linux/drivers/net/cnic.c

/* cnic.c: Broadcom CNIC core network driver.
 *
 * 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
 * the Free Software Foundation.
 *
 * Original skeleton written by: John(Zongxi) Chen (zongxi@broadcom.com)
 * 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 <linux/errno.h>
#include <linux/list.h>
#include <linux/slab.h>
#include <linux/pci.h>
#include <linux/init.h>
#include <linux/netdevice.h>
#include <linux/uio_driver.h>
#include <linux/in.h>
#include <linux/dma-mapping.h>
#include <linux/delay.h>
#include <linux/ethtool.h>
#include <linux/if_vlan.h>
#if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
#define BCM_VLAN 1
#endif
#include <net/ip.h>
#include <net/tcp.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/bnx2x_reg.h"
#include "bnx2x/bnx2x_fw_defs.h"
#include "bnx2x/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"

static char version[] __devinitdata =
	"Broadcom NetXtreme II CNIC Driver " DRV_MODULE_NAME " v" CNIC_MODULE_VERSION " (" CNIC_MODULE_RELDATE ")\n";

MODULE_AUTHOR("Michael Chan <mchan@broadcom.com> and John(Zongxi) "
	      "Chen (zongxi@broadcom.com");
MODULE_DESCRIPTION("Broadcom NetXtreme II CNIC Driver");
MODULE_LICENSE("GPL");
MODULE_VERSION(CNIC_MODULE_VERSION);

static LIST_HEAD(cnic_dev_list);
static LIST_HEAD(cnic_udev_list);
static DEFINE_RWLOCK(cnic_dev_lock);
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 = {
	.cnic_owner	= THIS_MODULE,
	.cnic_handler	= cnic_service_bnx2,
	.cnic_ctl	= cnic_ctl,
};

static struct cnic_ops cnic_bnx2x_ops = {
	.cnic_owner	= THIS_MODULE,
	.cnic_handler	= cnic_service_bnx2x,
	.cnic_ctl	= cnic_ctl,
};

static struct workqueue_struct *cnic_wq;

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)
{
	struct cnic_uio_dev *udev = uinfo->priv;
	struct cnic_dev *dev;

	if (!capable(CAP_NET_ADMIN))
		return -EPERM;

	if (udev->uio_dev != -1)
		return -EBUSY;

	rtnl_lock();
	dev = udev->dev;

	if (!dev || !test_bit(CNIC_F_CNIC_UP, &dev->flags)) {
		rtnl_unlock();
		return -ENODEV;
	}

	udev->uio_dev = iminor(inode);

	cnic_shutdown_rings(dev);
	cnic_init_rings(dev);
	rtnl_unlock();

	return 0;
}

static int cnic_uio_close(struct uio_info *uinfo, struct inode *inode)
{
	struct cnic_uio_dev *udev = uinfo->priv;

	udev->uio_dev = -1;
	return 0;
}

static inline void cnic_hold(struct cnic_dev *dev)
{
	atomic_inc(&dev->ref_count);
}

static inline void cnic_put(struct cnic_dev *dev)
{
	atomic_dec(&dev->ref_count);
}

static inline void csk_hold(struct cnic_sock *csk)
{
	atomic_inc(&csk->ref_count);
}

static inline void csk_put(struct cnic_sock *csk)
{
	atomic_dec(&csk->ref_count);
}

static struct cnic_dev *cnic_from_netdev(struct net_device *netdev)
{
	struct cnic_dev *cdev;

	read_lock(&cnic_dev_lock);
	list_for_each_entry(cdev, &cnic_dev_list, list) {
		if (netdev == cdev->netdev) {
			cnic_hold(cdev);
			read_unlock(&cnic_dev_lock);
			return cdev;
		}
	}
	read_unlock(&cnic_dev_lock);
	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;
	struct cnic_eth_dev *ethdev = cp->ethdev;
	struct drv_ctl_info info;
	struct drv_ctl_io *io = &info.data.io;

