Merge commit 'tip/tracing/core' into oprofile/core
[safe/jmp/linux-2.6] / drivers / net / cnic.c
index a9e2fd3..4b451a7 100644 (file)
@@ -1,6 +1,6 @@
 /* cnic.c: Broadcom CNIC core network driver.
  *
- * Copyright (c) 2006-2009 Broadcom Corporation
+ * Copyright (c) 2006-2010 Broadcom 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
@@ -10,6 +10,8 @@
  * Modified and maintained by: Michael Chan <mchan@broadcom.com>
  */
 
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
 #include <linux/module.h>
 
 #include <linux/kernel.h>
 #include <net/route.h>
 #include <net/ipv6.h>
 #include <net/ip6_route.h>
+#include <net/ip6_checksum.h>
 #include <scsi/iscsi_if.h>
 
 #include "cnic_if.h"
 #include "bnx2.h"
+#include "bnx2x_reg.h"
+#include "bnx2x_fw_defs.h"
+#include "bnx2x_hsi.h"
+#include "../scsi/bnx2i/57xx_iscsi_constants.h"
+#include "../scsi/bnx2i/57xx_iscsi_hsi.h"
 #include "cnic.h"
 #include "cnic_defs.h"
 
 #define DRV_MODULE_NAME                "cnic"
-#define PFX DRV_MODULE_NAME    ": "
 
 static char version[] __devinitdata =
        "Broadcom NetXtreme II CNIC Driver " DRV_MODULE_NAME " v" CNIC_MODULE_VERSION " (" CNIC_MODULE_RELDATE ")\n";
@@ -59,6 +66,7 @@ static DEFINE_MUTEX(cnic_lock);
 static struct cnic_ulp_ops *cnic_ulp_tbl[MAX_CNIC_ULP_TYPE];
 
 static int cnic_service_bnx2(void *, void *);
+static int cnic_service_bnx2x(void *, void *);
 static int cnic_ctl(void *, struct cnic_ctl_info *);
 
 static struct cnic_ops cnic_bnx2_ops = {
@@ -67,9 +75,14 @@ static struct cnic_ops cnic_bnx2_ops = {
        .cnic_ctl       = cnic_ctl,
 };
 
-static void cnic_shutdown_bnx2_rx_ring(struct cnic_dev *);
-static void cnic_init_bnx2_tx_ring(struct cnic_dev *);
-static void cnic_init_bnx2_rx_ring(struct cnic_dev *);
+static struct cnic_ops cnic_bnx2x_ops = {
+       .cnic_owner     = THIS_MODULE,
+       .cnic_handler   = cnic_service_bnx2x,
+       .cnic_ctl       = cnic_ctl,
+};
+
+static void cnic_shutdown_rings(struct cnic_dev *);
+static void cnic_init_rings(struct cnic_dev *);
 static int cnic_cm_set_pg(struct cnic_sock *);
 
 static int cnic_uio_open(struct uio_info *uinfo, struct inode *inode)
@@ -83,12 +96,16 @@ static int cnic_uio_open(struct uio_info *uinfo, struct inode *inode)
        if (cp->uio_dev != -1)
                return -EBUSY;
 
-       cp->uio_dev = iminor(inode);
+       rtnl_lock();
+       if (!test_bit(CNIC_F_CNIC_UP, &dev->flags)) {
+               rtnl_unlock();
+               return -ENODEV;
+       }
 
-       cnic_shutdown_bnx2_rx_ring(dev);
+       cp->uio_dev = iminor(inode);
 
-       cnic_init_bnx2_tx_ring(dev);
-       cnic_init_bnx2_rx_ring(dev);
+       cnic_init_rings(dev);
+       rtnl_unlock();
 
        return 0;
 }
@@ -98,6 +115,8 @@ static int cnic_uio_close(struct uio_info *uinfo, struct inode *inode)
        struct cnic_dev *dev = uinfo->priv;
        struct cnic_local *cp = dev->cnic_priv;
 
+       cnic_shutdown_rings(dev);
+
        cp->uio_dev = -1;
        return 0;
 }
@@ -138,6 +157,16 @@ static struct cnic_dev *cnic_from_netdev(struct net_device *netdev)
        return NULL;
 }
 
+static inline void ulp_get(struct cnic_ulp_ops *ulp_ops)
+{
+       atomic_inc(&ulp_ops->ref_count);
+}
+
+static inline void ulp_put(struct cnic_ulp_ops *ulp_ops)
+{
+       atomic_dec(&ulp_ops->ref_count);
+}
+
 static void cnic_ctx_wr(struct cnic_dev *dev, u32 cid_addr, u32 off, u32 val)
 {
        struct cnic_local *cp = dev->cnic_priv;
@@ -152,6 +181,36 @@ static void cnic_ctx_wr(struct cnic_dev *dev, u32 cid_addr, u32 off, u32 val)
        ethdev->drv_ctl(dev->netdev, &info);
 }
 
+static void cnic_ctx_tbl_wr(struct cnic_dev *dev, u32 off, dma_addr_t addr)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct cnic_eth_dev *ethdev = cp->ethdev;
+       struct drv_ctl_info info;
+       struct drv_ctl_io *io = &info.data.io;
+
+       info.cmd = DRV_CTL_CTXTBL_WR_CMD;
+       io->offset = off;
+       io->dma_addr = addr;
+       ethdev->drv_ctl(dev->netdev, &info);
+}
+
+static void cnic_ring_ctl(struct cnic_dev *dev, u32 cid, u32 cl_id, int start)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct cnic_eth_dev *ethdev = cp->ethdev;
+       struct drv_ctl_info info;
+       struct drv_ctl_l2_ring *ring = &info.data.ring;
+
+       if (start)
+               info.cmd = DRV_CTL_START_L2_CMD;
+       else
+               info.cmd = DRV_CTL_STOP_L2_CMD;
+
+       ring->cid = cid;
+       ring->client_id = cl_id;
+       ethdev->drv_ctl(dev->netdev, &info);
+}
+
 static void cnic_reg_wr_ind(struct cnic_dev *dev, u32 off, u32 val)
 {
        struct cnic_local *cp = dev->cnic_priv;
@@ -194,6 +253,19 @@ static void cnic_kwq_completion(struct cnic_dev *dev, u32 count)
        ethdev->drv_ctl(dev->netdev, &info);
 }
 
+static int cnic_get_l5_cid(struct cnic_local *cp, u32 cid, u32 *l5_cid)
+{
+       u32 i;
+
+       for (i = 0; i < MAX_ISCSI_TBL_SZ; i++) {
+               if (cp->ctx_tbl[i].cid == cid) {
+                       *l5_cid = i;
+                       return 0;
+               }
+       }
+       return -EINVAL;
+}
+
 static int cnic_send_nlmsg(struct cnic_local *cp, u32 type,
                           struct cnic_sock *csk)
 {
@@ -227,7 +299,7 @@ static int cnic_send_nlmsg(struct cnic_local *cp, u32 type,
        }
 
        rcu_read_lock();
-       ulp_ops = rcu_dereference(cp->ulp_ops[CNIC_ULP_ISCSI]);
+       ulp_ops = rcu_dereference(cnic_ulp_tbl[CNIC_ULP_ISCSI]);
        if (ulp_ops)
                ulp_ops->iscsi_nl_send_msg(cp->dev, msg_type, buf, len);
        rcu_read_unlock();
@@ -255,6 +327,12 @@ static int cnic_iscsi_nl_msg_recv(struct cnic_dev *dev, u32 msg_type,
                if (l5_cid >= MAX_CM_SK_TBL_SZ)
                        break;
 
+               rcu_read_lock();
+               if (!rcu_dereference(cp->ulp_ops[CNIC_ULP_L4])) {
+                       rc = -ENODEV;
+                       rcu_read_unlock();
+                       break;
+               }
                csk = &cp->csk_tbl[l5_cid];
                csk_hold(csk);
                if (cnic_in_use(csk)) {
@@ -269,6 +347,7 @@ static int cnic_iscsi_nl_msg_recv(struct cnic_dev *dev, u32 msg_type,
                                cnic_cm_set_pg(csk);
                }
                csk_put(csk);
+               rcu_read_unlock();
                rc = 0;
        }
        }
@@ -319,19 +398,32 @@ static int cnic_abort_prep(struct cnic_sock *csk)
        return 0;
 }
 
+static void cnic_uio_stop(void)
+{
+       struct cnic_dev *dev;
+
+       read_lock(&cnic_dev_lock);
+       list_for_each_entry(dev, &cnic_dev_list, list) {
+               struct cnic_local *cp = dev->cnic_priv;
+
+               if (cp->cnic_uinfo)
+                       cnic_send_nlmsg(cp, ISCSI_KEVENT_IF_DOWN, NULL);
+       }
+       read_unlock(&cnic_dev_lock);
+}
+
 int cnic_register_driver(int ulp_type, struct cnic_ulp_ops *ulp_ops)
 {
        struct cnic_dev *dev;
 
-       if (ulp_type >= MAX_CNIC_ULP_TYPE) {
-               printk(KERN_ERR PFX "cnic_register_driver: Bad type %d\n",
-                      ulp_type);
+       if (ulp_type < 0 || ulp_type >= MAX_CNIC_ULP_TYPE) {
+               pr_err("%s: Bad type %d\n", __func__, ulp_type);
                return -EINVAL;
        }
        mutex_lock(&cnic_lock);
        if (cnic_ulp_tbl[ulp_type]) {
-               printk(KERN_ERR PFX "cnic_register_driver: Type %d has already "
-                                   "been registered\n", ulp_type);
+               pr_err("%s: Type %d has already been registered\n",
+                      __func__, ulp_type);
                mutex_unlock(&cnic_lock);
                return -EBUSY;
        }
@@ -344,6 +436,7 @@ int cnic_register_driver(int ulp_type, struct cnic_ulp_ops *ulp_ops)
        }
        read_unlock(&cnic_dev_lock);
 
+       atomic_set(&ulp_ops->ref_count, 0);
        rcu_assign_pointer(cnic_ulp_tbl[ulp_type], ulp_ops);
        mutex_unlock(&cnic_lock);
 
@@ -365,16 +458,18 @@ int cnic_register_driver(int ulp_type, struct cnic_ulp_ops *ulp_ops)
 int cnic_unregister_driver(int ulp_type)
 {
        struct cnic_dev *dev;
+       struct cnic_ulp_ops *ulp_ops;
+       int i = 0;
 
-       if (ulp_type >= MAX_CNIC_ULP_TYPE) {
-               printk(KERN_ERR PFX "cnic_unregister_driver: Bad type %d\n",
-                      ulp_type);
+       if (ulp_type < 0 || ulp_type >= MAX_CNIC_ULP_TYPE) {
+               pr_err("%s: Bad type %d\n", __func__, ulp_type);
                return -EINVAL;
        }
        mutex_lock(&cnic_lock);
-       if (!cnic_ulp_tbl[ulp_type]) {
-               printk(KERN_ERR PFX "cnic_unregister_driver: Type %d has not "
-                                   "been registered\n", ulp_type);
+       ulp_ops = cnic_ulp_tbl[ulp_type];
+       if (!ulp_ops) {
+               pr_err("%s: Type %d has not been registered\n",
+                      __func__, ulp_type);
                goto out_unlock;
        }
        read_lock(&cnic_dev_lock);
@@ -382,18 +477,28 @@ int cnic_unregister_driver(int ulp_type)
                struct cnic_local *cp = dev->cnic_priv;
 
                if (rcu_dereference(cp->ulp_ops[ulp_type])) {
-                       printk(KERN_ERR PFX "cnic_unregister_driver: Type %d "
-                              "still has devices registered\n", ulp_type);
+                       pr_err("%s: Type %d still has devices registered\n",
+                              __func__, ulp_type);
                        read_unlock(&cnic_dev_lock);
                        goto out_unlock;
                }
        }
        read_unlock(&cnic_dev_lock);
 
+       if (ulp_type == CNIC_ULP_ISCSI)
+               cnic_uio_stop();
+
        rcu_assign_pointer(cnic_ulp_tbl[ulp_type], NULL);
 
        mutex_unlock(&cnic_lock);
        synchronize_rcu();
+       while ((atomic_read(&ulp_ops->ref_count) != 0) && (i < 20)) {
+               msleep(100);
+               i++;
+       }
+
+       if (atomic_read(&ulp_ops->ref_count) != 0)
+               netdev_warn(dev->netdev, "Failed waiting for ref count to go to zero\n");
        return 0;
 
 out_unlock:
@@ -410,21 +515,20 @@ static int cnic_register_device(struct cnic_dev *dev, int ulp_type,
        struct cnic_local *cp = dev->cnic_priv;
        struct cnic_ulp_ops *ulp_ops;
 
-       if (ulp_type >= MAX_CNIC_ULP_TYPE) {
-               printk(KERN_ERR PFX "cnic_register_device: Bad type %d\n",
-                      ulp_type);
+       if (ulp_type < 0 || ulp_type >= MAX_CNIC_ULP_TYPE) {
+               pr_err("%s: Bad type %d\n", __func__, ulp_type);
                return -EINVAL;
        }
        mutex_lock(&cnic_lock);
        if (cnic_ulp_tbl[ulp_type] == NULL) {
-               printk(KERN_ERR PFX "cnic_register_device: Driver with type %d "
-                                   "has not been registered\n", ulp_type);
+               pr_err("%s: Driver with type %d has not been registered\n",
+                      __func__, ulp_type);
                mutex_unlock(&cnic_lock);
                return -EAGAIN;
        }
        if (rcu_dereference(cp->ulp_ops[ulp_type])) {
-               printk(KERN_ERR PFX "cnic_register_device: Type %d has already "
-                      "been registered to this device\n", ulp_type);
+               pr_err("%s: Type %d has already been registered to this device\n",
+                      __func__, ulp_type);
                mutex_unlock(&cnic_lock);
                return -EBUSY;
        }
@@ -449,10 +553,10 @@ EXPORT_SYMBOL(cnic_register_driver);
 static int cnic_unregister_device(struct cnic_dev *dev, int ulp_type)
 {
        struct cnic_local *cp = dev->cnic_priv;
+       int i = 0;
 
-       if (ulp_type >= MAX_CNIC_ULP_TYPE) {
-               printk(KERN_ERR PFX "cnic_unregister_device: Bad type %d\n",
-                      ulp_type);
+       if (ulp_type < 0 || ulp_type >= MAX_CNIC_ULP_TYPE) {
+               pr_err("%s: Bad type %d\n", __func__, ulp_type);
                return -EINVAL;
        }
        mutex_lock(&cnic_lock);
@@ -460,8 +564,8 @@ static int cnic_unregister_device(struct cnic_dev *dev, int ulp_type)
                rcu_assign_pointer(cp->ulp_ops[ulp_type], NULL);
                cnic_put(dev);
        } else {
-               printk(KERN_ERR PFX "cnic_unregister_device: device not "
-                      "registered to this ulp type %d\n", ulp_type);
+               pr_err("%s: device not registered to this ulp type %d\n",
+                      __func__, ulp_type);
                mutex_unlock(&cnic_lock);
                return -EINVAL;
        }
@@ -469,6 +573,14 @@ static int cnic_unregister_device(struct cnic_dev *dev, int ulp_type)
 
        synchronize_rcu();
 
+       while (test_bit(ULP_F_CALL_PENDING, &cp->ulp_flags[ulp_type]) &&
+              i < 20) {
+               msleep(100);
+               i++;
+       }
+       if (test_bit(ULP_F_CALL_PENDING, &cp->ulp_flags[ulp_type]))
+               netdev_warn(dev->netdev, "Failed waiting for ULP up call to complete\n");
+
        return 0;
 }
 EXPORT_SYMBOL(cnic_unregister_driver);
@@ -557,14 +669,14 @@ static void cnic_free_dma(struct cnic_dev *dev, struct cnic_dma *dma)
 
        for (i = 0; i < dma->num_pages; i++) {
                if (dma->pg_arr[i]) {
-                       pci_free_consistent(dev->pcidev, BCM_PAGE_SIZE,
-                                           dma->pg_arr[i], dma->pg_map_arr[i]);
+                       dma_free_coherent(&dev->pcidev->dev, BCM_PAGE_SIZE,
+                                         dma->pg_arr[i], dma->pg_map_arr[i]);
                        dma->pg_arr[i] = NULL;
                }
        }
        if (dma->pgtbl) {
-               pci_free_consistent(dev->pcidev, dma->pgtbl_size,
-                                   dma->pgtbl, dma->pgtbl_map);
+               dma_free_coherent(&dev->pcidev->dev, dma->pgtbl_size,
+                                 dma->pgtbl, dma->pgtbl_map);
                dma->pgtbl = NULL;
        }
        kfree(dma->pg_arr);
@@ -586,6 +698,20 @@ static void cnic_setup_page_tbl(struct cnic_dev *dev, struct cnic_dma *dma)
        }
 }
 
+static void cnic_setup_page_tbl_le(struct cnic_dev *dev, struct cnic_dma *dma)
+{
+       int i;
+       u32 *page_table = dma->pgtbl;
+
+       for (i = 0; i < dma->num_pages; i++) {
+               /* Each entry needs to be in little endian format. */
+               *page_table = dma->pg_map_arr[i] & 0xffffffff;
+               page_table++;
+               *page_table = (u32) ((u64) dma->pg_map_arr[i] >> 32);
+               page_table++;
+       }
+}
+
 static int cnic_alloc_dma(struct cnic_dev *dev, struct cnic_dma *dma,
                          int pages, int use_pg_tbl)
 {
@@ -601,9 +727,10 @@ static int cnic_alloc_dma(struct cnic_dev *dev, struct cnic_dma *dma,
        dma->num_pages = pages;
 
        for (i = 0; i < pages; i++) {
-               dma->pg_arr[i] = pci_alloc_consistent(dev->pcidev,
-                                                     BCM_PAGE_SIZE,
-                                                     &dma->pg_map_arr[i]);
+               dma->pg_arr[i] = dma_alloc_coherent(&dev->pcidev->dev,
+                                                   BCM_PAGE_SIZE,
+                                                   &dma->pg_map_arr[i],
+                                                   GFP_ATOMIC);
                if (dma->pg_arr[i] == NULL)
                        goto error;
        }
@@ -612,8 +739,8 @@ static int cnic_alloc_dma(struct cnic_dev *dev, struct cnic_dma *dma,
 
        dma->pgtbl_size = ((pages * 8) + BCM_PAGE_SIZE - 1) &
                          ~(BCM_PAGE_SIZE - 1);
-       dma->pgtbl = pci_alloc_consistent(dev->pcidev, dma->pgtbl_size,
-                                         &dma->pgtbl_map);
+       dma->pgtbl = dma_alloc_coherent(&dev->pcidev->dev, dma->pgtbl_size,
+                                       &dma->pgtbl_map, GFP_ATOMIC);
        if (dma->pgtbl == NULL)
                goto error;
 
