]> www.pilppa.org Git - linux-2.6-omap-h63xx.git/blobdiff - drivers/scsi/fcoe/fcoe.c
Merge branch 'omap-pool'
[linux-2.6-omap-h63xx.git] / drivers / scsi / fcoe / fcoe.c
diff --git a/drivers/scsi/fcoe/fcoe.c b/drivers/scsi/fcoe/fcoe.c
new file mode 100644 (file)
index 0000000..94e1e31
--- /dev/null
@@ -0,0 +1,1878 @@
+/*
+ * Copyright(c) 2007 - 2008 Intel Corporation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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.,
+ * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Maintained at www.Open-FCoE.org
+ */
+
+#include <linux/module.h>
+#include <linux/version.h>
+#include <linux/spinlock.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/ethtool.h>
+#include <linux/if_ether.h>
+#include <linux/if_vlan.h>
+#include <linux/crc32.h>
+#include <linux/cpu.h>
+#include <linux/fs.h>
+#include <linux/sysfs.h>
+#include <linux/ctype.h>
+#include <scsi/scsi_tcq.h>
+#include <scsi/scsicam.h>
+#include <scsi/scsi_transport.h>
+#include <scsi/scsi_transport_fc.h>
+#include <net/rtnetlink.h>
+
+#include <scsi/fc/fc_encaps.h>
+#include <scsi/fc/fc_fip.h>
+
+#include <scsi/libfc.h>
+#include <scsi/fc_frame.h>
+#include <scsi/libfcoe.h>
+
+#include "fcoe.h"
+
+static int debug_fcoe;
+
+MODULE_AUTHOR("Open-FCoE.org");
+MODULE_DESCRIPTION("FCoE");
+MODULE_LICENSE("GPL v2");
+
+/* fcoe host list */
+LIST_HEAD(fcoe_hostlist);
+DEFINE_RWLOCK(fcoe_hostlist_lock);
+DEFINE_TIMER(fcoe_timer, NULL, 0, 0);
+DEFINE_PER_CPU(struct fcoe_percpu_s, fcoe_percpu);
+
+/* Function Prototyes */
+static int fcoe_reset(struct Scsi_Host *shost);
+static int fcoe_xmit(struct fc_lport *, struct fc_frame *);
+static int fcoe_rcv(struct sk_buff *, struct net_device *,
+                   struct packet_type *, struct net_device *);
+static int fcoe_percpu_receive_thread(void *arg);
+static void fcoe_clean_pending_queue(struct fc_lport *lp);
+static void fcoe_percpu_clean(struct fc_lport *lp);
+static int fcoe_link_ok(struct fc_lport *lp);
+
+static struct fc_lport *fcoe_hostlist_lookup(const struct net_device *);
+static int fcoe_hostlist_add(const struct fc_lport *);
+static int fcoe_hostlist_remove(const struct fc_lport *);
+
+static int fcoe_check_wait_queue(struct fc_lport *);
+static int fcoe_device_notification(struct notifier_block *, ulong, void *);
+static void fcoe_dev_setup(void);
+static void fcoe_dev_cleanup(void);
+
+/* notification function from net device */
+static struct notifier_block fcoe_notifier = {
+       .notifier_call = fcoe_device_notification,
+};
+
+static struct scsi_transport_template *scsi_transport_fcoe_sw;
+
+struct fc_function_template fcoe_transport_function = {
+       .show_host_node_name = 1,
+       .show_host_port_name = 1,
+       .show_host_supported_classes = 1,
+       .show_host_supported_fc4s = 1,
+       .show_host_active_fc4s = 1,
+       .show_host_maxframe_size = 1,
+
+       .show_host_port_id = 1,
+       .show_host_supported_speeds = 1,
+       .get_host_speed = fc_get_host_speed,
+       .show_host_speed = 1,
+       .show_host_port_type = 1,
+       .get_host_port_state = fc_get_host_port_state,
+       .show_host_port_state = 1,
+       .show_host_symbolic_name = 1,
+
+       .dd_fcrport_size = sizeof(struct fc_rport_libfc_priv),
+       .show_rport_maxframe_size = 1,
+       .show_rport_supported_classes = 1,
+
+       .show_host_fabric_name = 1,
+       .show_starget_node_name = 1,
+       .show_starget_port_name = 1,
+       .show_starget_port_id = 1,
+       .set_rport_dev_loss_tmo = fc_set_rport_loss_tmo,
+       .show_rport_dev_loss_tmo = 1,
+       .get_fc_host_stats = fc_get_host_stats,
+       .issue_fc_host_lip = fcoe_reset,
+
+       .terminate_rport_io = fc_rport_terminate_io,
+};
+
+static struct scsi_host_template fcoe_shost_template = {
+       .module = THIS_MODULE,
+       .name = "FCoE Driver",
+       .proc_name = FCOE_NAME,
+       .queuecommand = fc_queuecommand,
+       .eh_abort_handler = fc_eh_abort,
+       .eh_device_reset_handler = fc_eh_device_reset,
+       .eh_host_reset_handler = fc_eh_host_reset,
+       .slave_alloc = fc_slave_alloc,
+       .change_queue_depth = fc_change_queue_depth,
+       .change_queue_type = fc_change_queue_type,
+       .this_id = -1,
+       .cmd_per_lun = 32,
+       .can_queue = FCOE_MAX_OUTSTANDING_COMMANDS,
+       .use_clustering = ENABLE_CLUSTERING,
+       .sg_tablesize = SG_ALL,
+       .max_sectors = 0xffff,
+};
+
+/**
+ * fcoe_lport_config() - sets up the fc_lport
+ * @lp: ptr to the fc_lport
+ * @shost: ptr to the parent scsi host
+ *
+ * Returns: 0 for success
+ */
+static int fcoe_lport_config(struct fc_lport *lp)
+{
+       lp->link_up = 0;
+       lp->qfull = 0;
+       lp->max_retry_count = 3;
+       lp->e_d_tov = 2 * 1000; /* FC-FS default */
+       lp->r_a_tov = 2 * 2 * 1000;
+       lp->service_params = (FCP_SPPF_INIT_FCN | FCP_SPPF_RD_XRDY_DIS |
+                             FCP_SPPF_RETRY | FCP_SPPF_CONF_COMPL);
+
+       fc_lport_init_stats(lp);
+
+       /* lport fc_lport related configuration */
+       fc_lport_config(lp);
+
+       /* offload related configuration */
+       lp->crc_offload = 0;
+       lp->seq_offload = 0;
+       lp->lro_enabled = 0;
+       lp->lro_xid = 0;
+       lp->lso_max = 0;
+
+       return 0;
+}
+
+/**
+ * fcoe_netdev_config() - Set up netdev for SW FCoE
+ * @lp : ptr to the fc_lport
+ * @netdev : ptr to the associated netdevice struct
+ *
+ * Must be called after fcoe_lport_config() as it will use lport mutex
+ *
+ * Returns : 0 for success
+ */
+static int fcoe_netdev_config(struct fc_lport *lp, struct net_device *netdev)
+{
+       u32 mfs;
+       u64 wwnn, wwpn;
+       struct fcoe_softc *fc;
+       u8 flogi_maddr[ETH_ALEN];
+
+       /* Setup lport private data to point to fcoe softc */
+       fc = lport_priv(lp);
+       fc->ctlr.lp = lp;
+       fc->real_dev = netdev;
+       fc->phys_dev = netdev;
+
+       /* Require support for get_pauseparam ethtool op. */
+       if (netdev->priv_flags & IFF_802_1Q_VLAN)
+               fc->phys_dev = vlan_dev_real_dev(netdev);
+
+       /* Do not support for bonding device */
+       if ((fc->real_dev->priv_flags & IFF_MASTER_ALB) ||
+           (fc->real_dev->priv_flags & IFF_SLAVE_INACTIVE) ||
+           (fc->real_dev->priv_flags & IFF_MASTER_8023AD)) {
+               return -EOPNOTSUPP;
+       }
+
+       /*
+        * Determine max frame size based on underlying device and optional
+        * user-configured limit.  If the MFS is too low, fcoe_link_ok()
+        * will return 0, so do this first.