	info.cmd = DRV_CTL_CTX_WR_CMD;
	io->cid_addr = cid_addr;
	io->offset = off;
	io->data = 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;
	struct cnic_eth_dev *ethdev = cp->ethdev;
	struct drv_ctl_info info;
	struct drv_ctl_io *io = &info.data.io;

	info.cmd = DRV_CTL_IO_WR_CMD;
	io->offset = off;
	io->data = val;
	ethdev->drv_ctl(dev->netdev, &info);
}

static u32 cnic_reg_rd_ind(struct cnic_dev *dev, u32 off)
{
	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_IO_RD_CMD;
	io->offset = off;
	ethdev->drv_ctl(dev->netdev, &info);
	return io->data;
}

static int cnic_in_use(struct cnic_sock *csk)
{
	return test_bit(SK_F_INUSE, &csk->flags);
}

static void cnic_spq_completion(struct cnic_dev *dev, int cmd, u32 count)
{
	struct cnic_local *cp = dev->cnic_priv;
	struct cnic_eth_dev *ethdev = cp->ethdev;
	struct drv_ctl_info info;

	info.cmd = cmd;
	info.data.credit.credit_count = 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 < cp->max_cid_space; 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)
{
	struct iscsi_path path_req;
	char *buf = NULL;
	u16 len = 0;
	u32 msg_type = ISCSI_KEVENT_IF_DOWN;
	struct cnic_ulp_ops *ulp_ops;
	struct cnic_uio_dev *udev = cp->udev;

	if (!udev || udev->uio_dev == -1)
		return -ENODEV;

	if (csk) {
		len = sizeof(path_req);
		buf = (char *) &path_req;
		memset(&path_req, 0, len);

		msg_type = ISCSI_KEVENT_PATH_REQ;
		path_req.handle = (u64) csk->l5_cid;
		if (test_bit(SK_F_IPV6, &csk->flags)) {
			memcpy(&path_req.dst.v6_addr, &csk->dst_ip[0],
			       sizeof(struct in6_addr));
			path_req.ip_addr_len = 16;
		} else {
			memcpy(&path_req.dst.v4_addr, &csk->dst_ip[0],
			       sizeof(struct in_addr));
			path_req.ip_addr_len = 4;
		}
		path_req.vlan_id = csk->vlan_id;
		path_req.pmtu = csk->mtu;
	}

	rcu_read_lock();
	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();
	return 0;
}

static int cnic_iscsi_nl_msg_recv(struct cnic_dev *dev, u32 msg_type,
				  char *buf, u16 len)
{
	int rc = -EINVAL;

	switch (msg_type) {
	case ISCSI_UEVENT_PATH_UPDATE: {
		struct cnic_local *cp;
		u32 l5_cid;
		struct cnic_sock *csk;
		struct iscsi_path *path_resp;

		if (len < sizeof(*path_resp))
			break;

		path_resp = (struct iscsi_path *) buf;
		cp = dev->cnic_priv;
		l5_cid = (u32) path_resp->handle;
		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)) {
			memcpy(csk->ha, path_resp->mac_addr, 6);
			if (test_bit(SK_F_IPV6, &csk->flags))
				memcpy(&csk->src_ip[0], &path_resp->src.v6_addr,
				       sizeof(struct in6_addr));
			else
				memcpy(&csk->src_ip[0], &path_resp->src.v4_addr,
				       sizeof(struct in_addr));
			if (is_valid_ether_addr(csk->ha))
				cnic_cm_set_pg(csk);
		}
		csk_put(csk);
		rcu_read_unlock();
		rc = 0;
	}
	}

	return rc;
}

static int cnic_offld_prep(struct cnic_sock *csk)
{
	if (test_and_set_bit(SK_F_OFFLD_SCHED, &csk->flags))
		return 0;

	if (!test_bit(SK_F_CONNECT_START, &csk->flags)) {
		clear_bit(SK_F_OFFLD_SCHED, &csk->flags);
		return 0;
	}