@@ -626,13 +753,27 @@ error:
        return -ENOMEM;
 }
 
+static void cnic_free_context(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       int i;
+
+       for (i = 0; i < cp->ctx_blks; i++) {
+               if (cp->ctx_arr[i].ctx) {
+                       dma_free_coherent(&dev->pcidev->dev, cp->ctx_blk_size,
+                                         cp->ctx_arr[i].ctx,
+                                         cp->ctx_arr[i].mapping);
+                       cp->ctx_arr[i].ctx = NULL;
+               }
+       }
+}
+
 static void cnic_free_resc(struct cnic_dev *dev)
 {
        struct cnic_local *cp = dev->cnic_priv;
        int i = 0;
 
        if (cp->cnic_uinfo) {
-               cnic_send_nlmsg(cp, ISCSI_KEVENT_IF_DOWN, NULL);
                while (cp->uio_dev != -1 && i < 15) {
                        msleep(100);
                        i++;
@@ -643,25 +784,18 @@ static void cnic_free_resc(struct cnic_dev *dev)
        }
 
        if (cp->l2_buf) {
-               pci_free_consistent(dev->pcidev, cp->l2_buf_size,
-                                   cp->l2_buf, cp->l2_buf_map);
+               dma_free_coherent(&dev->pcidev->dev, cp->l2_buf_size,
+                                 cp->l2_buf, cp->l2_buf_map);
                cp->l2_buf = NULL;
        }
 
        if (cp->l2_ring) {
-               pci_free_consistent(dev->pcidev, cp->l2_ring_size,
-                                   cp->l2_ring, cp->l2_ring_map);
+               dma_free_coherent(&dev->pcidev->dev, cp->l2_ring_size,
+                                 cp->l2_ring, cp->l2_ring_map);
                cp->l2_ring = NULL;
        }
 
-       for (i = 0; i < cp->ctx_blks; i++) {
-               if (cp->ctx_arr[i].ctx) {
-                       pci_free_consistent(dev->pcidev, cp->ctx_blk_size,
-                                           cp->ctx_arr[i].ctx,
-                                           cp->ctx_arr[i].mapping);
-                       cp->ctx_arr[i].ctx = NULL;
-               }
-       }
+       cnic_free_context(dev);
        kfree(cp->ctx_arr);
        cp->ctx_arr = NULL;
        cp->ctx_blks = 0;
@@ -669,6 +803,7 @@ static void cnic_free_resc(struct cnic_dev *dev)
        cnic_free_dma(dev, &cp->gbl_buf_info);
        cnic_free_dma(dev, &cp->conn_buf_info);
        cnic_free_dma(dev, &cp->kwq_info);
+       cnic_free_dma(dev, &cp->kwq_16_data_info);
        cnic_free_dma(dev, &cp->kcq_info);
        kfree(cp->iscsi_tbl);
        cp->iscsi_tbl = NULL;
@@ -717,8 +852,10 @@ static int cnic_alloc_context(struct cnic_dev *dev)
 
                for (i = 0; i < cp->ctx_blks; i++) {
                        cp->ctx_arr[i].ctx =
-                               pci_alloc_consistent(dev->pcidev, BCM_PAGE_SIZE,
-                                                    &cp->ctx_arr[i].mapping);
+                               dma_alloc_coherent(&dev->pcidev->dev,
+                                                  BCM_PAGE_SIZE,
+                                                  &cp->ctx_arr[i].mapping,
+                                                  GFP_KERNEL);
                        if (cp->ctx_arr[i].ctx == NULL)
                                return -ENOMEM;
                }
@@ -726,53 +863,59 @@ static int cnic_alloc_context(struct cnic_dev *dev)
        return 0;
 }
 
-static int cnic_alloc_bnx2_resc(struct cnic_dev *dev)
+static int cnic_alloc_l2_rings(struct cnic_dev *dev, int pages)
 {
        struct cnic_local *cp = dev->cnic_priv;
-       struct uio_info *uinfo;
-       int ret;
-
-       ret = cnic_alloc_dma(dev, &cp->kwq_info, KWQ_PAGE_CNT, 1);
-       if (ret)
-               goto error;
-       cp->kwq = (struct kwqe **) cp->kwq_info.pg_arr;
-
-       ret = cnic_alloc_dma(dev, &cp->kcq_info, KCQ_PAGE_CNT, 1);
-       if (ret)
-               goto error;
-       cp->kcq = (struct kcqe **) cp->kcq_info.pg_arr;
-
-       ret = cnic_alloc_context(dev);
-       if (ret)
-               goto error;
 
-       cp->l2_ring_size = 2 * BCM_PAGE_SIZE;
-       cp->l2_ring = pci_alloc_consistent(dev->pcidev, cp->l2_ring_size,
-                                          &cp->l2_ring_map);
+       cp->l2_ring_size = pages * BCM_PAGE_SIZE;
+       cp->l2_ring = dma_alloc_coherent(&dev->pcidev->dev, cp->l2_ring_size,
+                                        &cp->l2_ring_map,
+                                        GFP_KERNEL | __GFP_COMP);
        if (!cp->l2_ring)
-               goto error;
+               return -ENOMEM;
 
        cp->l2_buf_size = (cp->l2_rx_ring_size + 1) * cp->l2_single_buf_size;
        cp->l2_buf_size = PAGE_ALIGN(cp->l2_buf_size);
-       cp->l2_buf = pci_alloc_consistent(dev->pcidev, cp->l2_buf_size,
-                                          &cp->l2_buf_map);
+       cp->l2_buf = dma_alloc_coherent(&dev->pcidev->dev, cp->l2_buf_size,
+                                       &cp->l2_buf_map,
+                                       GFP_KERNEL | __GFP_COMP);
        if (!cp->l2_buf)
-               goto error;
+               return -ENOMEM;
+
+       return 0;
+}
+
+static int cnic_alloc_uio(struct cnic_dev *dev) {
+       struct cnic_local *cp = dev->cnic_priv;
+       struct uio_info *uinfo;
+       int ret;
 
        uinfo = kzalloc(sizeof(*uinfo), GFP_ATOMIC);
        if (!uinfo)
-               goto error;
+               return -ENOMEM;
 
        uinfo->mem[0].addr = dev->netdev->base_addr;
        uinfo->mem[0].internal_addr = dev->regview;
        uinfo->mem[0].size = dev->netdev->mem_end - dev->netdev->mem_start;
        uinfo->mem[0].memtype = UIO_MEM_PHYS;
 
-       uinfo->mem[1].addr = (unsigned long) cp->status_blk & PAGE_MASK;
-       if (cp->ethdev->drv_state & CNIC_DRV_STATE_USING_MSIX)
-               uinfo->mem[1].size = BNX2_SBLK_MSIX_ALIGN_SIZE * 9;
-       else
-               uinfo->mem[1].size = BNX2_SBLK_MSIX_ALIGN_SIZE;
+       if (test_bit(CNIC_F_BNX2_CLASS, &dev->flags)) {
+               uinfo->mem[1].addr = (unsigned long) cp->status_blk.gen &
+                       PAGE_MASK;
+               if (cp->ethdev->drv_state & CNIC_DRV_STATE_USING_MSIX)
+                       uinfo->mem[1].size = BNX2_SBLK_MSIX_ALIGN_SIZE * 9;
+               else
+                       uinfo->mem[1].size = BNX2_SBLK_MSIX_ALIGN_SIZE;
+
+               uinfo->name = "bnx2_cnic";
+       } else if (test_bit(CNIC_F_BNX2X_CLASS, &dev->flags)) {
+               uinfo->mem[1].addr = (unsigned long) cp->bnx2x_def_status_blk &
+                       PAGE_MASK;
+               uinfo->mem[1].size = sizeof(struct host_def_status_block);
+
+               uinfo->name = "bnx2x_cnic";
+       }
+
        uinfo->mem[1].memtype = UIO_MEM_LOGICAL;
 
        uinfo->mem[2].addr = (unsigned long) cp->l2_ring;
@@ -783,7 +926,6 @@ static int cnic_alloc_bnx2_resc(struct cnic_dev *dev)
        uinfo->mem[3].size = cp->l2_buf_size;
        uinfo->mem[3].memtype = UIO_MEM_LOGICAL;
 
-       uinfo->name = "bnx2_cnic";
        uinfo->version = CNIC_MODULE_VERSION;
        uinfo->irq = UIO_IRQ_CUSTOM;
 
@@ -795,10 +937,39 @@ static int cnic_alloc_bnx2_resc(struct cnic_dev *dev)
        ret = uio_register_device(&dev->pcidev->dev, uinfo);
        if (ret) {
                kfree(uinfo);
-               goto error;
+               return ret;
        }
 
        cp->cnic_uinfo = uinfo;
+       return 0;
+}
+
+static int cnic_alloc_bnx2_resc(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       int ret;
+
+       ret = cnic_alloc_dma(dev, &cp->kwq_info, KWQ_PAGE_CNT, 1);
+       if (ret)
+               goto error;
+       cp->kwq = (struct kwqe **) cp->kwq_info.pg_arr;
+
+       ret = cnic_alloc_dma(dev, &cp->kcq_info, KCQ_PAGE_CNT, 1);
+       if (ret)
+               goto error;
+       cp->kcq = (struct kcqe **) cp->kcq_info.pg_arr;
+
+       ret = cnic_alloc_context(dev);
+       if (ret)
+               goto error;
+
+       ret = cnic_alloc_l2_rings(dev, 2);
+       if (ret)
+               goto error;
+
+       ret = cnic_alloc_uio(dev);
+       if (ret)
+               goto error;
 
        return 0;
 
@@ -807,6 +978,153 @@ error:
        return ret;
 }
 
+static int cnic_alloc_bnx2x_context(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct cnic_eth_dev *ethdev = cp->ethdev;
+       int ctx_blk_size = cp->ethdev->ctx_blk_size;
+       int total_mem, blks, i, cid_space;
+
+       if (BNX2X_ISCSI_START_CID < ethdev->starting_cid)
+               return -EINVAL;
+
+       cid_space = MAX_ISCSI_TBL_SZ +
+                   (BNX2X_ISCSI_START_CID - ethdev->starting_cid);
+
+       total_mem = BNX2X_CONTEXT_MEM_SIZE * cid_space;
+       blks = total_mem / ctx_blk_size;
+       if (total_mem % ctx_blk_size)
+               blks++;
+
+       if (blks > cp->ethdev->ctx_tbl_len)
+               return -ENOMEM;
+
+       cp->ctx_arr = kzalloc(blks * sizeof(struct cnic_ctx), GFP_KERNEL);
+       if (cp->ctx_arr == NULL)
+               return -ENOMEM;
+
+       cp->ctx_blks = blks;
+       cp->ctx_blk_size = ctx_blk_size;
+       if (BNX2X_CHIP_IS_E1H(cp->chip_id))
+               cp->ctx_align = 0;
+       else
+               cp->ctx_align = ctx_blk_size;
+
+       cp->cids_per_blk = ctx_blk_size / BNX2X_CONTEXT_MEM_SIZE;
+
+       for (i = 0; i < blks; i++) {
+               cp->ctx_arr[i].ctx =
+                       dma_alloc_coherent(&dev->pcidev->dev, cp->ctx_blk_size,
+                                          &cp->ctx_arr[i].mapping,
+                                          GFP_KERNEL);
+               if (cp->ctx_arr[i].ctx == NULL)
+                       return -ENOMEM;
+
+               if (cp->ctx_align && cp->ctx_blk_size == ctx_blk_size) {
+                       if (cp->ctx_arr[i].mapping & (cp->ctx_align - 1)) {
+                               cnic_free_context(dev);
+                               cp->ctx_blk_size += cp->ctx_align;
+                               i = -1;
+                               continue;
+                       }
+               }
+       }
+       return 0;
+}
+
+static int cnic_alloc_bnx2x_resc(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       int i, j, n, ret, pages;
+       struct cnic_dma *kwq_16_dma = &cp->kwq_16_data_info;
+
+       cp->iscsi_tbl = kzalloc(sizeof(struct cnic_iscsi) * MAX_ISCSI_TBL_SZ,
+                               GFP_KERNEL);
+       if (!cp->iscsi_tbl)
+               goto error;
+
+       cp->ctx_tbl = kzalloc(sizeof(struct cnic_context) *
+                                 MAX_CNIC_L5_CONTEXT, GFP_KERNEL);
+       if (!cp->ctx_tbl)
+               goto error;
+
+       for (i = 0; i < MAX_ISCSI_TBL_SZ; i++) {
+               cp->ctx_tbl[i].proto.iscsi = &cp->iscsi_tbl[i];
+               cp->ctx_tbl[i].ulp_proto_id = CNIC_ULP_ISCSI;
+       }
+
+       pages = PAGE_ALIGN(MAX_CNIC_L5_CONTEXT * CNIC_KWQ16_DATA_SIZE) /
+               PAGE_SIZE;
+
+       ret = cnic_alloc_dma(dev, kwq_16_dma, pages, 0);
+       if (ret)
+               return -ENOMEM;
+
+       n = PAGE_SIZE / CNIC_KWQ16_DATA_SIZE;
+       for (i = 0, j = 0; i < MAX_ISCSI_TBL_SZ; i++) {
+               long off = CNIC_KWQ16_DATA_SIZE * (i % n);
+
+               cp->ctx_tbl[i].kwqe_data = kwq_16_dma->pg_arr[j] + off;
+               cp->ctx_tbl[i].kwqe_data_mapping = kwq_16_dma->pg_map_arr[j] +
+                                                  off;
+
+               if ((i % n) == (n - 1))
+                       j++;
+       }
+
+       ret = cnic_alloc_dma(dev, &cp->kcq_info, KCQ_PAGE_CNT, 0);
+       if (ret)
+               goto error;
+       cp->kcq = (struct kcqe **) cp->kcq_info.pg_arr;
+
+       for (i = 0; i < KCQ_PAGE_CNT; i++) {
+               struct bnx2x_bd_chain_next *next =
+                       (struct bnx2x_bd_chain_next *)
+                       &cp->kcq[i][MAX_KCQE_CNT];
+               int j = i + 1;
+
+               if (j >= KCQ_PAGE_CNT)
+                       j = 0;
+               next->addr_hi = (u64) cp->kcq_info.pg_map_arr[j] >> 32;
+               next->addr_lo = cp->kcq_info.pg_map_arr[j] & 0xffffffff;
+       }
+
+       pages = PAGE_ALIGN(BNX2X_ISCSI_NUM_CONNECTIONS *
+                          BNX2X_ISCSI_CONN_BUF_SIZE) / PAGE_SIZE;
+       ret = cnic_alloc_dma(dev, &cp->conn_buf_info, pages, 1);
+       if (ret)
+               goto error;
+
+       pages = PAGE_ALIGN(BNX2X_ISCSI_GLB_BUF_SIZE) / PAGE_SIZE;
+       ret = cnic_alloc_dma(dev, &cp->gbl_buf_info, pages, 0);
+       if (ret)
+               goto error;
+
+       ret = cnic_alloc_bnx2x_context(dev);
+       if (ret)
+               goto error;
+
+       cp->bnx2x_def_status_blk = cp->ethdev->irq_arr[1].status_blk;
+
+       memset(cp->status_blk.bnx2x, 0, sizeof(*cp->status_blk.bnx2x));
+
+       cp->l2_rx_ring_size = 15;
+
+       ret = cnic_alloc_l2_rings(dev, 4);
+       if (ret)
+               goto error;
+
+       ret = cnic_alloc_uio(dev);
+       if (ret)
+               goto error;
+
+       return 0;
+
+error:
+       cnic_free_resc(dev);
+       return -ENOMEM;
+}
+
 static inline u32 cnic_kwq_avail(struct cnic_local *cp)
 {
        return cp->max_kwq_idx -
@@ -848,20 +1166,893 @@ static int cnic_submit_bnx2_kwqes(struct cnic_dev *dev, struct kwqe *wqes[],
        return 0;
 }
 
-static void service_kcqes(struct cnic_dev *dev, int num_cqes)
+static void *cnic_get_kwqe_16_data(struct cnic_local *cp, u32 l5_cid,
+                                  union l5cm_specific_data *l5_data)
+{
+       struct cnic_context *ctx = &cp->ctx_tbl[l5_cid];
+       dma_addr_t map;
+
+       map = ctx->kwqe_data_mapping;
+       l5_data->phy_address.lo = (u64) map & 0xffffffff;
+       l5_data->phy_address.hi = (u64) map >> 32;
+       return ctx->kwqe_data;
+}
+
+static int cnic_submit_kwqe_16(struct cnic_dev *dev, u32 cmd, u32 cid,
+                               u32 type, union l5cm_specific_data *l5_data)
 {
        struct cnic_local *cp = dev->cnic_priv;
-       int i, j;
+       struct l5cm_spe kwqe;
+       struct kwqe_16 *kwq[1];
+       int ret;
 
-       i = 0;
-       j = 1;
-       while (num_cqes) {
-               struct cnic_ulp_ops *ulp_ops;
-               int ulp_type;
-               u32 kcqe_op_flag = cp->completed_kcq[i]->kcqe_op_flag;
-               u32 kcqe_layer = kcqe_op_flag & KCQE_FLAGS_LAYER_MASK;
+       kwqe.hdr.conn_and_cmd_data =
+               cpu_to_le32(((cmd << SPE_HDR_CMD_ID_SHIFT) |
+                            BNX2X_HW_CID(cid, cp->func)));
+       kwqe.hdr.type = cpu_to_le16(type);
+       kwqe.hdr.reserved = 0;
+       kwqe.data.phy_address.lo = cpu_to_le32(l5_data->phy_address.lo);
+       kwqe.data.phy_address.hi = cpu_to_le32(l5_data->phy_address.hi);
 