+        */
+       mfs = fc->real_dev->mtu - (sizeof(struct fcoe_hdr) +
+                                  sizeof(struct fcoe_crc_eof));
+       if (fc_set_mfs(lp, mfs))
+               return -EINVAL;
+
+       /* offload features support */
+       if (fc->real_dev->features & NETIF_F_SG)
+               lp->sg_supp = 1;
+
+#ifdef NETIF_F_FCOE_CRC
+       if (netdev->features & NETIF_F_FCOE_CRC) {
+               lp->crc_offload = 1;
+               printk(KERN_DEBUG "fcoe:%s supports FCCRC offload\n",
+                      netdev->name);
+       }
+#endif
+#ifdef NETIF_F_FSO
+       if (netdev->features & NETIF_F_FSO) {
+               lp->seq_offload = 1;
+               lp->lso_max = netdev->gso_max_size;
+               printk(KERN_DEBUG "fcoe:%s supports LSO for max len 0x%x\n",
+                      netdev->name, lp->lso_max);
+       }
+#endif
+       if (netdev->fcoe_ddp_xid) {
+               lp->lro_enabled = 1;
+               lp->lro_xid = netdev->fcoe_ddp_xid;
+               printk(KERN_DEBUG "fcoe:%s supports LRO for max xid 0x%x\n",
+                      netdev->name, lp->lro_xid);
+       }
+       skb_queue_head_init(&fc->fcoe_pending_queue);
+       fc->fcoe_pending_queue_active = 0;
+
+       /* setup Source Mac Address */
+       memcpy(fc->ctlr.ctl_src_addr, fc->real_dev->dev_addr,
+              fc->real_dev->addr_len);
+
+       wwnn = fcoe_wwn_from_mac(fc->real_dev->dev_addr, 1, 0);
+       fc_set_wwnn(lp, wwnn);
+       /* XXX - 3rd arg needs to be vlan id */
+       wwpn = fcoe_wwn_from_mac(fc->real_dev->dev_addr, 2, 0);
+       fc_set_wwpn(lp, wwpn);
+
+       /*
+        * Add FCoE MAC address as second unicast MAC address
+        * or enter promiscuous mode if not capable of listening
+        * for multiple unicast MACs.
+        */
+       rtnl_lock();
+       memcpy(flogi_maddr, (u8[6]) FC_FCOE_FLOGI_MAC, ETH_ALEN);
+       dev_unicast_add(fc->real_dev, flogi_maddr, ETH_ALEN);
+       rtnl_unlock();
+
+       /*
+        * setup the receive function from ethernet driver
+        * on the ethertype for the given device
+        */
+       fc->fcoe_packet_type.func = fcoe_rcv;
+       fc->fcoe_packet_type.type = __constant_htons(ETH_P_FCOE);
+       fc->fcoe_packet_type.dev = fc->real_dev;
+       dev_add_pack(&fc->fcoe_packet_type);
+
+       return 0;
+}
+
+/**
+ * fcoe_shost_config() - Sets up fc_lport->host
+ * @lp : ptr to the fc_lport
+ * @shost : ptr to the associated scsi host
+ * @dev : device associated to scsi host
+ *
+ * Must be called after fcoe_lport_config() and fcoe_netdev_config()
+ *
+ * Returns : 0 for success
+ */
+static int fcoe_shost_config(struct fc_lport *lp, struct Scsi_Host *shost,
+                               struct device *dev)
+{
+       int rc = 0;
+
+       /* lport scsi host config */
+       lp->host = shost;
+
+       lp->host->max_lun = FCOE_MAX_LUN;
+       lp->host->max_id = FCOE_MAX_FCP_TARGET;
+       lp->host->max_channel = 0;
+       lp->host->transportt = scsi_transport_fcoe_sw;
+
+       /* add the new host to the SCSI-ml */
+       rc = scsi_add_host(lp->host, dev);
+       if (rc) {
+               FC_DBG("fcoe_shost_config:error on scsi_add_host\n");
+               return rc;
+       }
+       sprintf(fc_host_symbolic_name(lp->host), "%s v%s over %s",
+               FCOE_NAME, FCOE_VERSION,
+               fcoe_netdev(lp)->name);
+
+       return 0;
+}
+
+/**
+ * fcoe_em_config() - allocates em for this lport
+ * @lp: the port that em is to allocated for
+ *
+ * Returns : 0 on success
+ */
+static inline int fcoe_em_config(struct fc_lport *lp)
+{
+       BUG_ON(lp->emp);
+
+       lp->emp = fc_exch_mgr_alloc(lp, FC_CLASS_3,
+                                   FCOE_MIN_XID, FCOE_MAX_XID);
+       if (!lp->emp)
+               return -ENOMEM;
+
+       return 0;
+}
+
+/**
+ * fcoe_if_destroy() - FCoE software HBA tear-down function
+ * @netdev: ptr to the associated net_device
+ *
+ * Returns: 0 if link is OK for use by FCoE.
+ */
+static int fcoe_if_destroy(struct net_device *netdev)
+{
+       struct fc_lport *lp = NULL;
+       struct fcoe_softc *fc;
+       u8 flogi_maddr[ETH_ALEN];
+
+       BUG_ON(!netdev);
+
+       printk(KERN_DEBUG "fcoe_if_destroy:interface on %s\n",
+              netdev->name);
+
+       lp = fcoe_hostlist_lookup(netdev);
+       if (!lp)
+               return -ENODEV;
+
+       fc = lport_priv(lp);
+
+       /* Logout of the fabric */
+       fc_fabric_logoff(lp);
+
+       /* Remove the instance from fcoe's list */
+       fcoe_hostlist_remove(lp);
+
+       /* Don't listen for Ethernet packets anymore */
+       dev_remove_pack(&fc->fcoe_packet_type);
+       dev_remove_pack(&fc->fip_packet_type);
+       fcoe_ctlr_destroy(&fc->ctlr);
+
+       /* Cleanup the fc_lport */
+       fc_lport_destroy(lp);
+       fc_fcp_destroy(lp);
+
+       /* Detach from the scsi-ml */
+       fc_remove_host(lp->host);
+       scsi_remove_host(lp->host);
+
+       /* There are no more rports or I/O, free the EM */
+       if (lp->emp)
+               fc_exch_mgr_free(lp->emp);
+
+       /* Delete secondary MAC addresses */
+       rtnl_lock();
+       memcpy(flogi_maddr, (u8[6]) FC_FCOE_FLOGI_MAC, ETH_ALEN);
+       dev_unicast_delete(fc->real_dev, flogi_maddr, ETH_ALEN);
+       if (!is_zero_ether_addr(fc->ctlr.data_src_addr))
+               dev_unicast_delete(fc->real_dev,
+                                  fc->ctlr.data_src_addr, ETH_ALEN);
+       dev_mc_delete(fc->real_dev, FIP_ALL_ENODE_MACS, ETH_ALEN, 0);
+       rtnl_unlock();
+
+       /* Free the per-CPU revieve threads */
+       fcoe_percpu_clean(lp);
+
+       /* Free existing skbs */
+       fcoe_clean_pending_queue(lp);
+
+       /* Free memory used by statistical counters */
+       fc_lport_free_stats(lp);
+
+       /* Release the net_device and Scsi_Host */
+       dev_put(fc->real_dev);
+       scsi_host_put(lp->host);
+
+       return 0;
+}
+
+/*
+ * fcoe_ddp_setup - calls LLD's ddp_setup through net_device
+ * @lp:        the corresponding fc_lport
+ * @xid: the exchange id for this ddp transfer
+ * @sgl: the scatterlist describing this transfer
+ * @sgc: number of sg items
+ *
+ * Returns : 0 no ddp
+ */
+static int fcoe_ddp_setup(struct fc_lport *lp, u16 xid,
+                            struct scatterlist *sgl, unsigned int sgc)
+{
+       struct net_device *n = fcoe_netdev(lp);
+
+       if (n->netdev_ops && n->netdev_ops->ndo_fcoe_ddp_setup)
+               return n->netdev_ops->ndo_fcoe_ddp_setup(n, xid, sgl, sgc);
+
+       return 0;
+}
+
+/*
+ * fcoe_ddp_done - calls LLD's ddp_done through net_device
+ * @lp:        the corresponding fc_lport
+ * @xid: the exchange id for this ddp transfer
+ *
+ * Returns : the length of data that have been completed by ddp
+ */
+static int fcoe_ddp_done(struct fc_lport *lp, u16 xid)
+{
+       struct net_device *n = fcoe_netdev(lp);
+
+       if (n->netdev_ops && n->netdev_ops->ndo_fcoe_ddp_done)
+               return n->netdev_ops->ndo_fcoe_ddp_done(n, xid);
+       return 0;
+}
+
+static struct libfc_function_template fcoe_libfc_fcn_templ = {
+       .frame_send = fcoe_xmit,
+       .ddp_setup = fcoe_ddp_setup,
+       .ddp_done = fcoe_ddp_done,
+};
+
+/**
+ * fcoe_fip_recv - handle a received FIP frame.