	return 1;
}

static int cnic_close_prep(struct cnic_sock *csk)
{
	clear_bit(SK_F_CONNECT_START, &csk->flags);
	smp_mb__after_clear_bit();

	if (test_and_clear_bit(SK_F_OFFLD_COMPLETE, &csk->flags)) {
		while (test_and_set_bit(SK_F_OFFLD_SCHED, &csk->flags))
			msleep(1);

		return 1;
	}
	return 0;
}

static int cnic_abort_prep(struct cnic_sock *csk)
{
	clear_bit(SK_F_CONNECT_START, &csk->flags);
	smp_mb__after_clear_bit();

	while (test_and_set_bit(SK_F_OFFLD_SCHED, &csk->flags))
		msleep(1);

	if (test_and_clear_bit(SK_F_OFFLD_COMPLETE, &csk->flags)) {
		csk->state = L4_KCQE_OPCODE_VALUE_RESET_COMP;
		return 1;
	}

	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;

		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 < 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]) {
		pr_err("%s: Type %d has already been registered\n",
		       __func__, ulp_type);
		mutex_unlock(&cnic_lock);
		return -EBUSY;
	}

	read_lock(&cnic_dev_lock);
	list_for_each_entry(dev, &cnic_dev_list, list) {
		struct cnic_local *cp = dev->cnic_priv;

		clear_bit(ULP_F_INIT, &cp->ulp_flags[ulp_type]);
	}
	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);

	/* Prevent race conditions with netdev_event */
	rtnl_lock();
	read_lock(&cnic_dev_lock);
	list_for_each_entry(dev, &cnic_dev_list, list) {
		struct cnic_local *cp = dev->cnic_priv;

		if (!test_and_set_bit(ULP_F_INIT, &cp->ulp_flags[ulp_type]))
			ulp_ops->cnic_init(dev);
	}
	read_unlock(&cnic_dev_lock);
	rtnl_unlock();

	return 0;
}

int cnic_unregister_driver(int ulp_type)
{
	struct cnic_dev *dev;
	struct cnic_ulp_ops *ulp_ops;
	int i = 0;

	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);
	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);
	list_for_each_entry(dev, &cnic_dev_list, list) {
		struct cnic_local *cp = dev->cnic_priv;

		if (rcu_dereference(cp->ulp_ops[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:
	mutex_unlock(&cnic_lock);
	return -EINVAL;
}

static int cnic_start_hw(struct cnic_dev *);
static void cnic_stop_hw(struct cnic_dev *);

static int cnic_register_device(struct cnic_dev *dev, int ulp_type,
				void *ulp_ctx)
{
	struct cnic_local *cp = dev->cnic_priv;
	struct cnic_ulp_ops *ulp_ops;

	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) {
		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])) {
		pr_err("%s: Type %d has already been registered to this device\n",
		       __func__, ulp_type);
		mutex_unlock(&cnic_lock);
		return -EBUSY;
	}

	clear_bit(ULP_F_START, &cp->ulp_flags[ulp_type]);
	cp->ulp_handle[ulp_type] = ulp_ctx;
	ulp_ops = cnic_ulp_tbl[ulp_type];
	rcu_assign_pointer(cp->ulp_ops[ulp_type], ulp_ops);
	cnic_hold(dev);

	if (test_bit(CNIC_F_CNIC_UP, &dev->flags))
		if (!test_and_set_bit(ULP_F_START, &cp->ulp_flags[ulp_type]))
			ulp_ops->cnic_start(cp->ulp_handle[ulp_type]);

	mutex_unlock(&cnic_lock);

	return 0;

}
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 < 0 || ulp_type >= MAX_CNIC_ULP_TYPE) {
		pr_err("%s: Bad type %d\n", __func__, ulp_type);
		return -EINVAL;
	}
	mutex_lock(&cnic_lock);
	if (rcu_dereference(cp->ulp_ops[ulp_type])) {
		rcu_assign_pointer(cp->ulp_ops[ulp_type], NULL);
		cnic_put(dev);
	} else {
		pr_err("%s: device not registered to this ulp type %d\n",
		       __func__, ulp_type);
		mutex_unlock(&cnic_lock);
		return -EINVAL;
	}
	mutex_unlock(&cnic_lock);