-               if (unlikely(kcqe_op_flag & KCQE_RAMROD_COMPLETION))
+       kwq[0] = (struct kwqe_16 *) &kwqe;
+
+       spin_lock_bh(&cp->cnic_ulp_lock);
+       ret = cp->ethdev->drv_submit_kwqes_16(dev->netdev, kwq, 1);
+       spin_unlock_bh(&cp->cnic_ulp_lock);
+
+       if (ret == 1)
+               return 0;
+
+       return -EBUSY;
+}
+
+static void cnic_reply_bnx2x_kcqes(struct cnic_dev *dev, int ulp_type,
+                                  struct kcqe *cqes[], u32 num_cqes)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct cnic_ulp_ops *ulp_ops;
+
+       rcu_read_lock();
+       ulp_ops = rcu_dereference(cp->ulp_ops[ulp_type]);
+       if (likely(ulp_ops)) {
+               ulp_ops->indicate_kcqes(cp->ulp_handle[ulp_type],
+                                         cqes, num_cqes);
+       }
+       rcu_read_unlock();
+}
+
+static int cnic_bnx2x_iscsi_init1(struct cnic_dev *dev, struct kwqe *kwqe)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct iscsi_kwqe_init1 *req1 = (struct iscsi_kwqe_init1 *) kwqe;
+       int func = cp->func, pages;
+       int hq_bds;
+
+       cp->num_iscsi_tasks = req1->num_tasks_per_conn;
+       cp->num_ccells = req1->num_ccells_per_conn;
+       cp->task_array_size = BNX2X_ISCSI_TASK_CONTEXT_SIZE *
+                             cp->num_iscsi_tasks;
+       cp->r2tq_size = cp->num_iscsi_tasks * BNX2X_ISCSI_MAX_PENDING_R2TS *
+                       BNX2X_ISCSI_R2TQE_SIZE;
+       cp->hq_size = cp->num_ccells * BNX2X_ISCSI_HQ_BD_SIZE;
+       pages = PAGE_ALIGN(cp->hq_size) / PAGE_SIZE;
+       hq_bds = pages * (PAGE_SIZE / BNX2X_ISCSI_HQ_BD_SIZE);
+       cp->num_cqs = req1->num_cqs;
+
+       if (!dev->max_iscsi_conn)
+               return 0;
+
+       /* init Tstorm RAM */
+       CNIC_WR16(dev, BAR_TSTRORM_INTMEM + TSTORM_ISCSI_RQ_SIZE_OFFSET(func),
+                 req1->rq_num_wqes);
+       CNIC_WR16(dev, BAR_TSTRORM_INTMEM + TSTORM_ISCSI_PAGE_SIZE_OFFSET(func),
+                 PAGE_SIZE);
+       CNIC_WR8(dev, BAR_TSTRORM_INTMEM +
+                TSTORM_ISCSI_PAGE_SIZE_LOG_OFFSET(func), PAGE_SHIFT);
+       CNIC_WR16(dev, BAR_TSTRORM_INTMEM +
+                 TSTORM_ISCSI_NUM_OF_TASKS_OFFSET(func),
+                 req1->num_tasks_per_conn);
+
+       /* init Ustorm RAM */
+       CNIC_WR16(dev, BAR_USTRORM_INTMEM +
+                 USTORM_ISCSI_RQ_BUFFER_SIZE_OFFSET(func),
+                 req1->rq_buffer_size);
+       CNIC_WR16(dev, BAR_USTRORM_INTMEM + USTORM_ISCSI_PAGE_SIZE_OFFSET(func),
+                 PAGE_SIZE);
+       CNIC_WR8(dev, BAR_USTRORM_INTMEM +
+                USTORM_ISCSI_PAGE_SIZE_LOG_OFFSET(func), PAGE_SHIFT);
+       CNIC_WR16(dev, BAR_USTRORM_INTMEM +
+                 USTORM_ISCSI_NUM_OF_TASKS_OFFSET(func),
+                 req1->num_tasks_per_conn);
+       CNIC_WR16(dev, BAR_USTRORM_INTMEM + USTORM_ISCSI_RQ_SIZE_OFFSET(func),
+                 req1->rq_num_wqes);
+       CNIC_WR16(dev, BAR_USTRORM_INTMEM + USTORM_ISCSI_CQ_SIZE_OFFSET(func),
+                 req1->cq_num_wqes);
+       CNIC_WR16(dev, BAR_USTRORM_INTMEM + USTORM_ISCSI_R2TQ_SIZE_OFFSET(func),
+                 cp->num_iscsi_tasks * BNX2X_ISCSI_MAX_PENDING_R2TS);
+
+       /* init Xstorm RAM */
+       CNIC_WR16(dev, BAR_XSTRORM_INTMEM + XSTORM_ISCSI_PAGE_SIZE_OFFSET(func),
+                 PAGE_SIZE);
+       CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
+                XSTORM_ISCSI_PAGE_SIZE_LOG_OFFSET(func), PAGE_SHIFT);
+       CNIC_WR16(dev, BAR_XSTRORM_INTMEM +
+                 XSTORM_ISCSI_NUM_OF_TASKS_OFFSET(func),
+                 req1->num_tasks_per_conn);
+       CNIC_WR16(dev, BAR_XSTRORM_INTMEM + XSTORM_ISCSI_HQ_SIZE_OFFSET(func),
+                 hq_bds);
+       CNIC_WR16(dev, BAR_XSTRORM_INTMEM + XSTORM_ISCSI_SQ_SIZE_OFFSET(func),
+                 req1->num_tasks_per_conn);
+       CNIC_WR16(dev, BAR_XSTRORM_INTMEM + XSTORM_ISCSI_R2TQ_SIZE_OFFSET(func),
+                 cp->num_iscsi_tasks * BNX2X_ISCSI_MAX_PENDING_R2TS);
+
+       /* init Cstorm RAM */
+       CNIC_WR16(dev, BAR_CSTRORM_INTMEM + CSTORM_ISCSI_PAGE_SIZE_OFFSET(func),
+                 PAGE_SIZE);
+       CNIC_WR8(dev, BAR_CSTRORM_INTMEM +
+                CSTORM_ISCSI_PAGE_SIZE_LOG_OFFSET(func), PAGE_SHIFT);
+       CNIC_WR16(dev, BAR_CSTRORM_INTMEM +
+                 CSTORM_ISCSI_NUM_OF_TASKS_OFFSET(func),
+                 req1->num_tasks_per_conn);
+       CNIC_WR16(dev, BAR_CSTRORM_INTMEM + CSTORM_ISCSI_CQ_SIZE_OFFSET(func),
+                 req1->cq_num_wqes);
+       CNIC_WR16(dev, BAR_CSTRORM_INTMEM + CSTORM_ISCSI_HQ_SIZE_OFFSET(func),
+                 hq_bds);
+
+       return 0;
+}
+
+static int cnic_bnx2x_iscsi_init2(struct cnic_dev *dev, struct kwqe *kwqe)
+{
+       struct iscsi_kwqe_init2 *req2 = (struct iscsi_kwqe_init2 *) kwqe;
+       struct cnic_local *cp = dev->cnic_priv;
+       int func = cp->func;
+       struct iscsi_kcqe kcqe;
+       struct kcqe *cqes[1];
+
+       memset(&kcqe, 0, sizeof(kcqe));
+       if (!dev->max_iscsi_conn) {
+               kcqe.completion_status =
+                       ISCSI_KCQE_COMPLETION_STATUS_ISCSI_NOT_SUPPORTED;
+               goto done;
+       }
+
+       CNIC_WR(dev, BAR_TSTRORM_INTMEM +
+               TSTORM_ISCSI_ERROR_BITMAP_OFFSET(func), req2->error_bit_map[0]);
+       CNIC_WR(dev, BAR_TSTRORM_INTMEM +
+               TSTORM_ISCSI_ERROR_BITMAP_OFFSET(func) + 4,
+               req2->error_bit_map[1]);
+
+       CNIC_WR16(dev, BAR_USTRORM_INTMEM +
+                 USTORM_ISCSI_CQ_SQN_SIZE_OFFSET(func), req2->max_cq_sqn);
+       CNIC_WR(dev, BAR_USTRORM_INTMEM +
+               USTORM_ISCSI_ERROR_BITMAP_OFFSET(func), req2->error_bit_map[0]);
+       CNIC_WR(dev, BAR_USTRORM_INTMEM +
+               USTORM_ISCSI_ERROR_BITMAP_OFFSET(func) + 4,
+               req2->error_bit_map[1]);
+
+       CNIC_WR16(dev, BAR_CSTRORM_INTMEM +
+                 CSTORM_ISCSI_CQ_SQN_SIZE_OFFSET(func), req2->max_cq_sqn);
+
+       kcqe.completion_status = ISCSI_KCQE_COMPLETION_STATUS_SUCCESS;
+
+done:
+       kcqe.op_code = ISCSI_KCQE_OPCODE_INIT;
+       cqes[0] = (struct kcqe *) &kcqe;
+       cnic_reply_bnx2x_kcqes(dev, CNIC_ULP_ISCSI, cqes, 1);
+
+       return 0;
+}
+
+static void cnic_free_bnx2x_conn_resc(struct cnic_dev *dev, u32 l5_cid)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct cnic_context *ctx = &cp->ctx_tbl[l5_cid];
+
+       if (ctx->ulp_proto_id == CNIC_ULP_ISCSI) {
+               struct cnic_iscsi *iscsi = ctx->proto.iscsi;
+
+               cnic_free_dma(dev, &iscsi->hq_info);
+               cnic_free_dma(dev, &iscsi->r2tq_info);
+               cnic_free_dma(dev, &iscsi->task_array_info);
+       }
+       cnic_free_id(&cp->cid_tbl, ctx->cid);
+       ctx->cid = 0;
+}
+
+static int cnic_alloc_bnx2x_conn_resc(struct cnic_dev *dev, u32 l5_cid)
+{
+       u32 cid;
+       int ret, pages;
+       struct cnic_local *cp = dev->cnic_priv;
+       struct cnic_context *ctx = &cp->ctx_tbl[l5_cid];
+       struct cnic_iscsi *iscsi = ctx->proto.iscsi;
+
+       cid = cnic_alloc_new_id(&cp->cid_tbl);
+       if (cid == -1) {
+               ret = -ENOMEM;
+               goto error;
+       }
+
+       ctx->cid = cid;
+       pages = PAGE_ALIGN(cp->task_array_size) / PAGE_SIZE;
+
+       ret = cnic_alloc_dma(dev, &iscsi->task_array_info, pages, 1);
+       if (ret)
+               goto error;
+
+       pages = PAGE_ALIGN(cp->r2tq_size) / PAGE_SIZE;
+       ret = cnic_alloc_dma(dev, &iscsi->r2tq_info, pages, 1);
+       if (ret)
+               goto error;
+
+       pages = PAGE_ALIGN(cp->hq_size) / PAGE_SIZE;
+       ret = cnic_alloc_dma(dev, &iscsi->hq_info, pages, 1);
+       if (ret)
+               goto error;
+
+       return 0;
+
+error:
+       cnic_free_bnx2x_conn_resc(dev, l5_cid);
+       return ret;
+}
+
+static void *cnic_get_bnx2x_ctx(struct cnic_dev *dev, u32 cid, int init,
+                               struct regpair *ctx_addr)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct cnic_eth_dev *ethdev = cp->ethdev;
+       int blk = (cid - ethdev->starting_cid) / cp->cids_per_blk;
+       int off = (cid - ethdev->starting_cid) % cp->cids_per_blk;
+       unsigned long align_off = 0;
+       dma_addr_t ctx_map;
+       void *ctx;
+
+       if (cp->ctx_align) {
+               unsigned long mask = cp->ctx_align - 1;
+
+               if (cp->ctx_arr[blk].mapping & mask)
+                       align_off = cp->ctx_align -
+                                   (cp->ctx_arr[blk].mapping & mask);
+       }
+       ctx_map = cp->ctx_arr[blk].mapping + align_off +
+               (off * BNX2X_CONTEXT_MEM_SIZE);
+       ctx = cp->ctx_arr[blk].ctx + align_off +
+             (off * BNX2X_CONTEXT_MEM_SIZE);
+       if (init)
+               memset(ctx, 0, BNX2X_CONTEXT_MEM_SIZE);
+
+       ctx_addr->lo = ctx_map & 0xffffffff;
+       ctx_addr->hi = (u64) ctx_map >> 32;
+       return ctx;
+}
+
+static int cnic_setup_bnx2x_ctx(struct cnic_dev *dev, struct kwqe *wqes[],
+                               u32 num)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct iscsi_kwqe_conn_offload1 *req1 =
+                       (struct iscsi_kwqe_conn_offload1 *) wqes[0];
+       struct iscsi_kwqe_conn_offload2 *req2 =
+                       (struct iscsi_kwqe_conn_offload2 *) wqes[1];
+       struct iscsi_kwqe_conn_offload3 *req3;
+       struct cnic_context *ctx = &cp->ctx_tbl[req1->iscsi_conn_id];
+       struct cnic_iscsi *iscsi = ctx->proto.iscsi;
+       u32 cid = ctx->cid;
+       u32 hw_cid = BNX2X_HW_CID(cid, cp->func);
+       struct iscsi_context *ictx;
+       struct regpair context_addr;
+       int i, j, n = 2, n_max;
+
+       ctx->ctx_flags = 0;
+       if (!req2->num_additional_wqes)
+               return -EINVAL;
+
+       n_max = req2->num_additional_wqes + 2;
+
+       ictx = cnic_get_bnx2x_ctx(dev, cid, 1, &context_addr);
+       if (ictx == NULL)
+               return -ENOMEM;
+
+       req3 = (struct iscsi_kwqe_conn_offload3 *) wqes[n++];
+
+       ictx->xstorm_ag_context.hq_prod = 1;
+
+       ictx->xstorm_st_context.iscsi.first_burst_length =
+               ISCSI_DEF_FIRST_BURST_LEN;
+       ictx->xstorm_st_context.iscsi.max_send_pdu_length =
+               ISCSI_DEF_MAX_RECV_SEG_LEN;
+       ictx->xstorm_st_context.iscsi.sq_pbl_base.lo =
+               req1->sq_page_table_addr_lo;
+       ictx->xstorm_st_context.iscsi.sq_pbl_base.hi =
+               req1->sq_page_table_addr_hi;
+       ictx->xstorm_st_context.iscsi.sq_curr_pbe.lo = req2->sq_first_pte.hi;
+       ictx->xstorm_st_context.iscsi.sq_curr_pbe.hi = req2->sq_first_pte.lo;
+       ictx->xstorm_st_context.iscsi.hq_pbl_base.lo =
+               iscsi->hq_info.pgtbl_map & 0xffffffff;
+       ictx->xstorm_st_context.iscsi.hq_pbl_base.hi =
+               (u64) iscsi->hq_info.pgtbl_map >> 32;
+       ictx->xstorm_st_context.iscsi.hq_curr_pbe_base.lo =
+               iscsi->hq_info.pgtbl[0];
+       ictx->xstorm_st_context.iscsi.hq_curr_pbe_base.hi =
+               iscsi->hq_info.pgtbl[1];
+       ictx->xstorm_st_context.iscsi.r2tq_pbl_base.lo =
+               iscsi->r2tq_info.pgtbl_map & 0xffffffff;
+       ictx->xstorm_st_context.iscsi.r2tq_pbl_base.hi =
+               (u64) iscsi->r2tq_info.pgtbl_map >> 32;
+       ictx->xstorm_st_context.iscsi.r2tq_curr_pbe_base.lo =
+               iscsi->r2tq_info.pgtbl[0];
+       ictx->xstorm_st_context.iscsi.r2tq_curr_pbe_base.hi =
+               iscsi->r2tq_info.pgtbl[1];
+       ictx->xstorm_st_context.iscsi.task_pbl_base.lo =
+               iscsi->task_array_info.pgtbl_map & 0xffffffff;
+       ictx->xstorm_st_context.iscsi.task_pbl_base.hi =
+               (u64) iscsi->task_array_info.pgtbl_map >> 32;
+       ictx->xstorm_st_context.iscsi.task_pbl_cache_idx =
+               BNX2X_ISCSI_PBL_NOT_CACHED;
+       ictx->xstorm_st_context.iscsi.flags.flags |=
+               XSTORM_ISCSI_CONTEXT_FLAGS_B_IMMEDIATE_DATA;
+       ictx->xstorm_st_context.iscsi.flags.flags |=
+               XSTORM_ISCSI_CONTEXT_FLAGS_B_INITIAL_R2T;
+
+       ictx->tstorm_st_context.iscsi.hdr_bytes_2_fetch = ISCSI_HEADER_SIZE;
+       /* TSTORM requires the base address of RQ DB & not PTE */
+       ictx->tstorm_st_context.iscsi.rq_db_phy_addr.lo =
+               req2->rq_page_table_addr_lo & PAGE_MASK;
+       ictx->tstorm_st_context.iscsi.rq_db_phy_addr.hi =
+               req2->rq_page_table_addr_hi;
+       ictx->tstorm_st_context.iscsi.iscsi_conn_id = req1->iscsi_conn_id;
+       ictx->tstorm_st_context.tcp.cwnd = 0x5A8;
+       ictx->tstorm_st_context.tcp.flags2 |=
+               TSTORM_TCP_ST_CONTEXT_SECTION_DA_EN;
+
+       ictx->timers_context.flags |= ISCSI_TIMERS_BLOCK_CONTEXT_CONN_VALID_FLG;
+
+       ictx->ustorm_st_context.ring.rq.pbl_base.lo =
+               req2->rq_page_table_addr_lo;
+       ictx->ustorm_st_context.ring.rq.pbl_base.hi =
+               req2->rq_page_table_addr_hi;
+       ictx->ustorm_st_context.ring.rq.curr_pbe.lo = req3->qp_first_pte[0].hi;
+       ictx->ustorm_st_context.ring.rq.curr_pbe.hi = req3->qp_first_pte[0].lo;
+       ictx->ustorm_st_context.ring.r2tq.pbl_base.lo =
+               iscsi->r2tq_info.pgtbl_map & 0xffffffff;
+       ictx->ustorm_st_context.ring.r2tq.pbl_base.hi =
+               (u64) iscsi->r2tq_info.pgtbl_map >> 32;
+       ictx->ustorm_st_context.ring.r2tq.curr_pbe.lo =
+               iscsi->r2tq_info.pgtbl[0];
+       ictx->ustorm_st_context.ring.r2tq.curr_pbe.hi =
+               iscsi->r2tq_info.pgtbl[1];
+       ictx->ustorm_st_context.ring.cq_pbl_base.lo =
+               req1->cq_page_table_addr_lo;
+       ictx->ustorm_st_context.ring.cq_pbl_base.hi =
+               req1->cq_page_table_addr_hi;
+       ictx->ustorm_st_context.ring.cq[0].cq_sn = ISCSI_INITIAL_SN;
+       ictx->ustorm_st_context.ring.cq[0].curr_pbe.lo = req2->cq_first_pte.hi;
+       ictx->ustorm_st_context.ring.cq[0].curr_pbe.hi = req2->cq_first_pte.lo;
+       ictx->ustorm_st_context.task_pbe_cache_index =
+               BNX2X_ISCSI_PBL_NOT_CACHED;
+       ictx->ustorm_st_context.task_pdu_cache_index =
+               BNX2X_ISCSI_PDU_HEADER_NOT_CACHED;
+
+       for (i = 1, j = 1; i < cp->num_cqs; i++, j++) {
+               if (j == 3) {
+                       if (n >= n_max)
+                               break;
+                       req3 = (struct iscsi_kwqe_conn_offload3 *) wqes[n++];
+                       j = 0;
+               }
+               ictx->ustorm_st_context.ring.cq[i].cq_sn = ISCSI_INITIAL_SN;
+               ictx->ustorm_st_context.ring.cq[i].curr_pbe.lo =
+                       req3->qp_first_pte[j].hi;
+               ictx->ustorm_st_context.ring.cq[i].curr_pbe.hi =
+                       req3->qp_first_pte[j].lo;
+       }
+
+       ictx->ustorm_st_context.task_pbl_base.lo =
+               iscsi->task_array_info.pgtbl_map & 0xffffffff;
+       ictx->ustorm_st_context.task_pbl_base.hi =
+               (u64) iscsi->task_array_info.pgtbl_map >> 32;
+       ictx->ustorm_st_context.tce_phy_addr.lo =
+               iscsi->task_array_info.pgtbl[0];
+       ictx->ustorm_st_context.tce_phy_addr.hi =
+               iscsi->task_array_info.pgtbl[1];
+       ictx->ustorm_st_context.iscsi_conn_id = req1->iscsi_conn_id;
+       ictx->ustorm_st_context.num_cqs = cp->num_cqs;
+       ictx->ustorm_st_context.negotiated_rx |= ISCSI_DEF_MAX_RECV_SEG_LEN;
+       ictx->ustorm_st_context.negotiated_rx_and_flags |=
+               ISCSI_DEF_MAX_BURST_LEN;
+       ictx->ustorm_st_context.negotiated_rx |=
+               ISCSI_DEFAULT_MAX_OUTSTANDING_R2T <<
+               USTORM_ISCSI_ST_CONTEXT_MAX_OUTSTANDING_R2TS_SHIFT;
+
+       ictx->cstorm_st_context.hq_pbl_base.lo =
+               iscsi->hq_info.pgtbl_map & 0xffffffff;
+       ictx->cstorm_st_context.hq_pbl_base.hi =
+               (u64) iscsi->hq_info.pgtbl_map >> 32;
+       ictx->cstorm_st_context.hq_curr_pbe.lo = iscsi->hq_info.pgtbl[0];
+       ictx->cstorm_st_context.hq_curr_pbe.hi = iscsi->hq_info.pgtbl[1];
+       ictx->cstorm_st_context.task_pbl_base.lo =
+               iscsi->task_array_info.pgtbl_map & 0xffffffff;
+       ictx->cstorm_st_context.task_pbl_base.hi =
+               (u64) iscsi->task_array_info.pgtbl_map >> 32;
+       /* CSTORM and USTORM initialization is different, CSTORM requires
+        * CQ DB base & not PTE addr */
+       ictx->cstorm_st_context.cq_db_base.lo =
+               req1->cq_page_table_addr_lo & PAGE_MASK;
+       ictx->cstorm_st_context.cq_db_base.hi = req1->cq_page_table_addr_hi;
+       ictx->cstorm_st_context.iscsi_conn_id = req1->iscsi_conn_id;
+       ictx->cstorm_st_context.cq_proc_en_bit_map = (1 << cp->num_cqs) - 1;
+       for (i = 0; i < cp->num_cqs; i++) {
+               ictx->cstorm_st_context.cq_c_prod_sqn_arr.sqn[i] =
+                       ISCSI_INITIAL_SN;
+               ictx->cstorm_st_context.cq_c_sqn_2_notify_arr.sqn[i] =
+                       ISCSI_INITIAL_SN;
+       }
+
+       ictx->xstorm_ag_context.cdu_reserved =
+               CDU_RSRVD_VALUE_TYPE_A(hw_cid, CDU_REGION_NUMBER_XCM_AG,
+                                      ISCSI_CONNECTION_TYPE);
+       ictx->ustorm_ag_context.cdu_usage =
+               CDU_RSRVD_VALUE_TYPE_A(hw_cid, CDU_REGION_NUMBER_UCM_AG,
+                                      ISCSI_CONNECTION_TYPE);
+       return 0;