+ * @skb: the receive skb
+ * @dev: associated &net_device
+ * @ptype: the &packet_type structure which was used to register this handler.
+ * @orig_dev: original receive &net_device, in case @dev is a bond.
+ *
+ * Returns: 0 for success
+ */
+static int fcoe_fip_recv(struct sk_buff *skb, struct net_device *dev,
+                        struct packet_type *ptype,
+                        struct net_device *orig_dev)
+{
+       struct fcoe_softc *fc;
+
+       fc = container_of(ptype, struct fcoe_softc, fip_packet_type);
+       fcoe_ctlr_recv(&fc->ctlr, skb);
+       return 0;
+}
+
+/**
+ * fcoe_fip_send() - send an Ethernet-encapsulated FIP frame.
+ * @fip: FCoE controller.
+ * @skb: FIP Packet.
+ */
+static void fcoe_fip_send(struct fcoe_ctlr *fip, struct sk_buff *skb)
+{
+       skb->dev = fcoe_from_ctlr(fip)->real_dev;
+       dev_queue_xmit(skb);
+}
+
+/**
+ * fcoe_update_src_mac() - Update Ethernet MAC filters.
+ * @fip: FCoE controller.
+ * @old: Unicast MAC address to delete if the MAC is non-zero.
+ * @new: Unicast MAC address to add.
+ *
+ * Remove any previously-set unicast MAC filter.
+ * Add secondary FCoE MAC address filter for our OUI.
+ */
+static void fcoe_update_src_mac(struct fcoe_ctlr *fip, u8 *old, u8 *new)
+{
+       struct fcoe_softc *fc;
+
+       fc = fcoe_from_ctlr(fip);
+       rtnl_lock();
+       if (!is_zero_ether_addr(old))
+               dev_unicast_delete(fc->real_dev, old, ETH_ALEN);
+       dev_unicast_add(fc->real_dev, new, ETH_ALEN);
+       rtnl_unlock();
+}
+
+/**
+ * fcoe_if_create() - this function creates the fcoe interface
+ * @netdev: pointer the associated netdevice
+ *
+ * Creates fc_lport struct and scsi_host for lport, configures lport
+ * and starts fabric login.
+ *
+ * Returns : 0 on success
+ */
+static int fcoe_if_create(struct net_device *netdev)
+{
+       int rc;
+       struct fc_lport *lp = NULL;
+       struct fcoe_softc *fc;
+       struct Scsi_Host *shost;
+
+       BUG_ON(!netdev);
+
+       printk(KERN_DEBUG "fcoe_if_create:interface on %s\n",
+              netdev->name);
+
+       lp = fcoe_hostlist_lookup(netdev);
+       if (lp)
+               return -EEXIST;
+
+       shost = libfc_host_alloc(&fcoe_shost_template,
+                                sizeof(struct fcoe_softc));
+       if (!shost) {
+               FC_DBG("Could not allocate host structure\n");
+               return -ENOMEM;
+       }
+       lp = shost_priv(shost);
+       fc = lport_priv(lp);
+
+       /* configure fc_lport, e.g., em */
+       rc = fcoe_lport_config(lp);
+       if (rc) {
+               FC_DBG("Could not configure lport\n");
+               goto out_host_put;
+       }
+
+       /* configure lport network properties */
+       rc = fcoe_netdev_config(lp, netdev);
+       if (rc) {
+               FC_DBG("Could not configure netdev for lport\n");
+               goto out_host_put;
+       }
+
+       /*
+        * Initialize FIP.
+        */
+       fcoe_ctlr_init(&fc->ctlr);
+       fc->ctlr.send = fcoe_fip_send;
+       fc->ctlr.update_mac = fcoe_update_src_mac;
+
+       fc->fip_packet_type.func = fcoe_fip_recv;
+       fc->fip_packet_type.type = htons(ETH_P_FIP);
+       fc->fip_packet_type.dev = fc->real_dev;
+       dev_add_pack(&fc->fip_packet_type);
+
+       /* configure lport scsi host properties */
+       rc = fcoe_shost_config(lp, shost, &netdev->dev);
+       if (rc) {
+               FC_DBG("Could not configure shost for lport\n");
+               goto out_host_put;
+       }
+
+       /* lport exch manager allocation */
+       rc = fcoe_em_config(lp);
+       if (rc) {
+               FC_DBG("Could not configure em for lport\n");
+               goto out_host_put;
+       }
+
+       /* Initialize the library */
+       rc = fcoe_libfc_config(lp, &fcoe_libfc_fcn_templ);
+       if (rc) {
+               FC_DBG("Could not configure libfc for lport!\n");
+               goto out_lp_destroy;
+       }
+
+       /* add to lports list */
+       fcoe_hostlist_add(lp);
+
+       lp->boot_time = jiffies;
+
+       fc_fabric_login(lp);
+
+       if (!fcoe_link_ok(lp))
+               fcoe_ctlr_link_up(&fc->ctlr);
+
+       dev_hold(netdev);
+
+       return rc;
+
+out_lp_destroy:
+       fc_exch_mgr_free(lp->emp); /* Free the EM */
+out_host_put:
+       scsi_host_put(lp->host);
+       return rc;
+}
+
+/**
+ * fcoe_if_init() - attach to scsi transport
+ *
+ * Returns : 0 on success
+ */
+static int __init fcoe_if_init(void)
+{
+       /* attach to scsi transport */
+       scsi_transport_fcoe_sw =
+               fc_attach_transport(&fcoe_transport_function);
+
+       if (!scsi_transport_fcoe_sw) {
+               printk(KERN_ERR "fcoe_init:fc_attach_transport() failed\n");
+               return -ENODEV;
+       }
+
+       return 0;
+}
+
+/**
+ * fcoe_if_exit() - detach from scsi transport
+ *
+ * Returns : 0 on success
+ */
+int __exit fcoe_if_exit(void)
+{
+       fc_release_transport(scsi_transport_fcoe_sw);
+       return 0;
+}
+
+/**
+ * fcoe_percpu_thread_create() - Create a receive thread for an online cpu
+ * @cpu: cpu index for the online cpu
+ */
+static void fcoe_percpu_thread_create(unsigned int cpu)
+{
+       struct fcoe_percpu_s *p;
+       struct task_struct *thread;
+
+       p = &per_cpu(fcoe_percpu, cpu);
+
+       thread = kthread_create(fcoe_percpu_receive_thread,
+                               (void *)p, "fcoethread/%d", cpu);
+
+       if (likely(!IS_ERR(p->thread))) {
+               kthread_bind(thread, cpu);
+               wake_up_process(thread);
+
+               spin_lock_bh(&p->fcoe_rx_list.lock);
+               p->thread = thread;
+               spin_unlock_bh(&p->fcoe_rx_list.lock);
+       }
+}
+
+/**
+ * fcoe_percpu_thread_destroy() - removes the rx thread for the given cpu
+ * @cpu: cpu index the rx thread is to be removed
+ *
+ * Destroys a per-CPU Rx thread. Any pending skbs are moved to the
+ * current CPU's Rx thread. If the thread being destroyed is bound to
+ * the CPU processing this context the skbs will be freed.
+ */
+static void fcoe_percpu_thread_destroy(unsigned int cpu)
+{
+       struct fcoe_percpu_s *p;
+       struct task_struct *thread;
+       struct page *crc_eof;
+       struct sk_buff *skb;
+#ifdef CONFIG_SMP
+       struct fcoe_percpu_s *p0;
+       unsigned targ_cpu = smp_processor_id();
+#endif /* CONFIG_SMP */
+
+       printk(KERN_DEBUG "fcoe: Destroying receive thread for CPU %d\n", cpu);
+
+       /* Prevent any new skbs from being queued for this CPU. */
+       p = &per_cpu(fcoe_percpu, cpu);
+       spin_lock_bh(&p->fcoe_rx_list.lock);
+       thread = p->thread;
+       p->thread = NULL;
+       crc_eof = p->crc_eof_page;
+       p->crc_eof_page = NULL;
+       p->crc_eof_offset = 0;
+       spin_unlock_bh(&p->fcoe_rx_list.lock);
+
+#ifdef CONFIG_SMP
+       /*
+        * Don't bother moving the skb's if this context is running
+        * on the same CPU that is having its thread destroyed. This
+        * can easily happen when the module is removed.
