_map & (1 << cnx->remote_lp)))
-                       continue;
-
-               spin_lock(&port->queue_lock);
-
-               netif_stop_queue(dev);
-               port->stopped_map |= (1 << cnx->remote_lp);
-
-               veth_debug("cnx %d: stopped queue for %s, map = 0x%x.\n",
-                               cnx->remote_lp, dev->name, port->stopped_map);
-
-               spin_unlock(&port->queue_lock);
-       }
-}
-
-static void veth_timed_reset(unsigned long ptr)
-{
-       struct veth_lpar_connection *cnx = (struct veth_lpar_connection *)ptr;
-       unsigned long trigger_time, flags;
-
-       /* FIXME is it possible this fires after veth_stop_connection()?
-        * That would reschedule the statemachine for 5 seconds and probably
-        * execute it after the module's been unloaded. Hmm. */
-
-       spin_lock_irqsave(&cnx->lock, flags);
-
-       if (cnx->outstanding_tx > 0) {
-               trigger_time = cnx->last_contact + cnx->reset_timeout;
-
-               if (trigger_time < jiffies) {
-                       cnx->state |= VETH_STATE_RESET;
-                       veth_kick_statemachine(cnx);
-                       veth_error("%d packets not acked by LPAR %d within %d "
-                                       "seconds, resetting.\n",
-                                       cnx->outstanding_tx, cnx->remote_lp,
-                                       cnx->reset_timeout / HZ);
-               } else {
-                       /* Reschedule the timer */
-                       trigger_time = jiffies + cnx->reset_timeout;
-                       mod_timer(&cnx->reset_timer, trigger_time);
-               }
-       }
-
-       spin_unlock_irqrestore(&cnx->lock, flags);
-}
-
-/*
- * Rx path
- */
-
-static inline int veth_frame_wanted(struct veth_port *port, u64 mac_addr)
-{
-       int wanted = 0;
-       int i;
-       unsigned long flags;
-
-       if ( (mac_addr == port->mac_addr) || (mac_addr == 0xffffffffffff0000) )
-               return 1;
-
-       read_lock_irqsave(&port->mcast_gate, flags);
-
-       if (port->promiscuous) {
-               wanted = 1;
-               goto out;
-       }
-
-       for (i = 0; i < port->num_mcast; ++i) {
-               if (port->mcast_addr[i] == mac_addr) {
-                       wanted = 1;
-                       break;
-               }
-       }
-
- out:
-       read_unlock_irqrestore(&port->mcast_gate, flags);
-
-       return wanted;
-}
-
-struct dma_chunk {
-       u64 addr;
-       u64 size;
-};
-
-#define VETH_MAX_PAGES_PER_FRAME ( (VETH_MAX_MTU+PAGE_SIZE-2)/PAGE_SIZE + 1 )
-
-static inline void veth_build_dma_list(struct dma_chunk *list,
-                                      unsigned char *p, unsigned long length)
-{
-       unsigned long done;
-       int i = 1;
-
-       /* FIXME: skbs are contiguous in real addresses.  Do we
-        * really need to break it into PAGE_SIZE chunks, or can we do
-        * it just at the granularity of iSeries real->absolute
-        * mapping?  Indeed, given the way the allocator works, can we
-        * count on them being absolutely contiguous? */
-       list[0].addr = iseries_hv_addr(p);
-       list[0].size = min(length,
-                          PAGE_SIZE - ((unsigned long)p & ~PAGE_MASK));
-
-       done = list[0].size;
-       while (done < length) {
-               list[i].addr = iseries_hv_addr(p + done);
-               list[i].size = min(length-done, PAGE_SIZE);
-               done += list[i].size;
-               i++;
-       }
-}
-
-static void veth_flush_acks(struct veth_lpar_connection *cnx)
-{
-       HvLpEvent_Rc rc;
-
-       rc = veth_signaldata(cnx, VETH_EVENT_FRAMES_ACK,
-                            0, &cnx->pending_acks);
-
-       if (rc != HvLpEvent_Rc_Good)
-               veth_error("Failed acking frames from LPAR %d, rc = %d\n",
-                               cnx->remote_lp, (int)rc);
-
-       cnx->num_pending_acks = 0;
-       memset(&cnx->pending_acks, 0xff, sizeof(cnx->pending_acks));
-}
-
-static void veth_receive(struct veth_lpar_connection *cnx,
-                        struct veth_lpevent *event)
-{
-       struct veth_frames_data *senddata = &event->u.frames_data;
-       int startchunk = 0;
-       int nchunks;
-       unsigned long flags;
-       HvLpDma_Rc rc;
-
-       do {
-               u16 length = 0;
-               struct sk_buff *skb;
-               struct dma_chunk local_list[VETH_MAX_PAGES_PER_FRAME];
-               struct dma_chunk remote_list[VETH_MAX_FRAMES_PER_MSG];
-               u64 dest;
-               HvLpVirtualLanIndex vlan;
-               struct net_device *dev;
-               struct veth_port *port;
-
-               /* FIXME: do we need this? */
-               memset(local_list, 0, sizeof(local_list));
-               memset(remote_list, 0, sizeof(remote_list));
-