+
+}
+
+static int cnic_bnx2x_iscsi_ofld1(struct cnic_dev *dev, struct kwqe *wqes[],
+                                  u32 num, int *work)
+{
+       struct iscsi_kwqe_conn_offload1 *req1;
+       struct iscsi_kwqe_conn_offload2 *req2;
+       struct cnic_local *cp = dev->cnic_priv;
+       struct iscsi_kcqe kcqe;
+       struct kcqe *cqes[1];
+       u32 l5_cid;
+       int ret;
+
+       if (num < 2) {
+               *work = num;
+               return -EINVAL;
+       }
+
+       req1 = (struct iscsi_kwqe_conn_offload1 *) wqes[0];
+       req2 = (struct iscsi_kwqe_conn_offload2 *) wqes[1];
+       if ((num - 2) < req2->num_additional_wqes) {
+               *work = num;
+               return -EINVAL;
+       }
+       *work = 2 + req2->num_additional_wqes;;
+
+       l5_cid = req1->iscsi_conn_id;
+       if (l5_cid >= MAX_ISCSI_TBL_SZ)
+               return -EINVAL;
+
+       memset(&kcqe, 0, sizeof(kcqe));
+       kcqe.op_code = ISCSI_KCQE_OPCODE_OFFLOAD_CONN;
+       kcqe.iscsi_conn_id = l5_cid;
+       kcqe.completion_status = ISCSI_KCQE_COMPLETION_STATUS_CTX_ALLOC_FAILURE;
+
+       if (atomic_inc_return(&cp->iscsi_conn) > dev->max_iscsi_conn) {
+               atomic_dec(&cp->iscsi_conn);
+               ret = 0;
+               goto done;
+       }
+       ret = cnic_alloc_bnx2x_conn_resc(dev, l5_cid);
+       if (ret) {
+               atomic_dec(&cp->iscsi_conn);
+               ret = 0;
+               goto done;
+       }
+       ret = cnic_setup_bnx2x_ctx(dev, wqes, num);
+       if (ret < 0) {
+               cnic_free_bnx2x_conn_resc(dev, l5_cid);
+               atomic_dec(&cp->iscsi_conn);
+               goto done;
+       }
+
+       kcqe.completion_status = ISCSI_KCQE_COMPLETION_STATUS_SUCCESS;
+       kcqe.iscsi_conn_context_id = BNX2X_HW_CID(cp->ctx_tbl[l5_cid].cid,
+                                                 cp->func);
+
+done:
+       cqes[0] = (struct kcqe *) &kcqe;
+       cnic_reply_bnx2x_kcqes(dev, CNIC_ULP_ISCSI, cqes, 1);
+       return ret;
+}
+
+
+static int cnic_bnx2x_iscsi_update(struct cnic_dev *dev, struct kwqe *kwqe)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct iscsi_kwqe_conn_update *req =
+               (struct iscsi_kwqe_conn_update *) kwqe;
+       void *data;
+       union l5cm_specific_data l5_data;
+       u32 l5_cid, cid = BNX2X_SW_CID(req->context_id);
+       int ret;
+
+       if (cnic_get_l5_cid(cp, cid, &l5_cid) != 0)
+               return -EINVAL;
+
+       data = cnic_get_kwqe_16_data(cp, l5_cid, &l5_data);
+       if (!data)
+               return -ENOMEM;
+
+       memcpy(data, kwqe, sizeof(struct kwqe));
+
+       ret = cnic_submit_kwqe_16(dev, ISCSI_RAMROD_CMD_ID_UPDATE_CONN,
+                       req->context_id, ISCSI_CONNECTION_TYPE, &l5_data);
+       return ret;
+}
+
+static int cnic_bnx2x_iscsi_destroy(struct cnic_dev *dev, struct kwqe *kwqe)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct iscsi_kwqe_conn_destroy *req =
+               (struct iscsi_kwqe_conn_destroy *) kwqe;
+       union l5cm_specific_data l5_data;
+       u32 l5_cid = req->reserved0;
+       struct cnic_context *ctx = &cp->ctx_tbl[l5_cid];
+       int ret = 0;
+       struct iscsi_kcqe kcqe;
+       struct kcqe *cqes[1];
+
+       if (!(ctx->ctx_flags & CTX_FL_OFFLD_START))
+               goto skip_cfc_delete;
+
+       while (!time_after(jiffies, ctx->timestamp + (2 * HZ)))
+               msleep(250);
+
+       init_waitqueue_head(&ctx->waitq);
+       ctx->wait_cond = 0;
+       memset(&l5_data, 0, sizeof(l5_data));
+       ret = cnic_submit_kwqe_16(dev, RAMROD_CMD_ID_ETH_CFC_DEL,
+                                 req->context_id,
+                                 ETH_CONNECTION_TYPE |
+                                 (1 << SPE_HDR_COMMON_RAMROD_SHIFT),
+                                 &l5_data);
+       if (ret == 0)
+               wait_event(ctx->waitq, ctx->wait_cond);
+
+skip_cfc_delete:
+       cnic_free_bnx2x_conn_resc(dev, l5_cid);
+
+       atomic_dec(&cp->iscsi_conn);
+
+       memset(&kcqe, 0, sizeof(kcqe));
+       kcqe.op_code = ISCSI_KCQE_OPCODE_DESTROY_CONN;
+       kcqe.iscsi_conn_id = l5_cid;
+       kcqe.completion_status = ISCSI_KCQE_COMPLETION_STATUS_SUCCESS;
+       kcqe.iscsi_conn_context_id = req->context_id;
+
+       cqes[0] = (struct kcqe *) &kcqe;
+       cnic_reply_bnx2x_kcqes(dev, CNIC_ULP_ISCSI, cqes, 1);
+
+       return ret;
+}
+
+static void cnic_init_storm_conn_bufs(struct cnic_dev *dev,
+                                     struct l4_kwq_connect_req1 *kwqe1,
+                                     struct l4_kwq_connect_req3 *kwqe3,
+                                     struct l5cm_active_conn_buffer *conn_buf)
+{
+       struct l5cm_conn_addr_params *conn_addr = &conn_buf->conn_addr_buf;
+       struct l5cm_xstorm_conn_buffer *xstorm_buf =
+               &conn_buf->xstorm_conn_buffer;
+       struct l5cm_tstorm_conn_buffer *tstorm_buf =
+               &conn_buf->tstorm_conn_buffer;
+       struct regpair context_addr;
+       u32 cid = BNX2X_SW_CID(kwqe1->cid);
+       struct in6_addr src_ip, dst_ip;
+       int i;
+       u32 *addrp;
+
+       addrp = (u32 *) &conn_addr->local_ip_addr;
+       for (i = 0; i < 4; i++, addrp++)
+               src_ip.in6_u.u6_addr32[i] = cpu_to_be32(*addrp);
+
+       addrp = (u32 *) &conn_addr->remote_ip_addr;
+       for (i = 0; i < 4; i++, addrp++)
+               dst_ip.in6_u.u6_addr32[i] = cpu_to_be32(*addrp);
+
+       cnic_get_bnx2x_ctx(dev, cid, 0, &context_addr);
+
+       xstorm_buf->context_addr.hi = context_addr.hi;
+       xstorm_buf->context_addr.lo = context_addr.lo;
+       xstorm_buf->mss = 0xffff;
+       xstorm_buf->rcv_buf = kwqe3->rcv_buf;
+       if (kwqe1->tcp_flags & L4_KWQ_CONNECT_REQ1_NAGLE_ENABLE)
+               xstorm_buf->params |= L5CM_XSTORM_CONN_BUFFER_NAGLE_ENABLE;
+       xstorm_buf->pseudo_header_checksum =
+               swab16(~csum_ipv6_magic(&src_ip, &dst_ip, 0, IPPROTO_TCP, 0));
+
+       if (!(kwqe1->tcp_flags & L4_KWQ_CONNECT_REQ1_NO_DELAY_ACK))
+               tstorm_buf->params |=
+                       L5CM_TSTORM_CONN_BUFFER_DELAYED_ACK_ENABLE;
+       if (kwqe3->ka_timeout) {
+               tstorm_buf->ka_enable = 1;
+               tstorm_buf->ka_timeout = kwqe3->ka_timeout;
+               tstorm_buf->ka_interval = kwqe3->ka_interval;
+               tstorm_buf->ka_max_probe_count = kwqe3->ka_max_probe_count;
+       }
+       tstorm_buf->rcv_buf = kwqe3->rcv_buf;
+       tstorm_buf->snd_buf = kwqe3->snd_buf;
+       tstorm_buf->max_rt_time = 0xffffffff;
+}
+
+static void cnic_init_bnx2x_mac(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       int func = CNIC_FUNC(cp);
+       u8 *mac = dev->mac_addr;
+
+       CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
+                XSTORM_ISCSI_LOCAL_MAC_ADDR0_OFFSET(func), mac[0]);
+       CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
+                XSTORM_ISCSI_LOCAL_MAC_ADDR1_OFFSET(func), mac[1]);
+       CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
+                XSTORM_ISCSI_LOCAL_MAC_ADDR2_OFFSET(func), mac[2]);
+       CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
+                XSTORM_ISCSI_LOCAL_MAC_ADDR3_OFFSET(func), mac[3]);
+       CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
+                XSTORM_ISCSI_LOCAL_MAC_ADDR4_OFFSET(func), mac[4]);
+       CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
+                XSTORM_ISCSI_LOCAL_MAC_ADDR5_OFFSET(func), mac[5]);
+
+       CNIC_WR8(dev, BAR_TSTRORM_INTMEM +
+                TSTORM_ISCSI_TCP_VARS_LSB_LOCAL_MAC_ADDR_OFFSET(func), mac[5]);
+       CNIC_WR8(dev, BAR_TSTRORM_INTMEM +
+                TSTORM_ISCSI_TCP_VARS_LSB_LOCAL_MAC_ADDR_OFFSET(func) + 1,
+                mac[4]);
+       CNIC_WR8(dev, BAR_TSTRORM_INTMEM +
+                TSTORM_ISCSI_TCP_VARS_MSB_LOCAL_MAC_ADDR_OFFSET(func), mac[3]);
+       CNIC_WR8(dev, BAR_TSTRORM_INTMEM +
+                TSTORM_ISCSI_TCP_VARS_MSB_LOCAL_MAC_ADDR_OFFSET(func) + 1,
+                mac[2]);
+       CNIC_WR8(dev, BAR_TSTRORM_INTMEM +
+                TSTORM_ISCSI_TCP_VARS_MSB_LOCAL_MAC_ADDR_OFFSET(func) + 2,
+                mac[1]);
+       CNIC_WR8(dev, BAR_TSTRORM_INTMEM +
+                TSTORM_ISCSI_TCP_VARS_MSB_LOCAL_MAC_ADDR_OFFSET(func) + 3,
+                mac[0]);
+}
+
+static void cnic_bnx2x_set_tcp_timestamp(struct cnic_dev *dev, int tcp_ts)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       u8 xstorm_flags = XSTORM_L5CM_TCP_FLAGS_WND_SCL_EN;
+       u16 tstorm_flags = 0;
+
+       if (tcp_ts) {
+               xstorm_flags |= XSTORM_L5CM_TCP_FLAGS_TS_ENABLED;
+               tstorm_flags |= TSTORM_L5CM_TCP_FLAGS_TS_ENABLED;
+       }
+
+       CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
+                XSTORM_ISCSI_TCP_VARS_FLAGS_OFFSET(cp->func), xstorm_flags);
+
+       CNIC_WR16(dev, BAR_TSTRORM_INTMEM +
+                 TSTORM_ISCSI_TCP_VARS_FLAGS_OFFSET(cp->func), tstorm_flags);
+}
+
+static int cnic_bnx2x_connect(struct cnic_dev *dev, struct kwqe *wqes[],
+                             u32 num, int *work)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct l4_kwq_connect_req1 *kwqe1 =
+               (struct l4_kwq_connect_req1 *) wqes[0];
+       struct l4_kwq_connect_req3 *kwqe3;
+       struct l5cm_active_conn_buffer *conn_buf;
+       struct l5cm_conn_addr_params *conn_addr;
+       union l5cm_specific_data l5_data;
+       u32 l5_cid = kwqe1->pg_cid;
+       struct cnic_sock *csk = &cp->csk_tbl[l5_cid];
+       struct cnic_context *ctx = &cp->ctx_tbl[l5_cid];
+       int ret;
+
+       if (num < 2) {
+               *work = num;
+               return -EINVAL;
+       }
+
+       if (kwqe1->conn_flags & L4_KWQ_CONNECT_REQ1_IP_V6)
+               *work = 3;
+       else
+               *work = 2;
+
+       if (num < *work) {
+               *work = num;
+               return -EINVAL;
+       }
+
+       if (sizeof(*conn_buf) > CNIC_KWQ16_DATA_SIZE) {
+               netdev_err(dev->netdev, "conn_buf size too big\n");
+               return -ENOMEM;
+       }
+       conn_buf = cnic_get_kwqe_16_data(cp, l5_cid, &l5_data);
+       if (!conn_buf)
+               return -ENOMEM;
+
+       memset(conn_buf, 0, sizeof(*conn_buf));
+
+       conn_addr = &conn_buf->conn_addr_buf;
+       conn_addr->remote_addr_0 = csk->ha[0];
+       conn_addr->remote_addr_1 = csk->ha[1];
+       conn_addr->remote_addr_2 = csk->ha[2];
+       conn_addr->remote_addr_3 = csk->ha[3];
+       conn_addr->remote_addr_4 = csk->ha[4];
+       conn_addr->remote_addr_5 = csk->ha[5];
+
+       if (kwqe1->conn_flags & L4_KWQ_CONNECT_REQ1_IP_V6) {
+               struct l4_kwq_connect_req2 *kwqe2 =
+                       (struct l4_kwq_connect_req2 *) wqes[1];
+
+               conn_addr->local_ip_addr.ip_addr_hi_hi = kwqe2->src_ip_v6_4;
+               conn_addr->local_ip_addr.ip_addr_hi_lo = kwqe2->src_ip_v6_3;
+               conn_addr->local_ip_addr.ip_addr_lo_hi = kwqe2->src_ip_v6_2;
+
+               conn_addr->remote_ip_addr.ip_addr_hi_hi = kwqe2->dst_ip_v6_4;
+               conn_addr->remote_ip_addr.ip_addr_hi_lo = kwqe2->dst_ip_v6_3;
+               conn_addr->remote_ip_addr.ip_addr_lo_hi = kwqe2->dst_ip_v6_2;
+               conn_addr->params |= L5CM_CONN_ADDR_PARAMS_IP_VERSION;
+       }
+       kwqe3 = (struct l4_kwq_connect_req3 *) wqes[*work - 1];
+
+       conn_addr->local_ip_addr.ip_addr_lo_lo = kwqe1->src_ip;
+       conn_addr->remote_ip_addr.ip_addr_lo_lo = kwqe1->dst_ip;
+       conn_addr->local_tcp_port = kwqe1->src_port;
+       conn_addr->remote_tcp_port = kwqe1->dst_port;
+
+       conn_addr->pmtu = kwqe3->pmtu;
+       cnic_init_storm_conn_bufs(dev, kwqe1, kwqe3, conn_buf);
+
+       CNIC_WR16(dev, BAR_XSTRORM_INTMEM +
+                 XSTORM_ISCSI_LOCAL_VLAN_OFFSET(cp->func), csk->vlan_id);
+
+       cnic_bnx2x_set_tcp_timestamp(dev,
+               kwqe1->tcp_flags & L4_KWQ_CONNECT_REQ1_TIME_STAMP);
+
+       ret = cnic_submit_kwqe_16(dev, L5CM_RAMROD_CMD_ID_TCP_CONNECT,
+                       kwqe1->cid, ISCSI_CONNECTION_TYPE, &l5_data);
+       if (!ret)
+               ctx->ctx_flags |= CTX_FL_OFFLD_START;
+
+       return ret;
+}
+
+static int cnic_bnx2x_close(struct cnic_dev *dev, struct kwqe *kwqe)
+{
+       struct l4_kwq_close_req *req = (struct l4_kwq_close_req *) kwqe;
+       union l5cm_specific_data l5_data;
+       int ret;
+
+       memset(&l5_data, 0, sizeof(l5_data));
+       ret = cnic_submit_kwqe_16(dev, L5CM_RAMROD_CMD_ID_CLOSE,
+                       req->cid, ISCSI_CONNECTION_TYPE, &l5_data);
+       return ret;
+}
+
+static int cnic_bnx2x_reset(struct cnic_dev *dev, struct kwqe *kwqe)
+{
+       struct l4_kwq_reset_req *req = (struct l4_kwq_reset_req *) kwqe;
+       union l5cm_specific_data l5_data;
+       int ret;
+
+       memset(&l5_data, 0, sizeof(l5_data));
+       ret = cnic_submit_kwqe_16(dev, L5CM_RAMROD_CMD_ID_ABORT,
+                       req->cid, ISCSI_CONNECTION_TYPE, &l5_data);
+       return ret;
+}
+static int cnic_bnx2x_offload_pg(struct cnic_dev *dev, struct kwqe *kwqe)
+{
+       struct l4_kwq_offload_pg *req = (struct l4_kwq_offload_pg *) kwqe;
+       struct l4_kcq kcqe;
+       struct kcqe *cqes[1];
+
+       memset(&kcqe, 0, sizeof(kcqe));
+       kcqe.pg_host_opaque = req->host_opaque;
+       kcqe.pg_cid = req->host_opaque;
+       kcqe.op_code = L4_KCQE_OPCODE_VALUE_OFFLOAD_PG;
+       cqes[0] = (struct kcqe *) &kcqe;
+       cnic_reply_bnx2x_kcqes(dev, CNIC_ULP_L4, cqes, 1);
+       return 0;
+}
+
+static int cnic_bnx2x_update_pg(struct cnic_dev *dev, struct kwqe *kwqe)
+{
+       struct l4_kwq_update_pg *req = (struct l4_kwq_update_pg *) kwqe;
+       struct l4_kcq kcqe;
+       struct kcqe *cqes[1];
+
+       memset(&kcqe, 0, sizeof(kcqe));
+       kcqe.pg_host_opaque = req->pg_host_opaque;
+       kcqe.pg_cid = req->pg_cid;
+       kcqe.op_code = L4_KCQE_OPCODE_VALUE_UPDATE_PG;
+       cqes[0] = (struct kcqe *) &kcqe;
+       cnic_reply_bnx2x_kcqes(dev, CNIC_ULP_L4, cqes, 1);
+       return 0;
+}
+
+static int cnic_submit_bnx2x_kwqes(struct cnic_dev *dev, struct kwqe *wqes[],
+                                  u32 num_wqes)
+{
+       int i, work, ret;
+       u32 opcode;
+       struct kwqe *kwqe;
+
+       if (!test_bit(CNIC_F_CNIC_UP, &dev->flags))
+               return -EAGAIN;         /* bnx2 is down */
+
+       for (i = 0; i < num_wqes; ) {
+               kwqe = wqes[i];
+               opcode = KWQE_OPCODE(kwqe->kwqe_op_flag);
+               work = 1;
+
+               switch (opcode) {
+               case ISCSI_KWQE_OPCODE_INIT1:
+                       ret = cnic_bnx2x_iscsi_init1(dev, kwqe);
+                       break;
+               case ISCSI_KWQE_OPCODE_INIT2:
+                       ret = cnic_bnx2x_iscsi_init2(dev, kwqe);
+                       break;
+               case ISCSI_KWQE_OPCODE_OFFLOAD_CONN1:
+                       ret = cnic_bnx2x_iscsi_ofld1(dev, &wqes[i],
+                                                    num_wqes - i, &work);
+                       break;
+               case ISCSI_KWQE_OPCODE_UPDATE_CONN:
+                       ret = cnic_bnx2x_iscsi_update(dev, kwqe);
+                       break;
+               case ISCSI_KWQE_OPCODE_DESTROY_CONN:
+                       ret = cnic_bnx2x_iscsi_destroy(dev, kwqe);
+                       break;
+               case L4_KWQE_OPCODE_VALUE_CONNECT1:
+                       ret = cnic_bnx2x_connect(dev, &wqes[i], num_wqes - i,
+                                                &work);
+                       break;
+               case L4_KWQE_OPCODE_VALUE_CLOSE:
+                       ret = cnic_bnx2x_close(dev, kwqe);
+                       break;
+               case L4_KWQE_OPCODE_VALUE_RESET:
+                       ret = cnic_bnx2x_reset(dev, kwqe);
+                       break;
+               case L4_KWQE_OPCODE_VALUE_OFFLOAD_PG:
+                       ret = cnic_bnx2x_offload_pg(dev, kwqe);
+                       break;
+               case L4_KWQE_OPCODE_VALUE_UPDATE_PG:
+                       ret = cnic_bnx2x_update_pg(dev, kwqe);
+                       break;
+               case L4_KWQE_OPCODE_VALUE_UPLOAD_PG:
+                       ret = 0;
+                       break;
+               default:
+                       ret = 0;
+                       netdev_err(dev->netdev, "Unknown type of KWQE(0x%x)\n",
+                                  opcode);
+                       break;
+               }
+               if (ret < 0)
+                       netdev_err(dev->netdev, "KWQE(0x%x) failed\n",
+                                  opcode);
+               i += work;
+       }
+       return 0;
+}
+
+static void service_kcqes(struct cnic_dev *dev, int num_cqes)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       int i, j;
+
+       i = 0;
+       j = 1;
+       while (num_cqes) {
+               struct cnic_ulp_ops *ulp_ops;
+               int ulp_type;
+               u32 kcqe_op_flag = cp->completed_kcq[i]->kcqe_op_flag;
+               u32 kcqe_layer = kcqe_op_flag & KCQE_FLAGS_LAYER_MASK;
+
+               if (unlikely(kcqe_op_flag & KCQE_RAMROD_COMPLETION))
                        cnic_kwq_completion(dev, 1);
 