+        */
+       if (cpu != targ_cpu) {
+               p0 = &per_cpu(fcoe_percpu, targ_cpu);
+               spin_lock_bh(&p0->fcoe_rx_list.lock);
+               if (p0->thread) {
+                       FC_DBG("Moving frames from CPU %d to CPU %d\n",
+                              cpu, targ_cpu);
+
+                       while ((skb = __skb_dequeue(&p->fcoe_rx_list)) != NULL)
+                               __skb_queue_tail(&p0->fcoe_rx_list, skb);
+                       spin_unlock_bh(&p0->fcoe_rx_list.lock);
+               } else {
+                       /*
+                        * The targeted CPU is not initialized and cannot accept
+                        * new  skbs. Unlock the targeted CPU and drop the skbs
+                        * on the CPU that is going offline.
+                        */
+                       while ((skb = __skb_dequeue(&p->fcoe_rx_list)) != NULL)
+                               kfree_skb(skb);
+                       spin_unlock_bh(&p0->fcoe_rx_list.lock);
+               }
+       } else {
+               /*
+                * This scenario occurs when the module is being removed
+                * and all threads are being destroyed. skbs will continue
+                * to be shifted from the CPU thread that is being removed
+                * to the CPU thread associated with the CPU that is processing
+                * the module removal. Once there is only one CPU Rx thread it
+                * will reach this case and we will drop all skbs and later
+                * stop the thread.
+                */
+               spin_lock_bh(&p->fcoe_rx_list.lock);
+               while ((skb = __skb_dequeue(&p->fcoe_rx_list)) != NULL)
+                       kfree_skb(skb);
+               spin_unlock_bh(&p->fcoe_rx_list.lock);
+       }
+#else
+       /*
+        * This a non-SMP scenario where the singluar Rx thread is
+        * being removed. Free all skbs and stop the thread.
+        */
+       spin_lock_bh(&p->fcoe_rx_list.lock);
+       while ((skb = __skb_dequeue(&p->fcoe_rx_list)) != NULL)
+               kfree_skb(skb);
+       spin_unlock_bh(&p->fcoe_rx_list.lock);
+#endif
+
+       if (thread)
+               kthread_stop(thread);
+
+       if (crc_eof)
+               put_page(crc_eof);
+}
+
+/**
+ * fcoe_cpu_callback() - fcoe cpu hotplug event callback
+ * @nfb: callback data block
+ * @action: event triggering the callback
+ * @hcpu: index for the cpu of this event
+ *
+ * This creates or destroys per cpu data for fcoe
+ *
+ * Returns NOTIFY_OK always.
+ */
+static int fcoe_cpu_callback(struct notifier_block *nfb,
+                            unsigned long action, void *hcpu)
+{
+       unsigned cpu = (unsigned long)hcpu;
+
+       switch (action) {
+       case CPU_ONLINE:
+       case CPU_ONLINE_FROZEN:
+               FC_DBG("CPU %x online: Create Rx thread\n", cpu);
+               fcoe_percpu_thread_create(cpu);
+               break;
+       case CPU_DEAD:
+       case CPU_DEAD_FROZEN:
+               FC_DBG("CPU %x offline: Remove Rx thread\n", cpu);
+               fcoe_percpu_thread_destroy(cpu);
+               break;
+       default:
+               break;
+       }
+       return NOTIFY_OK;
+}
+
+static struct notifier_block fcoe_cpu_notifier = {
+       .notifier_call = fcoe_cpu_callback,
+};
+
+/**
+ * fcoe_rcv() - this is the fcoe receive function called by NET_RX_SOFTIRQ
+ * @skb: the receive skb
+ * @dev: associated net device
+ * @ptype: context
+ * @odldev: last device
+ *
+ * this function will receive the packet and build fc frame and pass it up
+ *
+ * Returns: 0 for success
+ */
+int fcoe_rcv(struct sk_buff *skb, struct net_device *dev,
+            struct packet_type *ptype, struct net_device *olddev)
+{
+       struct fc_lport *lp;
+       struct fcoe_rcv_info *fr;
+       struct fcoe_softc *fc;
+       struct fc_frame_header *fh;
+       struct fcoe_percpu_s *fps;
+       unsigned short oxid;
+       unsigned int cpu = 0;
+
+       fc = container_of(ptype, struct fcoe_softc, fcoe_packet_type);
+       lp = fc->ctlr.lp;
+       if (unlikely(lp == NULL)) {
+               FC_DBG("cannot find hba structure");
+               goto err2;
+       }
+       if (!lp->link_up)
+               goto err2;
+
+       if (unlikely(debug_fcoe)) {
+               FC_DBG("skb_info: len:%d data_len:%d head:%p data:%p tail:%p "
+                      "end:%p sum:%d dev:%s", skb->len, skb->data_len,
+                      skb->head, skb->data, skb_tail_pointer(skb),
+                      skb_end_pointer(skb), skb->csum,
+                      skb->dev ? skb->dev->name : "<NULL>");
+
+       }
+
+       /* check for FCOE packet type */
+       if (unlikely(eth_hdr(skb)->h_proto != htons(ETH_P_FCOE))) {
+               FC_DBG("wrong FC type frame");
+               goto err;
+       }
+
+       /*
+        * Check for minimum frame length, and make sure required FCoE
+        * and FC headers are pulled into the linear data area.
+        */
+       if (unlikely((skb->len < FCOE_MIN_FRAME) ||
+           !pskb_may_pull(skb, FCOE_HEADER_LEN)))
+               goto err;
+
+       skb_set_transport_header(skb, sizeof(struct fcoe_hdr));
+       fh = (struct fc_frame_header *) skb_transport_header(skb);
+
+       oxid = ntohs(fh->fh_ox_id);
+
+       fr = fcoe_dev_from_skb(skb);
+       fr->fr_dev = lp;
+       fr->ptype = ptype;
+
+#ifdef CONFIG_SMP
+       /*
+        * The incoming frame exchange id(oxid) is ANDed with num of online
+        * cpu bits to get cpu and then this cpu is used for selecting
+        * a per cpu kernel thread from fcoe_percpu.
+        */
+       cpu = oxid & (num_online_cpus() - 1);
+#endif
+
+       fps = &per_cpu(fcoe_percpu, cpu);
+       spin_lock_bh(&fps->fcoe_rx_list.lock);
+       if (unlikely(!fps->thread)) {
+               /*
+                * The targeted CPU is not ready, let's target
+                * the first CPU now. For non-SMP systems this
+                * will check the same CPU twice.
+                */
+               FC_DBG("CPU is online, but no receive thread ready "
+                      "for incoming skb- using first online CPU.\n");
+
+               spin_unlock_bh(&fps->fcoe_rx_list.lock);
+               cpu = first_cpu(cpu_online_map);
+               fps = &per_cpu(fcoe_percpu, cpu);
+               spin_lock_bh(&fps->fcoe_rx_list.lock);
+               if (!fps->thread) {
+                       spin_unlock_bh(&fps->fcoe_rx_list.lock);
+                       goto err;
+               }
+       }
+
+       /*
+        * We now have a valid CPU that we're targeting for
+        * this skb. We also have this receive thread locked,
+        * so we're free to queue skbs into it's queue.