-               /* a 0 address marks the end of the valid entries */
-               if (senddata->addr[startchunk] == 0)
-                       break;
-
-               /* make sure that we have at least 1 EOF entry in the
-                * remaining entries */
-               if (! (senddata->eofmask >> (startchunk + VETH_EOF_SHIFT))) {
-                       veth_error("Missing EOF fragment in event "
-                                       "eofmask = 0x%x startchunk = %d\n",
-                                       (unsigned)senddata->eofmask,
-                                       startchunk);
-                       break;
-               }
-
-               /* build list of chunks in this frame */
-               nchunks = 0;
-               do {
-                       remote_list[nchunks].addr =
-                               (u64) senddata->addr[startchunk+nchunks] << 32;
-                       remote_list[nchunks].size =
-                               senddata->len[startchunk+nchunks];
-                       length += remote_list[nchunks].size;
-               } while (! (senddata->eofmask &
-                           (1 << (VETH_EOF_SHIFT + startchunk + nchunks++))));
-
-               /* length == total length of all chunks */
-               /* nchunks == # of chunks in this frame */
-
-               if ((length - ETH_HLEN) > VETH_MAX_MTU) {
-                       veth_error("Received oversize frame from LPAR %d "
-                                       "(length = %d)\n",
-                                       cnx->remote_lp, length);
-                       continue;
-               }
-
-               skb = alloc_skb(length, GFP_ATOMIC);
-               if (!skb)
-                       continue;
-
-               veth_build_dma_list(local_list, skb->data, length);
-
-               rc = HvCallEvent_dmaBufList(HvLpEvent_Type_VirtualLan,
-                                           event->base_event.xSourceLp,
-                                           HvLpDma_Direction_RemoteToLocal,
-                                           cnx->src_inst,
-                                           cnx->dst_inst,
-                                           HvLpDma_AddressType_RealAddress,
-                                           HvLpDma_AddressType_TceIndex,
-                                           iseries_hv_addr(&local_list),
-                                           iseries_hv_addr(&remote_list),
-                                           length);
-               if (rc != HvLpDma_Rc_Good) {
-                       dev_kfree_skb_irq(skb);
-                       continue;
-               }
-
-               vlan = skb->data[9];
-               dev = veth_dev[vlan];
-               if (! dev) {
-                       /*
-                        * Some earlier versions of the driver sent
-                        * broadcasts down all connections, even to lpars
-                        * that weren't on the relevant vlan. So ignore
-                        * packets belonging to a vlan we're not on.
-                        * We can also be here if we receive packets while
-                        * the driver is going down, because then dev is NULL.
-                        */
-                       dev_kfree_skb_irq(skb);
-                       continue;
-               }
-
-               port = netdev_priv(dev);
-               dest = *((u64 *) skb->data) & 0xFFFFFFFFFFFF0000;
-
-               if ((vlan > HVMAXARCHITECTEDVIRTUALLANS) || !port) {
-                       dev_kfree_skb_irq(skb);
-                       continue;
-               }
-               if (! veth_frame_wanted(port, dest)) {
-                       dev_kfree_skb_irq(skb);
-                       continue;
-               }
-
-               skb_put(skb, length);
-               skb->protocol = eth_type_trans(skb, dev);
-               skb_checksum_none_assert(skb);
-               netif_rx(skb);  /* send it up */
-               dev->stats.rx_packets++;
-               dev->stats.rx_bytes += length;
-       } while (startchunk += nchunks, startchunk < VETH_MAX_FRAMES_PER_MSG);
-
-       /* Ack it */
-       spin_lock_irqsave(&cnx->lock, flags);
-       BUG_ON(cnx->num_pending_acks > VETH_MAX_ACKS_PER_MSG);
-
-       cnx->pending_acks[cnx->num_pending_acks++] =
-               event->base_event.xCorrelationToken;
-
-       if ( (cnx->num_pending_acks >= cnx->remote_caps.ack_threshold) ||
-            (cnx->num_pending_acks >= VETH_MAX_ACKS_PER_MSG) )
-               veth_flush_acks(cnx);
-
-       spin_unlock_irqrestore(&cnx->lock, flags);
-}
-
-static void veth_timed_ack(unsigned long ptr)
-{
-       struct veth_lpar_connection *cnx = (struct veth_lpar_connection *) ptr;
-       unsigned long flags;
-
-       /* Ack all the events */
-       spin_lock_irqsave(&cnx->lock, flags);
-       if (cnx->num_pending_acks > 0)