                while (j < num_cqes) {
@@ -884,8 +2075,8 @@ static void service_kcqes(struct cnic_dev *dev, int num_cqes)
                else if (kcqe_layer == KCQE_FLAGS_LAYER_MASK_L2)
                        goto end;
                else {
-                       printk(KERN_ERR PFX "%s: Unknown type of KCQE(0x%x)\n",
-                              dev->netdev->name, kcqe_op_flag);
+                       netdev_err(dev->netdev, "Unknown type of KCQE(0x%x)\n",
+                                  kcqe_op_flag);
                        goto end;
                }
 
@@ -914,6 +2105,22 @@ static u16 cnic_bnx2_hw_idx(u16 idx)
        return idx;
 }
 
+static u16 cnic_bnx2x_next_idx(u16 idx)
+{
+       idx++;
+       if ((idx & MAX_KCQE_CNT) == MAX_KCQE_CNT)
+               idx++;
+
+       return idx;
+}
+
+static u16 cnic_bnx2x_hw_idx(u16 idx)
+{
+       if ((idx & MAX_KCQE_CNT) == MAX_KCQE_CNT)
+               idx++;
+       return idx;
+}
+
 static int cnic_get_kcqes(struct cnic_dev *dev, u16 hw_prod, u16 *sw_prod)
 {
        struct cnic_local *cp = dev->cnic_priv;
@@ -939,7 +2146,7 @@ static int cnic_get_kcqes(struct cnic_dev *dev, u16 hw_prod, u16 *sw_prod)
        return last_cnt;
 }
 
-static void cnic_chk_bnx2_pkt_rings(struct cnic_local *cp)
+static void cnic_chk_pkt_rings(struct cnic_local *cp)
 {
        u16 rx_cons = *cp->rx_cons_ptr;
        u16 tx_cons = *cp->tx_cons_ptr;
@@ -947,6 +2154,7 @@ static void cnic_chk_bnx2_pkt_rings(struct cnic_local *cp)
        if (cp->tx_cons != tx_cons || cp->rx_cons != rx_cons) {
                cp->tx_cons = tx_cons;
                cp->rx_cons = rx_cons;
+
                uio_event_notify(cp->cnic_uinfo);
        }
 }
@@ -989,7 +2197,7 @@ done:
 
        cp->kcq_prod_idx = sw_prod;
 
-       cnic_chk_bnx2_pkt_rings(cp);
+       cnic_chk_pkt_rings(cp);
        return status_idx;
 }
 
@@ -997,7 +2205,7 @@ static void cnic_service_bnx2_msix(unsigned long data)
 {
        struct cnic_dev *dev = (struct cnic_dev *) data;
        struct cnic_local *cp = dev->cnic_priv;
-       struct status_block_msix *status_blk = cp->bnx2_status_blk;
+       struct status_block_msix *status_blk = cp->status_blk.bnx2;
        u32 status_idx = status_blk->status_idx;
        u16 hw_prod, sw_prod;
        int kcqe_cnt;
@@ -1027,7 +2235,7 @@ done:
        CNIC_WR16(dev, cp->kcq_io_addr, sw_prod);
        cp->kcq_prod_idx = sw_prod;
 
-       cnic_chk_bnx2_pkt_rings(cp);
+       cnic_chk_pkt_rings(cp);
 
        cp->last_status_idx = status_idx;
        CNIC_WR(dev, BNX2_PCICFG_INT_ACK_CMD, cp->int_num |
@@ -1043,7 +2251,7 @@ static irqreturn_t cnic_irq(int irq, void *dev_instance)
        if (cp->ack_int)
                cp->ack_int(dev);
 
-       prefetch(cp->status_blk);
+       prefetch(cp->status_blk.gen);
        prefetch(&cp->kcq[KCQ_PG(prod)][KCQ_IDX(prod)]);
 
        if (likely(test_bit(CNIC_F_CNIC_UP, &dev->flags)))
@@ -1052,23 +2260,116 @@ static irqreturn_t cnic_irq(int irq, void *dev_instance)
        return IRQ_HANDLED;
 }
 
+static inline void cnic_ack_bnx2x_int(struct cnic_dev *dev, u8 id, u8 storm,
+                                     u16 index, u8 op, u8 update)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       u32 hc_addr = (HC_REG_COMMAND_REG + CNIC_PORT(cp) * 32 +
+                      COMMAND_REG_INT_ACK);
+       struct igu_ack_register igu_ack;
+
+       igu_ack.status_block_index = index;
+       igu_ack.sb_id_and_flags =
+                       ((id << IGU_ACK_REGISTER_STATUS_BLOCK_ID_SHIFT) |
+                        (storm << IGU_ACK_REGISTER_STORM_ID_SHIFT) |
+                        (update << IGU_ACK_REGISTER_UPDATE_INDEX_SHIFT) |
+                        (op << IGU_ACK_REGISTER_INTERRUPT_MODE_SHIFT));
+
+       CNIC_WR(dev, hc_addr, (*(u32 *)&igu_ack));
+}
+
+static void cnic_ack_bnx2x_msix(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+
+       cnic_ack_bnx2x_int(dev, cp->status_blk_num, CSTORM_ID, 0,
+                          IGU_INT_DISABLE, 0);
+}
+
+static void cnic_service_bnx2x_bh(unsigned long data)
+{
+       struct cnic_dev *dev = (struct cnic_dev *) data;
+       struct cnic_local *cp = dev->cnic_priv;
+       u16 hw_prod, sw_prod;
+       struct cstorm_status_block_c *sblk =
+               &cp->status_blk.bnx2x->c_status_block;
+       u32 status_idx = sblk->status_block_index;
+       int kcqe_cnt;
+
+       if (unlikely(!test_bit(CNIC_F_CNIC_UP, &dev->flags)))
+               return;
+
+       hw_prod = sblk->index_values[HC_INDEX_C_ISCSI_EQ_CONS];
+       hw_prod = cp->hw_idx(hw_prod);
+       sw_prod = cp->kcq_prod_idx;
+       while (sw_prod != hw_prod) {
+               kcqe_cnt = cnic_get_kcqes(dev, hw_prod, &sw_prod);
+               if (kcqe_cnt == 0)
+                       goto done;
+
+               service_kcqes(dev, kcqe_cnt);
+
+               /* Tell compiler that sblk fields can change. */
+               barrier();
+               if (status_idx == sblk->status_block_index)
+                       break;
+
+               status_idx = sblk->status_block_index;
+               hw_prod = sblk->index_values[HC_INDEX_C_ISCSI_EQ_CONS];
+               hw_prod = cp->hw_idx(hw_prod);
+       }
+
+done:
+       CNIC_WR16(dev, cp->kcq_io_addr, sw_prod + MAX_KCQ_IDX);
+       cnic_ack_bnx2x_int(dev, cp->status_blk_num, CSTORM_ID,
+                          status_idx, IGU_INT_ENABLE, 1);
+
+       cp->kcq_prod_idx = sw_prod;
+       return;
+}
+
+static int cnic_service_bnx2x(void *data, void *status_blk)
+{
+       struct cnic_dev *dev = data;
+       struct cnic_local *cp = dev->cnic_priv;
+       u16 prod = cp->kcq_prod_idx & MAX_KCQ_IDX;
+
+       if (likely(test_bit(CNIC_F_CNIC_UP, &dev->flags))) {
+               prefetch(cp->status_blk.bnx2x);
+               prefetch(&cp->kcq[KCQ_PG(prod)][KCQ_IDX(prod)]);
+
+               tasklet_schedule(&cp->cnic_irq_task);
+               cnic_chk_pkt_rings(cp);
+       }
+
+       return 0;
+}
+
 static void cnic_ulp_stop(struct cnic_dev *dev)
 {
        struct cnic_local *cp = dev->cnic_priv;
        int if_type;
 
-       rcu_read_lock();
+       if (cp->cnic_uinfo)
+               cnic_send_nlmsg(cp, ISCSI_KEVENT_IF_DOWN, NULL);
+
        for (if_type = 0; if_type < MAX_CNIC_ULP_TYPE; if_type++) {
                struct cnic_ulp_ops *ulp_ops;
 
-               ulp_ops = rcu_dereference(cp->ulp_ops[if_type]);
-               if (!ulp_ops)
+               mutex_lock(&cnic_lock);
+               ulp_ops = cp->ulp_ops[if_type];
+               if (!ulp_ops) {
+                       mutex_unlock(&cnic_lock);
                        continue;
+               }
+               set_bit(ULP_F_CALL_PENDING, &cp->ulp_flags[if_type]);
+               mutex_unlock(&cnic_lock);
 
                if (test_and_clear_bit(ULP_F_START, &cp->ulp_flags[if_type]))
                        ulp_ops->cnic_stop(cp->ulp_handle[if_type]);
+
+               clear_bit(ULP_F_CALL_PENDING, &cp->ulp_flags[if_type]);
        }
-       rcu_read_unlock();
 }
 
 static void cnic_ulp_start(struct cnic_dev *dev)
@@ -1076,18 +2377,23 @@ static void cnic_ulp_start(struct cnic_dev *dev)
        struct cnic_local *cp = dev->cnic_priv;
        int if_type;
 
-       rcu_read_lock();
        for (if_type = 0; if_type < MAX_CNIC_ULP_TYPE; if_type++) {
                struct cnic_ulp_ops *ulp_ops;
 