+        */
+       __skb_queue_tail(&fps->fcoe_rx_list, skb);
+       if (fps->fcoe_rx_list.qlen == 1)
+               wake_up_process(fps->thread);
+
+       spin_unlock_bh(&fps->fcoe_rx_list.lock);
+
+       return 0;
+err:
+       fc_lport_get_stats(lp)->ErrorFrames++;
+
+err2:
+       kfree_skb(skb);
+       return -1;
+}
+EXPORT_SYMBOL_GPL(fcoe_rcv);
+
+/**
+ * fcoe_start_io() - pass to netdev to start xmit for fcoe
+ * @skb: the skb to be xmitted
+ *
+ * Returns: 0 for success
+ */
+static inline int fcoe_start_io(struct sk_buff *skb)
+{
+       int rc;
+
+       skb_get(skb);
+       rc = dev_queue_xmit(skb);
+       if (rc != 0)
+               return rc;
+       kfree_skb(skb);
+       return 0;
+}
+
+/**
+ * fcoe_get_paged_crc_eof() - in case we need alloc a page for crc_eof
+ * @skb: the skb to be xmitted
+ * @tlen: total len
+ *
+ * Returns: 0 for success
+ */
+static int fcoe_get_paged_crc_eof(struct sk_buff *skb, int tlen)
+{
+       struct fcoe_percpu_s *fps;
+       struct page *page;
+
+       fps = &get_cpu_var(fcoe_percpu);
+       page = fps->crc_eof_page;
+       if (!page) {
+               page = alloc_page(GFP_ATOMIC);
+               if (!page) {
+                       put_cpu_var(fcoe_percpu);
+                       return -ENOMEM;
+               }
+               fps->crc_eof_page = page;
+               fps->crc_eof_offset = 0;
+       }
+
+       get_page(page);
+       skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags, page,
+                          fps->crc_eof_offset, tlen);
+       skb->len += tlen;
+       skb->data_len += tlen;
+       skb->truesize += tlen;
+       fps->crc_eof_offset += sizeof(struct fcoe_crc_eof);
+
+       if (fps->crc_eof_offset >= PAGE_SIZE) {
+               fps->crc_eof_page = NULL;
+               fps->crc_eof_offset = 0;
+               put_page(page);
+       }
+       put_cpu_var(fcoe_percpu);
+       return 0;
+}
+
+/**
+ * fcoe_fc_crc() - calculates FC CRC in this fcoe skb
+ * @fp: the fc_frame containg data to be checksummed
+ *
+ * This uses crc32() to calculate the crc for fc frame
+ * Return   : 32 bit crc
+ */
+u32 fcoe_fc_crc(struct fc_frame *fp)
+{
+       struct sk_buff *skb = fp_skb(fp);
+       struct skb_frag_struct *frag;
+       unsigned char *data;
+       unsigned long off, len, clen;
+       u32 crc;
+       unsigned i;
+
+       crc = crc32(~0, skb->data, skb_headlen(skb));
+
+       for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
+               frag = &skb_shinfo(skb)->frags[i];
+               off = frag->page_offset;
+               len = frag->size;
+               while (len > 0) {
+                       clen = min(len, PAGE_SIZE - (off & ~PAGE_MASK));
+                       data = kmap_atomic(frag->page + (off >> PAGE_SHIFT),
+                                          KM_SKB_DATA_SOFTIRQ);
+                       crc = crc32(crc, data + (off & ~PAGE_MASK), clen);
+                       kunmap_atomic(data, KM_SKB_DATA_SOFTIRQ);
+                       off += clen;
+                       len -= clen;
+               }
+       }
+       return crc;
+}
+
+/**
+ * fcoe_xmit() - FCoE frame transmit function
+ * @lp:        the associated local port
+ * @fp: the fc_frame to be transmitted
+ *
+ * Return   : 0 for success
+ */
+int fcoe_xmit(struct fc_lport *lp, struct fc_frame *fp)
+{
+       int wlen, rc = 0;
+       u32 crc;
+       struct ethhdr *eh;
+       struct fcoe_crc_eof *cp;
+       struct sk_buff *skb;
+       struct fcoe_dev_stats *stats;
+       struct fc_frame_header *fh;
+       unsigned int hlen;              /* header length implies the version */
+       unsigned int tlen;              /* trailer length */
+       unsigned int elen;              /* eth header, may include vlan */
+       struct fcoe_softc *fc;
+       u8 sof, eof;
+       struct fcoe_hdr *hp;
+
+       WARN_ON((fr_len(fp) % sizeof(u32)) != 0);
+
+       fc = lport_priv(lp);
+       fh = fc_frame_header_get(fp);
+       skb = fp_skb(fp);
+       wlen = skb->len / FCOE_WORD_TO_BYTE;
+
+       if (!lp->link_up) {
+               kfree(skb);
+               return 0;
+       }
+
+       if (unlikely(fh->fh_r_ctl == FC_RCTL_ELS_REQ) &&
+           fcoe_ctlr_els_send(&fc->ctlr, skb))
+               return 0;
+
+       sof = fr_sof(fp);
+       eof = fr_eof(fp);
+
+       elen = (fc->real_dev->priv_flags & IFF_802_1Q_VLAN) ?
+               sizeof(struct vlan_ethhdr) : sizeof(struct ethhdr);
+       hlen = sizeof(struct fcoe_hdr);
+       tlen = sizeof(struct fcoe_crc_eof);
+       wlen = (skb->len - tlen + sizeof(crc)) / FCOE_WORD_TO_BYTE;
+
+       /* crc offload */
+       if (likely(lp->crc_offload)) {
+               skb->ip_summed = CHECKSUM_PARTIAL;
+               skb->csum_start = skb_headroom(skb);
+               skb->csum_offset = skb->len;
+               crc = 0;
+       } else {
+               skb->ip_summed = CHECKSUM_NONE;
+               crc = fcoe_fc_crc(fp);
+       }
+
+       /* copy fc crc and eof to the skb buff */
+       if (skb_is_nonlinear(skb)) {
+               skb_frag_t *frag;
+               if (fcoe_get_paged_crc_eof(skb, tlen)) {
+                       kfree_skb(skb);
+                       return -ENOMEM;
+               }
+               frag = &skb_shinfo(skb)->frags[skb_shinfo(skb)->nr_frags - 1];
+               cp = kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ)
+                       + frag->page_offset;
+       } else {
+               cp = (struct fcoe_crc_eof *)skb_put(skb, tlen);
+       }
+
+       memset(cp, 0, sizeof(*cp));
+       cp->fcoe_eof = eof;
+       cp->fcoe_crc32 = cpu_to_le32(~crc);
+
+       if (skb_is_nonlinear(skb)) {
+               kunmap_atomic(cp, KM_SKB_DATA_SOFTIRQ);
+               cp = NULL;
+       }
+
+       /* adjust skb netowrk/transport offsets to match mac/fcoe/fc */
+       skb_push(skb, elen + hlen);
+       skb_reset_mac_header(skb);
+       skb_reset_network_header(skb);
+       skb->mac_len = elen;
+       skb->protocol = htons(ETH_P_FCOE);
+       skb->dev = fc->real_dev;
+
+       /* fill up mac and fcoe headers */
+       eh = eth_hdr(skb);
+       eh->h_proto = htons(ETH_P_FCOE);
+       if (fc->ctlr.map_dest)
+               fc_fcoe_set_mac(eh->h_dest, fh->fh_d_id);
+       else
+               /* insert GW address */
+               memcpy(eh->h_dest, fc->ctlr.dest_addr, ETH_ALEN);
+
+       if (unlikely(fc->ctlr.flogi_oxid != FC_XID_UNKNOWN))
+               memcpy(eh->h_source, fc->ctlr.ctl_src_addr, ETH_ALEN);
+       else
+               memcpy(eh->h_source, fc->ctlr.data_src_addr, ETH_ALEN);
+
+       hp = (struct fcoe_hdr *)(eh + 1);
+       memset(hp, 0, sizeof(*hp));
+       if (FC_FCOE_VER)
+               FC_FCOE_ENCAPS_VER(hp, FC_FCOE_VER);
+       hp->fcoe_sof = sof;
+
+#ifdef NETIF_F_FSO
+       /* fcoe lso, mss is in max_payload which is non-zero for FCP data */
+       if (lp->seq_offload && fr_max_payload(fp)) {
+               skb_shinfo(skb)->gso_type = SKB_GSO_FCOE;
+               skb_shinfo(skb)->gso_size = fr_max_payload(fp);
+       } else {
+               skb_shinfo(skb)->gso_type = 0;
+               skb_shinfo(skb)->gso_size = 0;
+       }
+#endif
+       /* update tx stats: regardless if LLD fails */
+       stats = fc_lport_get_stats(lp);
+       stats->TxFrames++;
+       stats->TxWords += wlen;
+
+       /* send down to lld */
+       fr_dev(fp) = lp;
+       if (fc->fcoe_pending_queue.qlen)
+               rc = fcoe_check_wait_queue(lp);
+
+       if (rc == 0)
+               rc = fcoe_start_io(skb);
+
+       if (rc) {
+               spin_lock_bh(&fc->fcoe_pending_queue.lock);
+               __skb_queue_tail(&fc->fcoe_pending_queue, skb);
+               spin_unlock_bh(&fc->fcoe_pending_queue.lock);
+               if (fc->fcoe_pending_queue.qlen > FCOE_MAX_QUEUE_DEPTH)
+                       lp->qfull = 1;
+       }
+
+       return 0;
+}
+EXPORT_SYMBOL_GPL(fcoe_xmit);
+
+/**
+ * fcoe_percpu_receive_thread() - recv thread per cpu
+ * @arg: ptr to the fcoe per cpu struct
+ *
+ * Return: 0 for success
+ */
+int fcoe_percpu_receive_thread(void *arg)
+{
+       struct fcoe_percpu_s *p = arg;
+       u32 fr_len;
+       struct fc_lport *lp;
+       struct fcoe_rcv_info *fr;
+       struct fcoe_dev_stats *stats;
+       struct fc_frame_header *fh;
+       struct sk_buff *skb;
+       struct fcoe_crc_eof crc_eof;
+       struct fc_frame *fp;
+       u8 *mac = NULL;
+       struct fcoe_softc *fc;
+       struct fcoe_hdr *hp;
+
+       set_user_nice(current, -20);
+
+       while (!kthread_should_stop()) {
+
+               spin_lock_bh(&p->fcoe_rx_list.lock);
+               while ((skb = __skb_dequeue(&p->fcoe_rx_list)) == NULL) {
+                       set_current_state(TASK_INTERRUPTIBLE);
+                       spin_unlock_bh(&p->fcoe_rx_list.lock);
+                       schedule();
+                       set_current_state(TASK_RUNNING);
+                       if (kthread_should_stop())
+                               return 0;
+                       spin_lock_bh(&p->fcoe_rx_list.lock);
+               }
+               spin_unlock_bh(&p->fcoe_rx_list.lock);
+               fr = fcoe_dev_from_skb(skb);
+               lp = fr->fr_dev;
+               if (unlikely(lp == NULL)) {
+                       FC_DBG("invalid HBA Structure");
+                       kfree_skb(skb);
+                       continue;
+               }
+
+               if (unlikely(debug_fcoe)) {
+                       FC_DBG("skb_info: len:%d data_len:%d head:%p data:%p "
+                              "tail:%p end:%p sum:%d dev:%s",
+                              skb->len, skb->data_len,
+                              skb->head, skb->data, skb_tail_pointer(skb),
+                              skb_end_pointer(skb), skb->csum,
+                              skb->dev ? skb->dev->name : "<NULL>");
+               }
+
+               /*
+                * Save source MAC address before discarding header.