	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);

static int cnic_init_id_tbl(struct cnic_id_tbl *id_tbl, u32 size, u32 start_id)
{
	id_tbl->start = start_id;
	id_tbl->max = size;
	id_tbl->next = 0;
	spin_lock_init(&id_tbl->lock);
	id_tbl->table = kzalloc(DIV_ROUND_UP(size, 32) * 4, GFP_KERNEL);
	if (!id_tbl->table)
		return -ENOMEM;

	return 0;
}

static void cnic_free_id_tbl(struct cnic_id_tbl *id_tbl)
{
	kfree(id_tbl->table);
	id_tbl->table = NULL;
}

static int cnic_alloc_id(struct cnic_id_tbl *id_tbl, u32 id)
{
	int ret = -1;

	id -= id_tbl->start;
	if (id >= id_tbl->max)
		return ret;

	spin_lock(&id_tbl->lock);
	if (!test_bit(id, id_tbl->table)) {
		set_bit(id, id_tbl->table);
		ret = 0;
	}
	spin_unlock(&id_tbl->lock);
	return ret;
}

/* Returns -1 if not successful */
static u32 cnic_alloc_new_id(struct cnic_id_tbl *id_tbl)
{
	u32 id;

	spin_lock(&id_tbl->lock);
	id = find_next_zero_bit(id_tbl->table, id_tbl->max, id_tbl->next);
	if (id >= id_tbl->max) {
		id = -1;
		if (id_tbl->next != 0) {
			id = find_first_zero_bit(id_tbl->table, id_tbl->next);
			if (id >= id_tbl->next)
				id = -1;
		}
	}

	if (id < id_tbl->max) {
		set_bit(id, id_tbl->table);
		id_tbl->next = (id + 1) & (id_tbl->max - 1);
		id += id_tbl->start;
	}

	spin_unlock(&id_tbl->lock);

	return id;
}

static void cnic_free_id(struct cnic_id_tbl *id_tbl, u32 id)
{
	if (id == -1)
		return;

	id -= id_tbl->start;
	if (id >= id_tbl->max)
		return;

	clear_bit(id, id_tbl->table);
}

static void cnic_free_dma(struct cnic_dev *dev, struct cnic_dma *dma)
{
	int i;

	if (!dma->pg_arr)
		return;

	for (i = 0; i < dma->num_pages; i++) {
		if (dma->pg_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) {
		dma_free_coherent(&dev->pcidev->dev, dma->pgtbl_size,
				  dma->pgtbl, dma->pgtbl_map);
		dma->pgtbl = NULL;
	}
	kfree(dma->pg_arr);
	dma->pg_arr = NULL;
	dma->num_pages = 0;
}

static void cnic_setup_page_tbl(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 big endian format. */
		*page_table = (u32) ((u64) dma->pg_map_arr[i] >> 32);
		page_table++;
		*page_table = (u32) dma->pg_map_arr[i];
		page_table++;
	}
}

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)
{
	int i, size;
	struct cnic_local *cp = dev->cnic_priv;

	size = pages * (sizeof(void *) + sizeof(dma_addr_t));
	dma->pg_arr = kzalloc(size, GFP_ATOMIC);
	if (dma->pg_arr == NULL)
		return -ENOMEM;

	dma->pg_map_arr = (dma_addr_t *) (dma->pg_arr + pages);
	dma->num_pages = pages;

	for (i = 0; i < pages; 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;
	}
	if (!use_pg_tbl)
		return 0;

	dma->pgtbl_size = ((pages * 8) + BCM_PAGE_SIZE - 1) &
			  ~(BCM_PAGE_SIZE - 1);
	dma->pgtbl = dma_alloc_coherent(&dev->pcidev->dev, dma->pgtbl_size,
					&dma->pgtbl_map, GFP_ATOMIC);
	if (dma->pgtbl == NULL)
		goto error;