-               ulp_ops = rcu_dereference(cp->ulp_ops[if_type]);
-               if (!ulp_ops || !ulp_ops->cnic_start)
+               mutex_lock(&cnic_lock);
+               ulp_ops = cp->ulp_ops[if_type];
+               if (!ulp_ops || !ulp_ops->cnic_start) {
+                       mutex_unlock(&cnic_lock);
                        continue;
+               }
+               set_bit(ULP_F_CALL_PENDING, &cp->ulp_flags[if_type]);
+               mutex_unlock(&cnic_lock);
 
                if (!test_and_set_bit(ULP_F_START, &cp->ulp_flags[if_type]))
                        ulp_ops->cnic_start(cp->ulp_handle[if_type]);
+
+               clear_bit(ULP_F_CALL_PENDING, &cp->ulp_flags[if_type]);
        }
-       rcu_read_unlock();
 }
 
 static int cnic_ctl(void *data, struct cnic_ctl_info *info)
@@ -1097,24 +2403,33 @@ static int cnic_ctl(void *data, struct cnic_ctl_info *info)
        switch (info->cmd) {
        case CNIC_CTL_STOP_CMD:
                cnic_hold(dev);
-               mutex_lock(&cnic_lock);
 
                cnic_ulp_stop(dev);
                cnic_stop_hw(dev);
 
-               mutex_unlock(&cnic_lock);
                cnic_put(dev);
                break;
        case CNIC_CTL_START_CMD:
                cnic_hold(dev);
-               mutex_lock(&cnic_lock);
 
                if (!cnic_start_hw(dev))
                        cnic_ulp_start(dev);
 
-               mutex_unlock(&cnic_lock);
                cnic_put(dev);
                break;
+       case CNIC_CTL_COMPLETION_CMD: {
+               u32 cid = BNX2X_SW_CID(info->data.comp.cid);
+               u32 l5_cid;
+               struct cnic_local *cp = dev->cnic_priv;
+
+               if (cnic_get_l5_cid(cp, cid, &l5_cid) == 0) {
+                       struct cnic_context *ctx = &cp->ctx_tbl[l5_cid];
+
+                       ctx->wait_cond = 1;
+                       wake_up(&ctx->waitq);
+               }
+               break;
+       }
        default:
                return -EINVAL;
        }
@@ -1126,19 +2441,23 @@ static void cnic_ulp_init(struct cnic_dev *dev)
        int i;
        struct cnic_local *cp = dev->cnic_priv;
 
-       rcu_read_lock();
        for (i = 0; i < MAX_CNIC_ULP_TYPE_EXT; i++) {
                struct cnic_ulp_ops *ulp_ops;
 
-               ulp_ops = rcu_dereference(cnic_ulp_tbl[i]);
-               if (!ulp_ops || !ulp_ops->cnic_init)
+               mutex_lock(&cnic_lock);
+               ulp_ops = cnic_ulp_tbl[i];
+               if (!ulp_ops || !ulp_ops->cnic_init) {
+                       mutex_unlock(&cnic_lock);
                        continue;
+               }
+               ulp_get(ulp_ops);
+               mutex_unlock(&cnic_lock);
 
                if (!test_and_set_bit(ULP_F_INIT, &cp->ulp_flags[i]))
                        ulp_ops->cnic_init(dev);
 
+               ulp_put(ulp_ops);
        }
-       rcu_read_unlock();
 }
 
 static void cnic_ulp_exit(struct cnic_dev *dev)
@@ -1146,19 +2465,23 @@ static void cnic_ulp_exit(struct cnic_dev *dev)
        int i;
        struct cnic_local *cp = dev->cnic_priv;
 
-       rcu_read_lock();
        for (i = 0; i < MAX_CNIC_ULP_TYPE_EXT; i++) {
                struct cnic_ulp_ops *ulp_ops;
 
-               ulp_ops = rcu_dereference(cnic_ulp_tbl[i]);
-               if (!ulp_ops || !ulp_ops->cnic_exit)
+               mutex_lock(&cnic_lock);
+               ulp_ops = cnic_ulp_tbl[i];
+               if (!ulp_ops || !ulp_ops->cnic_exit) {
+                       mutex_unlock(&cnic_lock);
                        continue;
+               }
+               ulp_get(ulp_ops);
+               mutex_unlock(&cnic_lock);
 
                if (test_and_clear_bit(ULP_F_INIT, &cp->ulp_flags[i]))
                        ulp_ops->cnic_exit(dev);
 
+               ulp_put(ulp_ops);
        }
-       rcu_read_unlock();
 }
 
 static int cnic_cm_offload_pg(struct cnic_sock *csk)
@@ -1191,7 +2514,7 @@ static int cnic_cm_offload_pg(struct cnic_sock *csk)
        l4kwqe->sa5 = dev->mac_addr[5];
 
        l4kwqe->etype = ETH_P_IP;
-       l4kwqe->ipid_count = DEF_IPID_COUNT;
+       l4kwqe->ipid_start = DEF_IPID_START;
        l4kwqe->host_opaque = csk->l5_cid;
 
        if (csk->vlan_id) {
@@ -1537,8 +2860,8 @@ static int cnic_get_route(struct cnic_sock *csk, struct cnic_sockaddr *saddr)
 {
        struct cnic_dev *dev = csk->dev;
        struct cnic_local *cp = dev->cnic_priv;
-       int is_v6, err, rc = -ENETUNREACH;
-       struct dst_entry *dst;
+       int is_v6, rc = 0;
+       struct dst_entry *dst = NULL;
        struct net_device *realdev;
        u32 local_port;
 
@@ -1554,39 +2877,31 @@ static int cnic_get_route(struct cnic_sock *csk, struct cnic_sockaddr *saddr)
        clear_bit(SK_F_IPV6, &csk->flags);
 
        if (is_v6) {
-#if defined(CONFIG_IPV6) || (defined(CONFIG_IPV6_MODULE) && defined(MODULE))
                set_bit(SK_F_IPV6, &csk->flags);
-               err = cnic_get_v6_route(&saddr->remote.v6, &dst);
-               if (err)
-                       return err;
-
-               if (!dst || dst->error || !dst->dev)
-                       goto err_out;
+               cnic_get_v6_route(&saddr->remote.v6, &dst);
 
                memcpy(&csk->dst_ip[0], &saddr->remote.v6.sin6_addr,
                       sizeof(struct in6_addr));
                csk->dst_port = saddr->remote.v6.sin6_port;
                local_port = saddr->local.v6.sin6_port;
-#else
-               return rc;
-#endif
-
-       } else {
-               err = cnic_get_v4_route(&saddr->remote.v4, &dst);
-               if (err)
-                       return err;
 
-               if (!dst || dst->error || !dst->dev)
-                       goto err_out;
+       } else {
+               cnic_get_v4_route(&saddr->remote.v4, &dst);
 
                csk->dst_ip[0] = saddr->remote.v4.sin_addr.s_addr;
                csk->dst_port = saddr->remote.v4.sin_port;
                local_port = saddr->local.v4.sin_port;
        }
 
-       csk->vlan_id = cnic_get_vlan(dst->dev, &realdev);
-       if (realdev != dev->netdev)
-               goto err_out;
+       csk->vlan_id = 0;
+       csk->mtu = dev->netdev->mtu;
+       if (dst && dst->dev) {
+               u16 vlan = cnic_get_vlan(dst->dev, &realdev);
+               if (realdev == dev->netdev) {
+                       csk->vlan_id = vlan;
+                       csk->mtu = dst_mtu(dst);
+               }
+       }
 
        if (local_port >= CNIC_LOCAL_PORT_MIN &&
            local_port < CNIC_LOCAL_PORT_MAX) {
@@ -1604,9 +2919,6 @@ static int cnic_get_route(struct cnic_sock *csk, struct cnic_sockaddr *saddr)
        }
        csk->src_port = local_port;
 
-       csk->mtu = dst_mtu(dst);
-       rc = 0;
-
 err_out:
        dst_release(dst);
        return rc;
@@ -1730,6 +3042,14 @@ static void cnic_cm_process_offld_pg(struct cnic_dev *dev, struct l4_kcq *kcqe)
                clear_bit(SK_F_OFFLD_SCHED, &csk->flags);
                goto done;
        }
+       /* Possible PG kcqe status:  SUCCESS, OFFLOADED_PG, or CTX_ALLOC_FAIL */
+       if (kcqe->status == L4_KCQE_COMPLETION_STATUS_CTX_ALLOC_FAIL) {
+               clear_bit(SK_F_OFFLD_SCHED, &csk->flags);
+               cnic_cm_upcall(cp, csk,
+                              L4_KCQE_OPCODE_VALUE_CONNECT_COMPLETE);
+               goto done;
+       }
+
        csk->pg_cid = kcqe->pg_cid;
        set_bit(SK_F_PG_OFFLD_COMPLETE, &csk->flags);
        cnic_cm_conn_req(csk);
@@ -1767,6 +3087,13 @@ static void cnic_cm_process_kcqe(struct cnic_dev *dev, struct kcqe *kcqe)
        }
 
        switch (opcode) {
+       case L5CM_RAMROD_CMD_ID_TCP_CONNECT:
+               if (l4kcqe->status != 0) {
+                       clear_bit(SK_F_OFFLD_SCHED, &csk->flags);
+                       cnic_cm_upcall(cp, csk,
+                                      L4_KCQE_OPCODE_VALUE_CONNECT_COMPLETE);
+               }
+               break;
        case L4_KCQE_OPCODE_VALUE_CONNECT_COMPLETE:
                if (l4kcqe->status == 0)
                        set_bit(SK_F_OFFLD_COMPLETE, &csk->flags);
@@ -1777,11 +3104,16 @@ static void cnic_cm_process_kcqe(struct cnic_dev *dev, struct kcqe *kcqe)
                break;
 
        case L4_KCQE_OPCODE_VALUE_RESET_RECEIVED:
-               if (test_and_clear_bit(SK_F_OFFLD_COMPLETE, &csk->flags))
+               if (test_bit(CNIC_F_BNX2_CLASS, &dev->flags)) {
+                       cnic_cm_upcall(cp, csk, opcode);
+                       break;
+               } else if (test_and_clear_bit(SK_F_OFFLD_COMPLETE, &csk->flags))
                        csk->state = opcode;
                /* fall through */
        case L4_KCQE_OPCODE_VALUE_CLOSE_COMP:
        case L4_KCQE_OPCODE_VALUE_RESET_COMP:
+       case L5CM_RAMROD_CMD_ID_SEARCHER_DELETE:
+       case L5CM_RAMROD_CMD_ID_TERMINATE_OFFLOAD:
                cp->close_conn(csk, opcode);
                break;
 
@@ -1839,6 +3171,16 @@ static int cnic_ready_to_close(struct cnic_sock *csk, u32 opcode)
                if (!test_and_set_bit(SK_F_CLOSING, &csk->flags))
                        return 1;
        }
+       /* 57710+ only  workaround to handle unsolicited RESET_COMP
+        * which will be treated like a RESET RCVD notification
+        * which triggers the clean up procedure
+        */
+       else if (opcode == L4_KCQE_OPCODE_VALUE_RESET_COMP) {
+               if (!test_and_set_bit(SK_F_CLOSING, &csk->flags)) {
+                       csk->state = L4_KCQE_OPCODE_VALUE_RESET_RECEIVED;
+                       return 1;
+               }
+       }
        return 0;
 }
 
@@ -1848,10 +3190,8 @@ static void cnic_close_bnx2_conn(struct cnic_sock *csk, u32 opcode)
        struct cnic_local *cp = dev->cnic_priv;
 
        clear_bit(SK_F_CONNECT_START, &csk->flags);
-       if (cnic_ready_to_close(csk, opcode)) {
-               cnic_close_conn(csk);
-               cnic_cm_upcall(cp, csk, opcode);
-       }
+       cnic_close_conn(csk);
+       cnic_cm_upcall(cp, csk, opcode);
 }
 
 static void cnic_cm_stop_bnx2_hw(struct cnic_dev *dev)
@@ -1867,6 +3207,76 @@ static int cnic_cm_init_bnx2_hw(struct cnic_dev *dev)
        return 0;
 }
 
+static void cnic_close_bnx2x_conn(struct cnic_sock *csk, u32 opcode)
+{
+       struct cnic_dev *dev = csk->dev;
+       struct cnic_local *cp = dev->cnic_priv;
+       struct cnic_context *ctx = &cp->ctx_tbl[csk->l5_cid];
+       union l5cm_specific_data l5_data;
+       u32 cmd = 0;
+       int close_complete = 0;
+
+       switch (opcode) {
+       case L4_KCQE_OPCODE_VALUE_RESET_RECEIVED:
+       case L4_KCQE_OPCODE_VALUE_CLOSE_COMP:
+       case L4_KCQE_OPCODE_VALUE_RESET_COMP:
+               if (cnic_ready_to_close(csk, opcode))
+                       cmd = L5CM_RAMROD_CMD_ID_SEARCHER_DELETE;
+               break;
+       case L5CM_RAMROD_CMD_ID_SEARCHER_DELETE:
+               cmd = L5CM_RAMROD_CMD_ID_TERMINATE_OFFLOAD;
+               break;
+       case L5CM_RAMROD_CMD_ID_TERMINATE_OFFLOAD:
+               close_complete = 1;
+               break;
+       }
+       if (cmd) {
+               memset(&l5_data, 0, sizeof(l5_data));
+
+               cnic_submit_kwqe_16(dev, cmd, csk->cid, ISCSI_CONNECTION_TYPE,
+                                   &l5_data);
+       } else if (close_complete) {
+               ctx->timestamp = jiffies;
+               cnic_close_conn(csk);
+               cnic_cm_upcall(cp, csk, csk->state);
+       }
+}
+
+static void cnic_cm_stop_bnx2x_hw(struct cnic_dev *dev)
+{
+}
+
+static int cnic_cm_init_bnx2x_hw(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       int func = CNIC_FUNC(cp);
+
+       cnic_init_bnx2x_mac(dev);
+       cnic_bnx2x_set_tcp_timestamp(dev, 1);
+
+       CNIC_WR16(dev, BAR_XSTRORM_INTMEM +
+                 XSTORM_ISCSI_LOCAL_VLAN_OFFSET(func), 0);
+
+       CNIC_WR(dev, BAR_XSTRORM_INTMEM +
+               XSTORM_TCP_GLOBAL_DEL_ACK_COUNTER_ENABLED_OFFSET(func), 1);
+       CNIC_WR(dev, BAR_XSTRORM_INTMEM +
+               XSTORM_TCP_GLOBAL_DEL_ACK_COUNTER_MAX_COUNT_OFFSET(func),
+               DEF_MAX_DA_COUNT);
+
+       CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
+                XSTORM_ISCSI_TCP_VARS_TTL_OFFSET(func), DEF_TTL);
+       CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
+                XSTORM_ISCSI_TCP_VARS_TOS_OFFSET(func), DEF_TOS);
+       CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
+                XSTORM_ISCSI_TCP_VARS_ADV_WND_SCL_OFFSET(func), 2);
+       CNIC_WR(dev, BAR_XSTRORM_INTMEM +
+               XSTORM_TCP_TX_SWS_TIMER_VAL_OFFSET(func), DEF_SWS_TIMER);
+
+       CNIC_WR(dev, BAR_TSTRORM_INTMEM + TSTORM_TCP_MAX_CWND_OFFSET(func),
+               DEF_MAX_CWND);
+       return 0;
+}
+
 static int cnic_cm_open(struct cnic_dev *dev)
 {
        struct cnic_local *cp = dev->cnic_priv;
@@ -1999,9 +3409,8 @@ static int cnic_init_bnx2_irq(struct cnic_dev *dev)
                CNIC_WR(dev, base + BNX2_HC_COM_TICKS_OFF, (64 << 16) | 220);
                CNIC_WR(dev, base + BNX2_HC_CMD_TICKS_OFF, (64 << 16) | 220);
 
-               cp->bnx2_status_blk = cp->status_blk;
-               cp->last_status_idx = cp->bnx2_status_blk->status_idx;
-               tasklet_init(&cp->cnic_irq_task, &cnic_service_bnx2_msix,
+               cp->last_status_idx = cp->status_blk.bnx2->status_idx;
+               tasklet_init(&cp->cnic_irq_task, cnic_service_bnx2_msix,
                             (unsigned long) dev);
                err = request_irq(ethdev->irq_arr[0].vector, cnic_irq, 0,
                                  "cnic", dev);
@@ -2009,7 +3418,7 @@ static int cnic_init_bnx2_irq(struct cnic_dev *dev)
                        tasklet_disable(&cp->cnic_irq_task);
                        return err;
                }
-               while (cp->bnx2_status_blk->status_completion_producer_index &&
+               while (cp->status_blk.bnx2->status_completion_producer_index &&
                       i < 10) {
                        CNIC_WR(dev, BNX2_HC_COALESCE_NOW,
                                1 << (11 + sblk_num));
@@ -2017,13 +3426,13 @@ static int cnic_init_bnx2_irq(struct cnic_dev *dev)
                        i++;
                        barrier();
                }
-               if (cp->bnx2_status_blk->status_completion_producer_index) {
+               if (cp->status_blk.bnx2->status_completion_producer_index) {
                        cnic_free_irq(dev);
                        goto failed;
                }
 
        } else {
-               struct status_block *sblk = cp->status_blk;
+               struct status_block *sblk = cp->status_blk.gen;
                u32 hc_cmd = CNIC_RD(dev, BNX2_HC_COMMAND);
                int i = 0;
 
@@ -2041,8 +3450,7 @@ static int cnic_init_bnx2_irq(struct cnic_dev *dev)
        return 0;
 
 failed:
-       printk(KERN_ERR PFX "%s: " "KCQ index not resetting to 0.\n",
-              dev->netdev->name);
+       netdev_err(dev->netdev, "KCQ index not resetting to 0\n");
        return -EBUSY;
 }
 