+                */
+               fc = lport_priv(lp);
+               if (skb_is_nonlinear(skb))
+                       skb_linearize(skb);     /* not ideal */
+               mac = eth_hdr(skb)->h_source;
+
+               /*
+                * Frame length checks and setting up the header pointers
+                * was done in fcoe_rcv already.
+                */
+               hp = (struct fcoe_hdr *) skb_network_header(skb);
+               fh = (struct fc_frame_header *) skb_transport_header(skb);
+
+               stats = fc_lport_get_stats(lp);
+               if (unlikely(FC_FCOE_DECAPS_VER(hp) != FC_FCOE_VER)) {
+                       if (stats->ErrorFrames < 5)
+                               printk(KERN_WARNING "FCoE version "
+                                      "mismatch: The frame has "
+                                      "version %x, but the "
+                                      "initiator supports version "
+                                      "%x\n", FC_FCOE_DECAPS_VER(hp),
+                                      FC_FCOE_VER);
+                       stats->ErrorFrames++;
+                       kfree_skb(skb);
+                       continue;
+               }
+
+               skb_pull(skb, sizeof(struct fcoe_hdr));
+               fr_len = skb->len - sizeof(struct fcoe_crc_eof);
+
+               stats->RxFrames++;
+               stats->RxWords += fr_len / FCOE_WORD_TO_BYTE;
+
+               fp = (struct fc_frame *)skb;
+               fc_frame_init(fp);
+               fr_dev(fp) = lp;
+               fr_sof(fp) = hp->fcoe_sof;
+
+               /* Copy out the CRC and EOF trailer for access */
+               if (skb_copy_bits(skb, fr_len, &crc_eof, sizeof(crc_eof))) {
+                       kfree_skb(skb);
+                       continue;
+               }
+               fr_eof(fp) = crc_eof.fcoe_eof;
+               fr_crc(fp) = crc_eof.fcoe_crc32;
+               if (pskb_trim(skb, fr_len)) {
+                       kfree_skb(skb);
+                       continue;
+               }
+
+               /*
+                * We only check CRC if no offload is available and if it is
+                * it's solicited data, in which case, the FCP layer would
+                * check it during the copy.
+                */
+               if (lp->crc_offload && skb->ip_summed == CHECKSUM_UNNECESSARY)
+                       fr_flags(fp) &= ~FCPHF_CRC_UNCHECKED;
+               else
+                       fr_flags(fp) |= FCPHF_CRC_UNCHECKED;
+
+               fh = fc_frame_header_get(fp);
+               if (fh->fh_r_ctl == FC_RCTL_DD_SOL_DATA &&
+                   fh->fh_type == FC_TYPE_FCP) {
+                       fc_exch_recv(lp, lp->emp, fp);
+                       continue;
+               }
+               if (fr_flags(fp) & FCPHF_CRC_UNCHECKED) {
+                       if (le32_to_cpu(fr_crc(fp)) !=
+                           ~crc32(~0, skb->data, fr_len)) {
+                               if (debug_fcoe || stats->InvalidCRCCount < 5)
+                                       printk(KERN_WARNING "fcoe: dropping "
+                                              "frame with CRC error\n");
+                               stats->InvalidCRCCount++;
+                               stats->ErrorFrames++;
+                               fc_frame_free(fp);
+                               continue;
+                       }
+                       fr_flags(fp) &= ~FCPHF_CRC_UNCHECKED;
+               }
+               if (unlikely(fc->ctlr.flogi_oxid != FC_XID_UNKNOWN) &&
+                   fcoe_ctlr_recv_flogi(&fc->ctlr, fp, mac)) {
+                       fc_frame_free(fp);
+                       continue;
+               }
+               fc_exch_recv(lp, lp->emp, fp);
+       }
+       return 0;
+}
+
+/**
+ * fcoe_watchdog() - fcoe timer callback
+ * @vp:
+ *
+ * This checks the pending queue length for fcoe and set lport qfull
+ * if the FCOE_MAX_QUEUE_DEPTH is reached. This is done for all fc_lport on the
+ * fcoe_hostlist.
+ *
+ * Returns: 0 for success
+ */
+void fcoe_watchdog(ulong vp)
+{
+       struct fcoe_softc *fc;
+
+       read_lock(&fcoe_hostlist_lock);
+       list_for_each_entry(fc, &fcoe_hostlist, list) {
+               if (fc->ctlr.lp)
+                       fcoe_check_wait_queue(fc->ctlr.lp);
+       }
+       read_unlock(&fcoe_hostlist_lock);
+
+       fcoe_timer.expires = jiffies + (1 * HZ);
+       add_timer(&fcoe_timer);
+}
+
+
+/**
+ * fcoe_check_wait_queue() - put the skb into fcoe pending xmit queue
+ * @lp: the fc_port for this skb
+ * @skb: the associated skb to be xmitted
+ *
+ * This empties the wait_queue, dequeue the head of the wait_queue queue
+ * and calls fcoe_start_io() for each packet, if all skb have been
+ * transmitted, return qlen or -1 if a error occurs, then restore
+ * wait_queue and  try again later.
+ *
+ * The wait_queue is used when the skb transmit fails. skb will go
+ * in the wait_queue which will be emptied by the time function OR
+ * by the next skb transmit.