-               veth_flush_acks(cnx);
-
-       /* Reschedule the timer */
-       cnx->ack_timer.expires = jiffies + cnx->ack_timeout;
-       add_timer(&cnx->ack_timer);
-       spin_unlock_irqrestore(&cnx->lock, flags);
-}
-
-static int veth_remove(struct vio_dev *vdev)
-{
-       struct veth_lpar_connection *cnx;
-       struct net_device *dev;
-       struct veth_port *port;
-       int i;
-
-       dev = veth_dev[vdev->unit_address];
-
-       if (! dev)
-               return 0;
-
-       port = netdev_priv(dev);
-
-       for (i = 0; i < HVMAXARCHITECTEDLPS; i++) {
-               cnx = veth_cnx[i];
-
-               if (cnx && (port->lpar_map & (1 << i))) {
-                       /* Drop our reference to connections on our VLAN */
-                       kobject_put(&cnx->kobject);
-               }
-       }
-
-       veth_dev[vdev->unit_address] = NULL;
-       kobject_del(&port->kobject);
-       kobject_put(&port->kobject);
-       unregister_netdev(dev);
-       free_netdev(dev);
-
-       return 0;
-}
-
-static int veth_probe(struct vio_dev *vdev, const struct vio_device_id *id)
-{
-       int i = vdev->unit_address;
-       struct net_device *dev;
-       struct veth_port *port;
-
-       dev = veth_probe_one(i, vdev);
-       if (dev == NULL) {
-               veth_remove(vdev);
-               return 1;
-       }
-       veth_dev[i] = dev;
-
-       port = netdev_priv(dev);
-
-       /* Start the state machine on each connection on this vlan. If we're
-        * the first dev to do so this will commence link negotiation */
-       for (i = 0; i < HVMAXARCHITECTEDLPS; i++) {
-               struct veth_lpar_connection *cnx;
-
-               if (! (port->lpar_map & (1 << i)))
-                       continue;
-
-               cnx = veth_cnx[i];
-               if (!cnx)
-                       continue;
-
-               kobject_get(&cnx->kobject);
-               veth_kick_statemachine(cnx);
-       }
-
-       return 0;
-}
-
-/**
- * veth_device_table: Used by vio.c to match devices that we
- * support.
- */
-static struct vio_device_id veth_device_table[] __devinitdata = {
-       { "network", "IBM,iSeries-l-lan" },
-       { "", "" }
-};
-MODULE_DEVICE_TABLE(vio, veth_device_table);
-
-static struct vio_driver veth_driver = {
-       .id_table = veth_device_table,
-       .probe = veth_probe,
-       .remove = veth_remove,
-       .driver = {
-               .name = DRV_NAME,
-               .owner = THIS_MODULE,
-       }
-};
-
-/*
- * Module initialization/cleanup
- */
-
-static void __exit veth_module_cleanup(void)
-{
-       int i;
-       struct veth_lpar_connection *cnx;
-
-       /* Disconnect our "irq" to stop events coming from the Hypervisor. */
-       HvLpEvent_unregisterHandler(HvLpEvent_Type_VirtualLan);
-
-       for (i = 0; i < HVMAXARCHITECTEDLPS; ++i) {
-               cnx = veth_cnx[i];
-
-               if (!cnx)
-                       continue;
-
-               /* Cancel work queued from Hypervisor callbacks */
-               cancel_delayed_work_sync(&cnx->statemachine_wq);
-               /* Remove the connection from sysfs */
-               kobject_del(&cnx->kobject);
-               /* Drop the driver's reference to the connection */
-               kobject_put(&cnx->kobject);
-       }
-
-       /* Unregister the driver, which will close all the netdevs and stop
-        * the connections when they're no longer referenced. */
-       vio_unregister_driver(&veth_driver);
-}
-module_exit(veth_module_cleanup);
-
-static int __init veth_module_init(void)
-{
-       int i;
-       int rc;
-
-       if (!firmware_has_feature(FW_FEATURE_ISERIES))
-               return -ENODEV;
-
-       this_lp = HvLpConfig_getLpIndex_outline();
-
-       for (i = 0; i < HVMAXARCHITECTEDLPS; ++i) {
-               rc = veth_init_connection(i);
-               if (rc != 0)
-                       goto error;
-       }
-
-       HvLpEvent_registerHandler(HvLpEvent_Type_VirtualLan,
-                                 &veth_handle_event);
-
-       rc = vio_register_driver(&veth_driver);
-       if (rc != 0)
-               goto error;
-
-       for (i = 0; i < HVMAXARCHITECTEDLPS; ++i) {
-               struct kobject *kobj;
-
-               if (!veth_cnx[i])
-                       continue;
-
-               kobj = &veth_cnx[i]->kobject;
-               /* If the add failes, complain but otherwise continue */
-               if (0 != driver_add_kobj(&veth_driver.driver, kobj,
-                                       "cnx%.2d", veth_cnx[i]->remote_lp))
-                       veth_error("cnx %d: Failed adding to sysfs.\n", i);
-       }
-
-       return 0;
-
-error:
-       for (i = 0; i < HVMAXARCHITECTEDLPS; ++i) {
-               veth_destroy_connection(veth_cnx[i]);
-       }
-
-       return rc;
-}
-module_init(veth_module_init);