@@ -2081,7 +3489,7 @@ static void cnic_init_bnx2_tx_ring(struct cnic_dev *dev)
        int i;
        struct tx_bd *txbd;
        dma_addr_t buf_map;
-       struct status_block *s_blk = cp->status_blk;
+       struct status_block *s_blk = cp->status_blk.gen;
 
        sb_id = cp->status_blk_num;
        tx_cid = 20;
@@ -2089,7 +3497,7 @@ static void cnic_init_bnx2_tx_ring(struct cnic_dev *dev)
        cnic_init_context(dev, tx_cid + 1);
        cp->tx_cons_ptr = &s_blk->status_tx_quick_consumer_index2;
        if (ethdev->drv_state & CNIC_DRV_STATE_USING_MSIX) {
-               struct status_block_msix *sblk = cp->status_blk;
+               struct status_block_msix *sblk = cp->status_blk.bnx2;
 
                tx_cid = TX_TSS_CID + sb_id - 1;
                cnic_init_context(dev, tx_cid);
@@ -2145,7 +3553,7 @@ static void cnic_init_bnx2_rx_ring(struct cnic_dev *dev)
        u32 cid_addr, sb_id, val, coal_reg, coal_val;
        int i;
        struct rx_bd *rxbd;
-       struct status_block *s_blk = cp->status_blk;
+       struct status_block *s_blk = cp->status_blk.gen;
 
        sb_id = cp->status_blk_num;
        cnic_init_context(dev, 2);
@@ -2153,7 +3561,7 @@ static void cnic_init_bnx2_rx_ring(struct cnic_dev *dev)
        coal_reg = BNX2_HC_COMMAND;
        coal_val = CNIC_RD(dev, coal_reg);
        if (ethdev->drv_state & CNIC_DRV_STATE_USING_MSIX) {
-               struct status_block_msix *sblk = cp->status_blk;
+               struct status_block_msix *sblk = cp->status_blk.bnx2;
 
                cp->rx_cons_ptr = &sblk->status_rx_quick_consumer_index;
                coal_reg = BNX2_HC_COALESCE_NOW;
@@ -2174,9 +3582,9 @@ static void cnic_init_bnx2_rx_ring(struct cnic_dev *dev)
        cnic_ctx_wr(dev, cid_addr, BNX2_L2CTX_CTX_TYPE, val);
 
        if (sb_id == 0)
-               val = 2 << BNX2_L2CTX_STATUSB_NUM_SHIFT;
+               val = 2 << BNX2_L2CTX_L2_STATUSB_NUM_SHIFT;
        else
-               val = BNX2_L2CTX_STATUSB_NUM(sb_id);
+               val = BNX2_L2CTX_L2_STATUSB_NUM(sb_id);
        cnic_ctx_wr(dev, cid_addr, BNX2_L2CTX_HOST_BDIDX, val);
 
        rxbd = (struct rx_bd *) (cp->l2_ring + BCM_PAGE_SIZE);
@@ -2252,7 +3660,7 @@ static int cnic_start_bnx2_hw(struct cnic_dev *dev)
 {
        struct cnic_local *cp = dev->cnic_priv;
        struct cnic_eth_dev *ethdev = cp->ethdev;
-       struct status_block *sblk = cp->status_blk;
+       struct status_block *sblk = cp->status_blk.gen;
        u32 val;
        int err;
 
@@ -2333,7 +3741,7 @@ static int cnic_start_bnx2_hw(struct cnic_dev *dev)
        cp->int_num = 0;
        if (ethdev->drv_state & CNIC_DRV_STATE_USING_MSIX) {
                u32 sb_id = cp->status_blk_num;
-               u32 sb = BNX2_L2CTX_STATUSB_NUM(sb_id);
+               u32 sb = BNX2_L2CTX_L5_STATUSB_NUM(sb_id);
 
                cp->int_num = sb_id << BNX2_PCICFG_INT_ACK_CMD_INT_NUM_SHIFT;
                cnic_ctx_wr(dev, cp->kwq_cid_addr, L5_KRNLQ_HOST_QIDX, sb);
@@ -2364,8 +3772,7 @@ static int cnic_start_bnx2_hw(struct cnic_dev *dev)
 
        err = cnic_init_bnx2_irq(dev);
        if (err) {
-               printk(KERN_ERR PFX "%s: cnic_init_irq failed\n",
-                      dev->netdev->name);
+               netdev_err(dev->netdev, "cnic_init_irq failed\n");
                cnic_reg_wr_ind(dev, BNX2_CP_SCRATCH + 0x20, 0);
                cnic_reg_wr_ind(dev, BNX2_COM_SCRATCH + 0x20, 0);
                return err;
@@ -2374,6 +3781,461 @@ static int cnic_start_bnx2_hw(struct cnic_dev *dev)
        return 0;
 }
 
+static void cnic_setup_bnx2x_context(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct cnic_eth_dev *ethdev = cp->ethdev;
+       u32 start_offset = ethdev->ctx_tbl_offset;
+       int i;
+
+       for (i = 0; i < cp->ctx_blks; i++) {
+               struct cnic_ctx *ctx = &cp->ctx_arr[i];
+               dma_addr_t map = ctx->mapping;
+
+               if (cp->ctx_align) {
+                       unsigned long mask = cp->ctx_align - 1;
+
+                       map = (map + mask) & ~mask;
+               }
+
+               cnic_ctx_tbl_wr(dev, start_offset + i, map);
+       }
+}
+
+static int cnic_init_bnx2x_irq(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct cnic_eth_dev *ethdev = cp->ethdev;
+       int err = 0;
+
+       tasklet_init(&cp->cnic_irq_task, cnic_service_bnx2x_bh,
+                    (unsigned long) dev);
+       if (ethdev->drv_state & CNIC_DRV_STATE_USING_MSIX) {
+               err = request_irq(ethdev->irq_arr[0].vector, cnic_irq, 0,
+                                 "cnic", dev);
+               if (err)
+                       tasklet_disable(&cp->cnic_irq_task);
+       }
+       return err;
+}
+
+static void cnic_enable_bnx2x_int(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       u8 sb_id = cp->status_blk_num;
+       int port = CNIC_PORT(cp);
+
+       CNIC_WR8(dev, BAR_CSTRORM_INTMEM +
+                CSTORM_SB_HC_TIMEOUT_C_OFFSET(port, sb_id,
+                                              HC_INDEX_C_ISCSI_EQ_CONS),
+                64 / 12);
+       CNIC_WR16(dev, BAR_CSTRORM_INTMEM +
+                 CSTORM_SB_HC_DISABLE_C_OFFSET(port, sb_id,
+                                               HC_INDEX_C_ISCSI_EQ_CONS), 0);
+}
+
+static void cnic_disable_bnx2x_int_sync(struct cnic_dev *dev)
+{
+}
+
+static void cnic_init_bnx2x_tx_ring(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       union eth_tx_bd_types *txbd = (union eth_tx_bd_types *) cp->l2_ring;
+       struct eth_context *context;
+       struct regpair context_addr;
+       dma_addr_t buf_map;
+       int func = CNIC_FUNC(cp);
+       int port = CNIC_PORT(cp);
+       int i;
+       int cli = BNX2X_ISCSI_CL_ID(CNIC_E1HVN(cp));
+       u32 val;
+
+       memset(txbd, 0, BCM_PAGE_SIZE);
+
+       buf_map = cp->l2_buf_map;
+       for (i = 0; i < MAX_TX_DESC_CNT; i += 3, txbd += 3) {
+               struct eth_tx_start_bd *start_bd = &txbd->start_bd;
+               struct eth_tx_bd *reg_bd = &((txbd + 2)->reg_bd);
+
+               start_bd->addr_hi = cpu_to_le32((u64) buf_map >> 32);
+               start_bd->addr_lo = cpu_to_le32(buf_map & 0xffffffff);
+               reg_bd->addr_hi = start_bd->addr_hi;
+               reg_bd->addr_lo = start_bd->addr_lo + 0x10;
+               start_bd->nbytes = cpu_to_le16(0x10);
+               start_bd->nbd = cpu_to_le16(3);
+               start_bd->bd_flags.as_bitfield = ETH_TX_BD_FLAGS_START_BD;
+               start_bd->general_data = (UNICAST_ADDRESS <<
+                       ETH_TX_START_BD_ETH_ADDR_TYPE_SHIFT);
+               start_bd->general_data |= (1 << ETH_TX_START_BD_HDR_NBDS_SHIFT);
+
+       }
+       context = cnic_get_bnx2x_ctx(dev, BNX2X_ISCSI_L2_CID, 1, &context_addr);
+
+       val = (u64) cp->l2_ring_map >> 32;
+       txbd->next_bd.addr_hi = cpu_to_le32(val);
+
+       context->xstorm_st_context.tx_bd_page_base_hi = val;
+
+       val = (u64) cp->l2_ring_map & 0xffffffff;
+       txbd->next_bd.addr_lo = cpu_to_le32(val);
+
+       context->xstorm_st_context.tx_bd_page_base_lo = val;
+
+       context->cstorm_st_context.sb_index_number =
+               HC_INDEX_DEF_C_ETH_ISCSI_CQ_CONS;
+       context->cstorm_st_context.status_block_id = BNX2X_DEF_SB_ID;
+
+       context->xstorm_st_context.statistics_data = (cli |
+                               XSTORM_ETH_ST_CONTEXT_STATISTICS_ENABLE);
+
+       context->xstorm_ag_context.cdu_reserved =
+               CDU_RSRVD_VALUE_TYPE_A(BNX2X_HW_CID(BNX2X_ISCSI_L2_CID, func),
+                                       CDU_REGION_NUMBER_XCM_AG,
+                                       ETH_CONNECTION_TYPE);
+
+       /* reset xstorm per client statistics */
+       val = BAR_XSTRORM_INTMEM +
+             XSTORM_PER_COUNTER_ID_STATS_OFFSET(port, cli);
+       for (i = 0; i < sizeof(struct xstorm_per_client_stats) / 4; i++)
+               CNIC_WR(dev, val + i * 4, 0);
+
+       cp->tx_cons_ptr =
+               &cp->bnx2x_def_status_blk->c_def_status_block.index_values[
+                       HC_INDEX_DEF_C_ETH_ISCSI_CQ_CONS];
+}
+
+static void cnic_init_bnx2x_rx_ring(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct eth_rx_bd *rxbd = (struct eth_rx_bd *) (cp->l2_ring +
+                               BCM_PAGE_SIZE);
+       struct eth_rx_cqe_next_page *rxcqe = (struct eth_rx_cqe_next_page *)
+                               (cp->l2_ring + (2 * BCM_PAGE_SIZE));
+       struct eth_context *context;
+       struct regpair context_addr;
+       int i;
+       int port = CNIC_PORT(cp);
+       int func = CNIC_FUNC(cp);
+       int cli = BNX2X_ISCSI_CL_ID(CNIC_E1HVN(cp));
+       u32 val;
+       struct tstorm_eth_client_config tstorm_client = {0};
+
+       for (i = 0; i < BNX2X_MAX_RX_DESC_CNT; i++, rxbd++) {
+               dma_addr_t buf_map;
+               int n = (i % cp->l2_rx_ring_size) + 1;
+
+               buf_map = cp->l2_buf_map + (n * cp->l2_single_buf_size);
+               rxbd->addr_hi = cpu_to_le32((u64) buf_map >> 32);
+               rxbd->addr_lo = cpu_to_le32(buf_map & 0xffffffff);
+       }
+       context = cnic_get_bnx2x_ctx(dev, BNX2X_ISCSI_L2_CID, 0, &context_addr);
+
+       val = (u64) (cp->l2_ring_map + BCM_PAGE_SIZE) >> 32;
+       rxbd->addr_hi = cpu_to_le32(val);
+
+       context->ustorm_st_context.common.bd_page_base_hi = val;
+
+       val = (u64) (cp->l2_ring_map + BCM_PAGE_SIZE) & 0xffffffff;
+       rxbd->addr_lo = cpu_to_le32(val);
+
+       context->ustorm_st_context.common.bd_page_base_lo = val;
+
+       context->ustorm_st_context.common.sb_index_numbers =
+                                               BNX2X_ISCSI_RX_SB_INDEX_NUM;
+       context->ustorm_st_context.common.clientId = cli;
+       context->ustorm_st_context.common.status_block_id = BNX2X_DEF_SB_ID;
+       context->ustorm_st_context.common.flags =
+               USTORM_ETH_ST_CONTEXT_CONFIG_ENABLE_STATISTICS;
+       context->ustorm_st_context.common.statistics_counter_id = cli;
+       context->ustorm_st_context.common.mc_alignment_log_size = 0;
+       context->ustorm_st_context.common.bd_buff_size =
+                                               cp->l2_single_buf_size;
+
+       context->ustorm_ag_context.cdu_usage =
+               CDU_RSRVD_VALUE_TYPE_A(BNX2X_HW_CID(BNX2X_ISCSI_L2_CID, func),
+                                       CDU_REGION_NUMBER_UCM_AG,
+                                       ETH_CONNECTION_TYPE);
+
+       rxcqe += BNX2X_MAX_RCQ_DESC_CNT;
+       val = (u64) (cp->l2_ring_map + (2 * BCM_PAGE_SIZE)) >> 32;
+       rxcqe->addr_hi = cpu_to_le32(val);
+
+       CNIC_WR(dev, BAR_USTRORM_INTMEM +
+               USTORM_CQE_PAGE_BASE_OFFSET(port, cli) + 4, val);
+
+       CNIC_WR(dev, BAR_USTRORM_INTMEM +
+               USTORM_CQE_PAGE_NEXT_OFFSET(port, cli) + 4, val);
+
+       val = (u64) (cp->l2_ring_map + (2 * BCM_PAGE_SIZE)) & 0xffffffff;
+       rxcqe->addr_lo = cpu_to_le32(val);
+
+       CNIC_WR(dev, BAR_USTRORM_INTMEM +
+               USTORM_CQE_PAGE_BASE_OFFSET(port, cli), val);
+
+       CNIC_WR(dev, BAR_USTRORM_INTMEM +
+               USTORM_CQE_PAGE_NEXT_OFFSET(port, cli), val);
+
+       /* client tstorm info */
+       tstorm_client.mtu = cp->l2_single_buf_size - 14;
+       tstorm_client.config_flags =
+                       (TSTORM_ETH_CLIENT_CONFIG_E1HOV_REM_ENABLE |
+                       TSTORM_ETH_CLIENT_CONFIG_STATSITICS_ENABLE);
+       tstorm_client.statistics_counter_id = cli;
+
+       CNIC_WR(dev, BAR_TSTRORM_INTMEM +
+                  TSTORM_CLIENT_CONFIG_OFFSET(port, cli),
+                  ((u32 *)&tstorm_client)[0]);
+       CNIC_WR(dev, BAR_TSTRORM_INTMEM +
+                  TSTORM_CLIENT_CONFIG_OFFSET(port, cli) + 4,
+                  ((u32 *)&tstorm_client)[1]);
+
+       /* reset tstorm per client statistics */
+       val = BAR_TSTRORM_INTMEM +
+             TSTORM_PER_COUNTER_ID_STATS_OFFSET(port, cli);
+       for (i = 0; i < sizeof(struct tstorm_per_client_stats) / 4; i++)
+               CNIC_WR(dev, val + i * 4, 0);
+
+       /* reset ustorm per client statistics */
+       val = BAR_USTRORM_INTMEM +
+             USTORM_PER_COUNTER_ID_STATS_OFFSET(port, cli);
+       for (i = 0; i < sizeof(struct ustorm_per_client_stats) / 4; i++)
+               CNIC_WR(dev, val + i * 4, 0);
+
+       cp->rx_cons_ptr =
+               &cp->bnx2x_def_status_blk->u_def_status_block.index_values[
+                       HC_INDEX_DEF_U_ETH_ISCSI_RX_CQ_CONS];
+}
+
+static void cnic_get_bnx2x_iscsi_info(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       u32 base, addr, val;
+       int port = CNIC_PORT(cp);
+
+       dev->max_iscsi_conn = 0;
+       base = CNIC_RD(dev, MISC_REG_SHARED_MEM_ADDR);
+       if (base < 0xa0000 || base >= 0xc0000)
+               return;
+
+       addr = BNX2X_SHMEM_ADDR(base,
+               dev_info.port_hw_config[port].iscsi_mac_upper);
+
+       val = CNIC_RD(dev, addr);
+
+       dev->mac_addr[0] = (u8) (val >> 8);
+       dev->mac_addr[1] = (u8) val;
+
+       addr = BNX2X_SHMEM_ADDR(base,
+               dev_info.port_hw_config[port].iscsi_mac_lower);
+
+       val = CNIC_RD(dev, addr);
+
+       dev->mac_addr[2] = (u8) (val >> 24);
+       dev->mac_addr[3] = (u8) (val >> 16);
+       dev->mac_addr[4] = (u8) (val >> 8);
+       dev->mac_addr[5] = (u8) val;
+
+       addr = BNX2X_SHMEM_ADDR(base, validity_map[port]);
+       val = CNIC_RD(dev, addr);
+
+       if (!(val & SHR_MEM_VALIDITY_LIC_NO_KEY_IN_EFFECT)) {
+               u16 val16;
+
+               addr = BNX2X_SHMEM_ADDR(base,
+                               drv_lic_key[port].max_iscsi_init_conn);
+               val16 = CNIC_RD16(dev, addr);
+
+               if (val16)
+                       val16 ^= 0x1e1e;
+               dev->max_iscsi_conn = val16;
+       }
+       if (BNX2X_CHIP_IS_E1H(cp->chip_id)) {
+               int func = CNIC_FUNC(cp);
+
+               addr = BNX2X_SHMEM_ADDR(base,
+                               mf_cfg.func_mf_config[func].e1hov_tag);
+               val = CNIC_RD(dev, addr);
+               val &= FUNC_MF_CFG_E1HOV_TAG_MASK;
+               if (val != FUNC_MF_CFG_E1HOV_TAG_DEFAULT) {
+                       addr = BNX2X_SHMEM_ADDR(base,
+                               mf_cfg.func_mf_config[func].config);
+                       val = CNIC_RD(dev, addr);
+                       val &= FUNC_MF_CFG_PROTOCOL_MASK;
+                       if (val != FUNC_MF_CFG_PROTOCOL_ISCSI)
+                               dev->max_iscsi_conn = 0;
+               }
+       }
+}
+
+static int cnic_start_bnx2x_hw(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       int func = CNIC_FUNC(cp), ret, i;
+       int port = CNIC_PORT(cp);
+       u16 eq_idx;
+       u8 sb_id = cp->status_blk_num;
+
+       ret = cnic_init_id_tbl(&cp->cid_tbl, MAX_ISCSI_TBL_SZ,
+                              BNX2X_ISCSI_START_CID);
+
+       if (ret)
+               return -ENOMEM;
+
+       cp->kcq_io_addr = BAR_CSTRORM_INTMEM +
+                         CSTORM_ISCSI_EQ_PROD_OFFSET(func, 0);
+       cp->kcq_prod_idx = 0;
+
+       cnic_get_bnx2x_iscsi_info(dev);
+
+       /* Only 1 EQ */
+       CNIC_WR16(dev, cp->kcq_io_addr, MAX_KCQ_IDX);
+       CNIC_WR(dev, BAR_CSTRORM_INTMEM +
+               CSTORM_ISCSI_EQ_CONS_OFFSET(func, 0), 0);
+       CNIC_WR(dev, BAR_CSTRORM_INTMEM +
+               CSTORM_ISCSI_EQ_NEXT_PAGE_ADDR_OFFSET(func, 0),
+               cp->kcq_info.pg_map_arr[1] & 0xffffffff);
+       CNIC_WR(dev, BAR_CSTRORM_INTMEM +
+               CSTORM_ISCSI_EQ_NEXT_PAGE_ADDR_OFFSET(func, 0) + 4,
+               (u64) cp->kcq_info.pg_map_arr[1] >> 32);
+       CNIC_WR(dev, BAR_CSTRORM_INTMEM +
+               CSTORM_ISCSI_EQ_NEXT_EQE_ADDR_OFFSET(func, 0),
+               cp->kcq_info.pg_map_arr[0] & 0xffffffff);
+       CNIC_WR(dev, BAR_CSTRORM_INTMEM +
+               CSTORM_ISCSI_EQ_NEXT_EQE_ADDR_OFFSET(func, 0) + 4,
+               (u64) cp->kcq_info.pg_map_arr[0] >> 32);
+       CNIC_WR8(dev, BAR_CSTRORM_INTMEM +
+               CSTORM_ISCSI_EQ_NEXT_PAGE_ADDR_VALID_OFFSET(func, 0), 1);
+       CNIC_WR16(dev, BAR_CSTRORM_INTMEM +
+               CSTORM_ISCSI_EQ_SB_NUM_OFFSET(func, 0), cp->status_blk_num);
+       CNIC_WR8(dev, BAR_CSTRORM_INTMEM +
+               CSTORM_ISCSI_EQ_SB_INDEX_OFFSET(func, 0),
+               HC_INDEX_C_ISCSI_EQ_CONS);
+
+       for (i = 0; i < cp->conn_buf_info.num_pages; i++) {
+               CNIC_WR(dev, BAR_TSTRORM_INTMEM +
+                       TSTORM_ISCSI_CONN_BUF_PBL_OFFSET(func, i),
+                       cp->conn_buf_info.pgtbl[2 * i]);
+               CNIC_WR(dev, BAR_TSTRORM_INTMEM +
+                       TSTORM_ISCSI_CONN_BUF_PBL_OFFSET(func, i) + 4,
+                       cp->conn_buf_info.pgtbl[(2 * i) + 1]);
+       }
+
+       CNIC_WR(dev, BAR_USTRORM_INTMEM +
+               USTORM_ISCSI_GLOBAL_BUF_PHYS_ADDR_OFFSET(func),
+               cp->gbl_buf_info.pg_map_arr[0] & 0xffffffff);
+       CNIC_WR(dev, BAR_USTRORM_INTMEM +
+               USTORM_ISCSI_GLOBAL_BUF_PHYS_ADDR_OFFSET(func) + 4,
+               (u64) cp->gbl_buf_info.pg_map_arr[0] >> 32);
+
+       cnic_setup_bnx2x_context(dev);
+
+       eq_idx = CNIC_RD16(dev, BAR_CSTRORM_INTMEM +
+                          CSTORM_SB_HOST_STATUS_BLOCK_C_OFFSET(port, sb_id) +
+                          offsetof(struct cstorm_status_block_c,
+                                   index_values[HC_INDEX_C_ISCSI_EQ_CONS]));
+       if (eq_idx != 0) {
+               netdev_err(dev->netdev, "EQ cons index %x != 0\n", eq_idx);
+               return -EBUSY;
+       }
+       ret = cnic_init_bnx2x_irq(dev);
+       if (ret)
+               return ret;
+
+       cnic_init_bnx2x_tx_ring(dev);
+       cnic_init_bnx2x_rx_ring(dev);
+
+       return 0;
+}
+
+static void cnic_init_rings(struct cnic_dev *dev)
+{
+       if (test_bit(CNIC_F_BNX2_CLASS, &dev->flags)) {
+               cnic_init_bnx2_tx_ring(dev);
+               cnic_init_bnx2_rx_ring(dev);
+       } else if (test_bit(CNIC_F_BNX2X_CLASS, &dev->flags)) {
+               struct cnic_local *cp = dev->cnic_priv;
+               u32 cli = BNX2X_ISCSI_CL_ID(CNIC_E1HVN(cp));
+               union l5cm_specific_data l5_data;
+               struct ustorm_eth_rx_producers rx_prods = {0};
+               u32 off, i;
+
+               rx_prods.bd_prod = 0;
+               rx_prods.cqe_prod = BNX2X_MAX_RCQ_DESC_CNT;
+               barrier();
+
+               off = BAR_USTRORM_INTMEM +
+                       USTORM_RX_PRODS_OFFSET(CNIC_PORT(cp), cli);
+
+               for (i = 0; i < sizeof(struct ustorm_eth_rx_producers) / 4; i++)
+                       CNIC_WR(dev, off + i * 4, ((u32 *) &rx_prods)[i]);
+
+               cnic_init_bnx2x_tx_ring(dev);
+               cnic_init_bnx2x_rx_ring(dev);
+
+               l5_data.phy_address.lo = cli;
+               l5_data.phy_address.hi = 0;
+               cnic_submit_kwqe_16(dev, RAMROD_CMD_ID_ETH_CLIENT_SETUP,
+                       BNX2X_ISCSI_L2_CID, ETH_CONNECTION_TYPE, &l5_data);
+               cnic_ring_ctl(dev, BNX2X_ISCSI_L2_CID, cli, 1);
+       }
+}
+
+static void cnic_shutdown_rings(struct cnic_dev *dev)
+{
+       if (test_bit(CNIC_F_BNX2_CLASS, &dev->flags)) {
+               cnic_shutdown_bnx2_rx_ring(dev);
+       } else if (test_bit(CNIC_F_BNX2X_CLASS, &dev->flags)) {
+               struct cnic_local *cp = dev->cnic_priv;
+               u32 cli = BNX2X_ISCSI_CL_ID(CNIC_E1HVN(cp));
+               union l5cm_specific_data l5_data;
+
+               cnic_ring_ctl(dev, BNX2X_ISCSI_L2_CID, cli, 0);
+
+               l5_data.phy_address.lo = cli;
+               l5_data.phy_address.hi = 0;
+               cnic_submit_kwqe_16(dev, RAMROD_CMD_ID_ETH_HALT,
+                       BNX2X_ISCSI_L2_CID, ETH_CONNECTION_TYPE, &l5_data);
+               msleep(10);
+
+               memset(&l5_data, 0, sizeof(l5_data));
+               cnic_submit_kwqe_16(dev, RAMROD_CMD_ID_ETH_CFC_DEL,
+                       BNX2X_ISCSI_L2_CID, ETH_CONNECTION_TYPE |
+                       (1 << SPE_HDR_COMMON_RAMROD_SHIFT), &l5_data);
+               msleep(10);
+       }
+}
+
+static int cnic_register_netdev(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct cnic_eth_dev *ethdev = cp->ethdev;
+       int err;
+
+       if (!ethdev)
+               return -ENODEV;
+
+       if (ethdev->drv_state & CNIC_DRV_STATE_REGD)
+               return 0;
+
+       err = ethdev->drv_register_cnic(dev->netdev, cp->cnic_ops, dev);
+       if (err)
+               netdev_err(dev->netdev, "register_cnic failed\n");
+
+       return err;
+}
+
+static void cnic_unregister_netdev(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       struct cnic_eth_dev *ethdev = cp->ethdev;
+
+       if (!ethdev)
+               return;
+
+       ethdev->drv_unregister_cnic(dev->netdev);
+}
+
 static int cnic_start_hw(struct cnic_dev *dev)
 {
        struct cnic_local *cp = dev->cnic_priv;
@@ -2383,24 +4245,16 @@ static int cnic_start_hw(struct cnic_dev *dev)
        if (test_bit(CNIC_F_CNIC_UP, &dev->flags))
                return -EALREADY;
 