+ *
+ * Returns: 0 for success
+ */
+static int fcoe_check_wait_queue(struct fc_lport *lp)
+{
+       struct fcoe_softc *fc = lport_priv(lp);
+       struct sk_buff *skb;
+       int rc = -1;
+
+       spin_lock_bh(&fc->fcoe_pending_queue.lock);
+       if (fc->fcoe_pending_queue_active)
+               goto out;
+       fc->fcoe_pending_queue_active = 1;
+
+       while (fc->fcoe_pending_queue.qlen) {
+               /* keep qlen > 0 until fcoe_start_io succeeds */
+               fc->fcoe_pending_queue.qlen++;
+               skb = __skb_dequeue(&fc->fcoe_pending_queue);
+
+               spin_unlock_bh(&fc->fcoe_pending_queue.lock);
+               rc = fcoe_start_io(skb);
+               spin_lock_bh(&fc->fcoe_pending_queue.lock);
+
+               if (rc) {
+                       __skb_queue_head(&fc->fcoe_pending_queue, skb);
+                       /* undo temporary increment above */
+                       fc->fcoe_pending_queue.qlen--;
+                       break;
+               }
+               /* undo temporary increment above */
+               fc->fcoe_pending_queue.qlen--;
+       }
+
+       if (fc->fcoe_pending_queue.qlen < FCOE_LOW_QUEUE_DEPTH)
+               lp->qfull = 0;
+       fc->fcoe_pending_queue_active = 0;
+       rc = fc->fcoe_pending_queue.qlen;
+out:
+       spin_unlock_bh(&fc->fcoe_pending_queue.lock);
+       return rc;
+}
+
+/**
+ * fcoe_dev_setup() - setup link change notification interface
+ */
+static void fcoe_dev_setup()
+{
+       /*
+        * here setup a interface specific wd time to
+        * monitor the link state
+        */
+       register_netdevice_notifier(&fcoe_notifier);
+}
+
+/**
+ * fcoe_dev_setup() - cleanup link change notification interface
+ */
+static void fcoe_dev_cleanup(void)
+{
+       unregister_netdevice_notifier(&fcoe_notifier);
+}
+
+/**
+ * fcoe_device_notification() - netdev event notification callback
+ * @notifier: context of the notification
+ * @event: type of event
+ * @ptr: fixed array for output parsed ifname
+ *
+ * This function is called by the ethernet driver in case of link change event
+ *
+ * Returns: 0 for success
+ */
+static int fcoe_device_notification(struct notifier_block *notifier,
+                                   ulong event, void *ptr)
+{
+       struct fc_lport *lp = NULL;
+       struct net_device *real_dev = ptr;
+       struct fcoe_softc *fc;
+       struct fcoe_dev_stats *stats;
+       u32 link_possible = 1;
+       u32 mfs;
+       int rc = NOTIFY_OK;
+
+       read_lock(&fcoe_hostlist_lock);
+       list_for_each_entry(fc, &fcoe_hostlist, list) {
+               if (fc->real_dev == real_dev) {
+                       lp = fc->ctlr.lp;
+                       break;
+               }
+       }
+       read_unlock(&fcoe_hostlist_lock);
+       if (lp == NULL) {
+               rc = NOTIFY_DONE;
+               goto out;
+       }
+
+       switch (event) {
+       case NETDEV_DOWN:
+       case NETDEV_GOING_DOWN:
+               link_possible = 0;
+               break;
+       case NETDEV_UP:
+       case NETDEV_CHANGE:
+               break;
+       case NETDEV_CHANGEMTU:
+               mfs = fc->real_dev->mtu -
+                       (sizeof(struct fcoe_hdr) +
+                        sizeof(struct fcoe_crc_eof));
+               if (mfs >= FC_MIN_MAX_FRAME)
+                       fc_set_mfs(lp, mfs);
+               break;
+       case NETDEV_REGISTER:
+               break;
+       default:
+               FC_DBG("Unknown event %ld from netdev netlink\n", event);
+       }
+       if (link_possible && !fcoe_link_ok(lp))
+               fcoe_ctlr_link_up(&fc->ctlr);
+       else if (fcoe_ctlr_link_down(&fc->ctlr)) {
+               stats = fc_lport_get_stats(lp);
+               stats->LinkFailureCount++;
+               fcoe_clean_pending_queue(lp);
+       }
+out:
+       return rc;
+}
+
+/**
+ * fcoe_if_to_netdev() - parse a name buffer to get netdev
+ * @ifname: fixed array for output parsed ifname
+ * @buffer: incoming buffer to be copied
+ *
+ * Returns: NULL or ptr to netdeive
+ */
+static struct net_device *fcoe_if_to_netdev(const char *buffer)
+{
+       char *cp;
+       char ifname[IFNAMSIZ + 2];
+
+       if (buffer) {
+               strlcpy(ifname, buffer, IFNAMSIZ);
+               cp = ifname + strlen(ifname);
+               while (--cp >= ifname && *cp == '\n')
+                       *cp = '\0';
+               return dev_get_by_name(&init_net, ifname);
+       }
+       return NULL;
+}
+
+/**
+ * fcoe_netdev_to_module_owner() - finds out the nic drive moddule of the netdev
+ * @netdev: the target netdev
+ *
+ * Returns: ptr to the struct module, NULL for failure
+ */
+static struct module *
+fcoe_netdev_to_module_owner(const struct net_device *netdev)
+{
+       struct device *dev;
+
+       if (!netdev)
+               return NULL;
+
+       dev = netdev->dev.parent;
+       if (!dev)
+               return NULL;
+
+       if (!dev->driver)
+               return NULL;
+
+       return dev->driver->owner;
+}
+
+/**
+ * fcoe_ethdrv_get() - Hold the Ethernet driver
+ * @netdev: the target netdev
+ *
+ * Holds the Ethernet driver module by try_module_get() for
+ * the corresponding netdev.
+ *
+ * Returns: 0 for succsss
+ */
+static int fcoe_ethdrv_get(const struct net_device *netdev)
+{
+       struct module *owner;
+
+       owner = fcoe_netdev_to_module_owner(netdev);
+       if (owner) {
+               printk(KERN_DEBUG "fcoe:hold driver module %s for %s\n",
+                      module_name(owner), netdev->name);
+               return  try_module_get(owner);
+       }
+       return -ENODEV;
+}
+
+/**
+ * fcoe_ethdrv_put() - Release the Ethernet driver
+ * @netdev: the target netdev
+ *
+ * Releases the Ethernet driver module by module_put for
+ * the corresponding netdev.
+ *
+ * Returns: 0 for succsss
+ */
+static int fcoe_ethdrv_put(const struct net_device *netdev)
+{
+       struct module *owner;
+
+       owner = fcoe_netdev_to_module_owner(netdev);
+       if (owner) {
+               printk(KERN_DEBUG "fcoe:release driver module %s for %s\n",
+                      module_name(owner), netdev->name);
+               module_put(owner);
+               return 0;
+       }
+       return -ENODEV;
+}
+
+/**
+ * fcoe_destroy() - handles the destroy from sysfs
+ * @buffer: expcted to be a eth if name
+ * @kp: associated kernel param
+ *
+ * Returns: 0 for success
+ */
+static int fcoe_destroy(const char *buffer, struct kernel_param *kp)
+{
+       int rc;
+       struct net_device *netdev;
+
+       netdev = fcoe_if_to_netdev(buffer);
+       if (!netdev) {
+               rc = -ENODEV;
+               goto out_nodev;
+       }
+       /* look for existing lport */
+       if (!fcoe_hostlist_lookup(netdev)) {
+               rc = -ENODEV;
+               goto out_putdev;
+       }
+       rc = fcoe_if_destroy(netdev);
+       if (rc) {
+               printk(KERN_ERR "fcoe: fcoe_if_destroy(%s) failed\n",
+                      netdev->name);
+               rc = -EIO;
+               goto out_putdev;
+       }
+       fcoe_ethdrv_put(netdev);
+       rc = 0;
+out_putdev:
+       dev_put(netdev);
+out_nodev:
+       return rc;
+}
+
+/**
+ * fcoe_create() - Handles the create call from sysfs
+ * @buffer: expcted to be a eth if name
+ * @kp: associated kernel param
+ *
+ * Returns: 0 for success
+ */
+static int fcoe_create(const char *buffer, struct kernel_param *kp)
+{
+       int rc;
+       struct net_device *netdev;
+
+       netdev = fcoe_if_to_netdev(buffer);
+       if (!netdev) {
+               rc = -ENODEV;
+               goto out_nodev;
+       }
+       /* look for existing lport */
+       if (fcoe_hostlist_lookup(netdev)) {
+               rc = -EEXIST;
+               goto out_putdev;
+       }
+       fcoe_ethdrv_get(netdev);
+
+       rc = fcoe_if_create(netdev);
+       if (rc) {
+               printk(KERN_ERR "fcoe: fcoe_if_create(%s) failed\n",
+                      netdev->name);
+               fcoe_ethdrv_put(netdev);
+               rc = -EIO;
+               goto out_putdev;
+       }
+       rc = 0;
+out_putdev:
+       dev_put(netdev);
+out_nodev:
+       return rc;
+}
+
+module_param_call(create, fcoe_create, NULL, NULL, S_IWUSR);
+__MODULE_PARM_TYPE(create, "string");
+MODULE_PARM_DESC(create, "Create fcoe port using net device passed in.");
+module_param_call(destroy, fcoe_destroy, NULL, NULL, S_IWUSR);
+__MODULE_PARM_TYPE(destroy, "string");
+MODULE_PARM_DESC(destroy, "Destroy fcoe port");
+
+/**
+ * fcoe_link_ok() - Check if link is ok for the fc_lport
+ * @lp: ptr to the fc_lport
+ *
+ * Any permanently-disqualifying conditions have been previously checked.
+ * This also updates the speed setting, which may change with link for 100/1000.
+ *
+ * This function should probably be checking for PAUSE support at some point
+ * in the future. Currently Per-priority-pause is not determinable using
+ * ethtool, so we shouldn't be restrictive until that problem is resolved.
+ *
+ * Returns: 0 if link is OK for use by FCoE.