-       err = ethdev->drv_register_cnic(dev->netdev, cp->cnic_ops, dev);
-       if (err) {
-               printk(KERN_ERR PFX "%s: register_cnic failed\n",
-                      dev->netdev->name);
-               goto err2;
-       }
-
        dev->regview = ethdev->io_base;
        cp->chip_id = ethdev->chip_id;
        pci_dev_get(dev->pcidev);
        cp->func = PCI_FUNC(dev->pcidev->devfn);
-       cp->status_blk = ethdev->irq_arr[0].status_blk;
+       cp->status_blk.gen = ethdev->irq_arr[0].status_blk;
        cp->status_blk_num = ethdev->irq_arr[0].status_blk_num;
 
        err = cp->alloc_resc(dev);
        if (err) {
-               printk(KERN_ERR PFX "%s: allocate resource failure\n",
-                      dev->netdev->name);
+               netdev_err(dev->netdev, "allocate resource failure\n");
                goto err1;
        }
 
@@ -2419,18 +4273,13 @@ static int cnic_start_hw(struct cnic_dev *dev)
        return 0;
 
 err1:
-       ethdev->drv_unregister_cnic(dev->netdev);
        cp->free_resc(dev);
        pci_dev_put(dev->pcidev);
-err2:
        return err;
 }
 
 static void cnic_stop_bnx2_hw(struct cnic_dev *dev)
 {
-       struct cnic_local *cp = dev->cnic_priv;
-       struct cnic_eth_dev *ethdev = cp->ethdev;
-
        cnic_disable_bnx2_int_sync(dev);
 
        cnic_reg_wr_ind(dev, BNX2_CP_SCRATCH + 0x20, 0);
@@ -2442,8 +4291,25 @@ static void cnic_stop_bnx2_hw(struct cnic_dev *dev)
        cnic_setup_5709_context(dev, 0);
        cnic_free_irq(dev);
 
-       ethdev->drv_unregister_cnic(dev->netdev);
+       cnic_free_resc(dev);
+}
+
+
+static void cnic_stop_bnx2x_hw(struct cnic_dev *dev)
+{
+       struct cnic_local *cp = dev->cnic_priv;
+       u8 sb_id = cp->status_blk_num;
+       int port = CNIC_PORT(cp);
 
+       cnic_free_irq(dev);
+       CNIC_WR16(dev, BAR_CSTRORM_INTMEM +
+                 CSTORM_SB_HOST_STATUS_BLOCK_C_OFFSET(port, sb_id) +
+                 offsetof(struct cstorm_status_block_c,
+                          index_values[HC_INDEX_C_ISCSI_EQ_CONS]),
+                 0);
+       CNIC_WR(dev, BAR_CSTRORM_INTMEM +
+               CSTORM_ISCSI_EQ_CONS_OFFSET(cp->func, 0), 0);
+       CNIC_WR16(dev, cp->kcq_io_addr, 0);
        cnic_free_resc(dev);
 }
 
@@ -2470,10 +4336,9 @@ static void cnic_free_dev(struct cnic_dev *dev)
                i++;
        }
        if (atomic_read(&dev->ref_count) != 0)
-               printk(KERN_ERR PFX "%s: Failed waiting for ref count to go"
-                                   " to zero.\n", dev->netdev->name);
+               netdev_err(dev->netdev, "Failed waiting for ref count to go to zero\n");
 
-       printk(KERN_INFO PFX "Removed CNIC device: %s\n", dev->netdev->name);
+       netdev_info(dev->netdev, "Removed CNIC device\n");
        dev_put(dev->netdev);
        kfree(dev);
 }
@@ -2489,8 +4354,7 @@ static struct cnic_dev *cnic_alloc_dev(struct net_device *dev,
 
        cdev = kzalloc(alloc_size , GFP_KERNEL);
        if (cdev == NULL) {
-               printk(KERN_ERR PFX "%s: allocate dev struct failure\n",
-                      dev->name);
+               netdev_err(dev, "allocate dev struct failure\n");
                return NULL;
        }
 
@@ -2508,7 +4372,7 @@ static struct cnic_dev *cnic_alloc_dev(struct net_device *dev,
 
        spin_lock_init(&cp->cnic_ulp_lock);
 
-       printk(KERN_INFO PFX "Added CNIC device: %s\n", dev->name);
+       netdev_info(dev, "Added CNIC device\n");
 
        return cdev;
 }
@@ -2519,12 +4383,12 @@ static struct cnic_dev *init_bnx2_cnic(struct net_device *dev)
        struct cnic_dev *cdev;
        struct cnic_local *cp;
        struct cnic_eth_dev *ethdev = NULL;
-       struct cnic_eth_dev *(*probe)(void *) = NULL;
+       struct cnic_eth_dev *(*probe)(struct net_device *) = NULL;
 
-       probe = __symbol_get("bnx2_cnic_probe");
+       probe = symbol_get(bnx2_cnic_probe);
        if (probe) {
                ethdev = (*probe)(dev);
-               symbol_put_addr(probe);
+               symbol_put(bnx2_cnic_probe);
        }
        if (!ethdev)
                return NULL;
@@ -2578,6 +4442,57 @@ cnic_err:
        return NULL;
 }
 
+static struct cnic_dev *init_bnx2x_cnic(struct net_device *dev)
+{
+       struct pci_dev *pdev;
+       struct cnic_dev *cdev;
+       struct cnic_local *cp;
+       struct cnic_eth_dev *ethdev = NULL;
+       struct cnic_eth_dev *(*probe)(struct net_device *) = NULL;
+
+       probe = symbol_get(bnx2x_cnic_probe);
+       if (probe) {
+               ethdev = (*probe)(dev);
+               symbol_put(bnx2x_cnic_probe);
+       }
+       if (!ethdev)
+               return NULL;
+
+       pdev = ethdev->pdev;
+       if (!pdev)
+               return NULL;
+
+       dev_hold(dev);
+       cdev = cnic_alloc_dev(dev, pdev);
+       if (cdev == NULL) {
+               dev_put(dev);
+               return NULL;
+       }
+
+       set_bit(CNIC_F_BNX2X_CLASS, &cdev->flags);
+       cdev->submit_kwqes = cnic_submit_bnx2x_kwqes;
+
+       cp = cdev->cnic_priv;
+       cp->ethdev = ethdev;
+       cdev->pcidev = pdev;
+
+       cp->cnic_ops = &cnic_bnx2x_ops;
+       cp->start_hw = cnic_start_bnx2x_hw;
+       cp->stop_hw = cnic_stop_bnx2x_hw;
+       cp->setup_pgtbl = cnic_setup_page_tbl_le;
+       cp->alloc_resc = cnic_alloc_bnx2x_resc;
+       cp->free_resc = cnic_free_resc;
+       cp->start_cm = cnic_cm_init_bnx2x_hw;
+       cp->stop_cm = cnic_cm_stop_bnx2x_hw;
+       cp->enable_int = cnic_enable_bnx2x_int;
+       cp->disable_int_sync = cnic_disable_bnx2x_int_sync;
+       cp->ack_int = cnic_ack_bnx2x_msix;
+       cp->close_conn = cnic_close_bnx2x_conn;
+       cp->next_idx = cnic_bnx2x_next_idx;
+       cp->hw_idx = cnic_bnx2x_hw_idx;
+       return cdev;
+}
+
 static struct cnic_dev *is_cnic_dev(struct net_device *dev)
 {
        struct ethtool_drvinfo drvinfo;
@@ -2589,6 +4504,8 @@ static struct cnic_dev *is_cnic_dev(struct net_device *dev)
 
                if (!strcmp(drvinfo.driver, "bnx2"))
                        cdev = init_bnx2_cnic(dev);
+               if (!strcmp(drvinfo.driver, "bnx2x"))
+                       cdev = init_bnx2x_cnic(dev);
                if (cdev) {
                        write_lock(&cnic_dev_lock);
                        list_add(&cdev->list, &cnic_dev_list);
@@ -2626,11 +4543,14 @@ static int cnic_netdev_event(struct notifier_block *this, unsigned long event,
                        cnic_ulp_init(dev);
                else if (event == NETDEV_UNREGISTER)
                        cnic_ulp_exit(dev);
-               else if (event == NETDEV_UP) {
-                       mutex_lock(&cnic_lock);
+
+               if (event == NETDEV_UP) {
+                       if (cnic_register_netdev(dev) != 0) {
+                               cnic_put(dev);
+                               goto done;
+                       }
                        if (!cnic_start_hw(dev))
                                cnic_ulp_start(dev);
-                       mutex_unlock(&cnic_lock);
                }
 
                rcu_read_lock();
@@ -2649,10 +4569,9 @@ static int cnic_netdev_event(struct notifier_block *this, unsigned long event,
                rcu_read_unlock();
 
                if (event == NETDEV_GOING_DOWN) {
-                       mutex_lock(&cnic_lock);
                        cnic_ulp_stop(dev);
                        cnic_stop_hw(dev);
-                       mutex_unlock(&cnic_lock);
+                       cnic_unregister_netdev(dev);
                } else if (event == NETDEV_UNREGISTER) {
                        write_lock(&cnic_dev_lock);
                        list_del_init(&dev->list);
@@ -2684,6 +4603,7 @@ static void cnic_release(void)
                }
 
                cnic_ulp_exit(dev);
+               cnic_unregister_netdev(dev);
                list_del_init(&dev->list);
                cnic_free_dev(dev);
        }
@@ -2693,7 +4613,7 @@ static int __init cnic_init(void)
 {
        int rc = 0;
 
-       printk(KERN_INFO "%s", version);
+       pr_info("%s", version);
 
        rc = register_netdevice_notifier(&cnic_netdev_notifier);
        if (rc) {