+ *
+ */
+int fcoe_link_ok(struct fc_lport *lp)
+{
+       struct fcoe_softc *fc = lport_priv(lp);
+       struct net_device *dev = fc->real_dev;
+       struct ethtool_cmd ecmd = { ETHTOOL_GSET };
+       int rc = 0;
+
+       if ((dev->flags & IFF_UP) && netif_carrier_ok(dev)) {
+               dev = fc->phys_dev;
+               if (dev->ethtool_ops->get_settings) {
+                       dev->ethtool_ops->get_settings(dev, &ecmd);
+                       lp->link_supported_speeds &=
+                               ~(FC_PORTSPEED_1GBIT | FC_PORTSPEED_10GBIT);
+                       if (ecmd.supported & (SUPPORTED_1000baseT_Half |
+                                             SUPPORTED_1000baseT_Full))
+                               lp->link_supported_speeds |= FC_PORTSPEED_1GBIT;
+                       if (ecmd.supported & SUPPORTED_10000baseT_Full)
+                               lp->link_supported_speeds |=
+                                       FC_PORTSPEED_10GBIT;
+                       if (ecmd.speed == SPEED_1000)
+                               lp->link_speed = FC_PORTSPEED_1GBIT;
+                       if (ecmd.speed == SPEED_10000)
+                               lp->link_speed = FC_PORTSPEED_10GBIT;
+               }
+       } else
+               rc = -1;
+
+       return rc;
+}
+EXPORT_SYMBOL_GPL(fcoe_link_ok);
+
+/**
+ * fcoe_percpu_clean() - Clear the pending skbs for an lport
+ * @lp: the fc_lport
+ */
+void fcoe_percpu_clean(struct fc_lport *lp)
+{
+       struct fcoe_percpu_s *pp;
+       struct fcoe_rcv_info *fr;
+       struct sk_buff_head *list;
+       struct sk_buff *skb, *next;
+       struct sk_buff *head;
+       unsigned int cpu;
+
+       for_each_possible_cpu(cpu) {
+               pp = &per_cpu(fcoe_percpu, cpu);
+               spin_lock_bh(&pp->fcoe_rx_list.lock);
+               list = &pp->fcoe_rx_list;
+               head = list->next;
+               for (skb = head; skb != (struct sk_buff *)list;
+                    skb = next) {
+                       next = skb->next;
+                       fr = fcoe_dev_from_skb(skb);
+                       if (fr->fr_dev == lp) {
+                               __skb_unlink(skb, list);
+                               kfree_skb(skb);
+                       }
+               }
+               spin_unlock_bh(&pp->fcoe_rx_list.lock);
+       }
+}
+EXPORT_SYMBOL_GPL(fcoe_percpu_clean);
+
+/**
+ * fcoe_clean_pending_queue() - Dequeue a skb and free it
+ * @lp: the corresponding fc_lport
+ *
+ * Returns: none
+ */
+void fcoe_clean_pending_queue(struct fc_lport *lp)
+{
+       struct fcoe_softc  *fc = lport_priv(lp);
+       struct sk_buff *skb;
+
+       spin_lock_bh(&fc->fcoe_pending_queue.lock);
+       while ((skb = __skb_dequeue(&fc->fcoe_pending_queue)) != NULL) {
+               spin_unlock_bh(&fc->fcoe_pending_queue.lock);
+               kfree_skb(skb);
+               spin_lock_bh(&fc->fcoe_pending_queue.lock);
+       }
+       spin_unlock_bh(&fc->fcoe_pending_queue.lock);
+}
+EXPORT_SYMBOL_GPL(fcoe_clean_pending_queue);
+
+/**
+ * fcoe_reset() - Resets the fcoe
+ * @shost: shost the reset is from
+ *
+ * Returns: always 0
+ */
+int fcoe_reset(struct Scsi_Host *shost)
+{
+       struct fc_lport *lport = shost_priv(shost);
+       fc_lport_reset(lport);
+       return 0;
+}
+EXPORT_SYMBOL_GPL(fcoe_reset);
+
+/**
+ * fcoe_hostlist_lookup_softc() - find the corresponding lport by a given device
+ * @device: this is currently ptr to net_device
+ *
+ * Returns: NULL or the located fcoe_softc
+ */
+static struct fcoe_softc *
+fcoe_hostlist_lookup_softc(const struct net_device *dev)
+{
+       struct fcoe_softc *fc;
+
+       read_lock(&fcoe_hostlist_lock);
+       list_for_each_entry(fc, &fcoe_hostlist, list) {
+               if (fc->real_dev == dev) {
+                       read_unlock(&fcoe_hostlist_lock);
+                       return fc;
+               }
+       }
+       read_unlock(&fcoe_hostlist_lock);
+       return NULL;
+}
+
+/**
+ * fcoe_hostlist_lookup() - Find the corresponding lport by netdev
+ * @netdev: ptr to net_device
+ *
+ * Returns: 0 for success
+ */
+struct fc_lport *fcoe_hostlist_lookup(const struct net_device *netdev)
+{
+       struct fcoe_softc *fc;
+
+       fc = fcoe_hostlist_lookup_softc(netdev);
+
+       return (fc) ? fc->ctlr.lp : NULL;
+}
+EXPORT_SYMBOL_GPL(fcoe_hostlist_lookup);
+
+/**
+ * fcoe_hostlist_add() - Add a lport to lports list
+ * @lp: ptr to the fc_lport to badded
+ *
+ * Returns: 0 for success
+ */
+int fcoe_hostlist_add(const struct fc_lport *lp)
+{
+       struct fcoe_softc *fc;
+
+       fc = fcoe_hostlist_lookup_softc(fcoe_netdev(lp));
+       if (!fc) {
+               fc = lport_priv(lp);
+               write_lock_bh(&fcoe_hostlist_lock);
+               list_add_tail(&fc->list, &fcoe_hostlist);
+               write_unlock_bh(&fcoe_hostlist_lock);
+       }
+       return 0;
+}
+EXPORT_SYMBOL_GPL(fcoe_hostlist_add);
+
+/**
+ * fcoe_hostlist_remove() - remove a lport from lports list
+ * @lp: ptr to the fc_lport to badded
+ *
+ * Returns: 0 for success
+ */
+int fcoe_hostlist_remove(const struct fc_lport *lp)
+{
+       struct fcoe_softc *fc;
+
+       fc = fcoe_hostlist_lookup_softc(fcoe_netdev(lp));
+       BUG_ON(!fc);
+       write_lock_bh(&fcoe_hostlist_lock);
+       list_del(&fc->list);
+       write_unlock_bh(&fcoe_hostlist_lock);
+
+       return 0;
+}
+EXPORT_SYMBOL_GPL(fcoe_hostlist_remove);
+
+/**
+ * fcoe_init() - fcoe module loading initialization
+ *
+ * Returns 0 on success, negative on failure
+ */
+static int __init fcoe_init(void)
+{
+       unsigned int cpu;
+       int rc = 0;
+       struct fcoe_percpu_s *p;
+
+       INIT_LIST_HEAD(&fcoe_hostlist);
+       rwlock_init(&fcoe_hostlist_lock);
+
+       for_each_possible_cpu(cpu) {
+               p = &per_cpu(fcoe_percpu, cpu);
+               skb_queue_head_init(&p->fcoe_rx_list);
+       }
+
+       for_each_online_cpu(cpu)
+               fcoe_percpu_thread_create(cpu);
+
+       /* Initialize per CPU interrupt thread */
+       rc = register_hotcpu_notifier(&fcoe_cpu_notifier);
+       if (rc)
+               goto out_free;
+
+       /* Setup link change notification */
+       fcoe_dev_setup();
+
+       setup_timer(&fcoe_timer, fcoe_watchdog, 0);
+
+       mod_timer(&fcoe_timer, jiffies + (10 * HZ));
+
+       fcoe_if_init();
+
+       return 0;
+
+out_free:
+       for_each_online_cpu(cpu) {
+               fcoe_percpu_thread_destroy(cpu);
+       }
+
+       return rc;
+}
+module_init(fcoe_init);
+
+/**
+ * fcoe_exit() - fcoe module unloading cleanup
+ *
+ * Returns 0 on success, negative on failure
+ */
+static void __exit fcoe_exit(void)
+{
+       unsigned int cpu;
+       struct fcoe_softc *fc, *tmp;
+
+       fcoe_dev_cleanup();
+
+       /* Stop the timer */
+       del_timer_sync(&fcoe_timer);
+
+       /* releases the associated fcoe hosts */
+       list_for_each_entry_safe(fc, tmp, &fcoe_hostlist, list)
+               fcoe_if_destroy(fc->real_dev);
+
+       unregister_hotcpu_notifier(&fcoe_cpu_notifier);
+
+       for_each_online_cpu(cpu) {
+               fcoe_percpu_thread_destroy(cpu);
+       }
+
+       /* detach from scsi transport */
+       fcoe_if_exit();
+}
+module_exit(fcoe_exit);