	cp->setup_pgtbl(dev, dma);

	return 0;

error:
	cnic_free_dma(dev, dma);
	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_uio(struct cnic_uio_dev *udev)
{
	uio_unregister_device(&udev->cnic_uinfo);

	if (udev->l2_buf) {
		dma_free_coherent(&udev->pdev->dev, udev->l2_buf_size,
				  udev->l2_buf, udev->l2_buf_map);
		udev->l2_buf = NULL;
	}

	if (udev->l2_ring) {
		dma_free_coherent(&udev->pdev->dev, udev->l2_ring_size,
				  udev->l2_ring, udev->l2_ring_map);
		udev->l2_ring = NULL;
	}

	pci_dev_put(udev->pdev);
	kfree(udev);
}

static void cnic_free_uio(struct cnic_uio_dev *udev)
{
	if (!udev)
		return;

	write_lock(&cnic_dev_lock);
	list_del_init(&udev->list);
	write_unlock(&cnic_dev_lock);
	__cnic_free_uio(udev);
}

static void cnic_free_resc(struct cnic_dev *dev)
{
	struct cnic_local *cp = dev->cnic_priv;
	struct cnic_uio_dev *udev = cp->udev;

	if (udev) {
		udev->dev = NULL;
		cp->udev = NULL;
	}

	cnic_free_context(dev);
	kfree(cp->ctx_arr);
	cp->ctx_arr = NULL;
	cp->ctx_blks = 0;

	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->kcq1.dma);
	kfree(cp->iscsi_tbl);
	cp->iscsi_tbl = NULL;
	kfree(cp->ctx_tbl);
	cp->ctx_tbl = NULL;

	cnic_free_id_tbl(&cp->cid_tbl);
}

static int cnic_alloc_context(struct cnic_dev *dev)
{
	struct cnic_local *cp = dev->cnic_priv;

	if (CHIP_NUM(cp) == CHIP_NUM_5709) {
		int i, k, arr_size;

		cp->ctx_blk_size = BCM_PAGE_SIZE;
		cp->cids_per_blk = BCM_PAGE_SIZE / 128;
		arr_size = BNX2_MAX_CID / cp->cids_per_blk *
			   sizeof(struct cnic_ctx);
		cp->ctx_arr = kzalloc(arr_size, GFP_KERNEL);
		if (cp->ctx_arr == NULL)
			return -ENOMEM;

		k = 0;
		for (i = 0; i < 2; i++) {
			u32 j, reg, off, lo, hi;

			if (i == 0)
				off = BNX2_PG_CTX_MAP;
			else
				off = BNX2_ISCSI_CTX_MAP;

			reg = cnic_reg_rd_ind(dev, off);
			lo = reg >> 16;
			hi = reg & 0xffff;
			for (j = lo; j < hi; j += cp->cids_per_blk, k++)
				cp->ctx_arr[k].cid = j;
		}

		cp->ctx_blks = k;
		if (cp->ctx_blks >= (BNX2_MAX_CID / cp->cids_per_blk)) {
			cp->ctx_blks = 0;
			return -ENOMEM;
		}

		for (i = 0; i < cp->ctx_blks; i++) {
			cp->ctx_arr[i].ctx =
				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;
		}
	}
	return 0;
}

static int cnic_alloc_kcq(struct cnic_dev *dev, struct kcq_info *info)
{
	int err, i, is_bnx2 = 0;
	struct kcqe **kcq;

	if (test_bit(CNIC_F_BNX2_CLASS, &dev->flags))
		is_bnx2 = 1;

	err = cnic_alloc_dma(dev, &info->dma, KCQ_PAGE_CNT, is_bnx2);
	if (err)
		return err;

	kcq = (struct kcqe **) info->dma.pg_arr;
	info->kcq = kcq;

	if (is_bnx2)
		return 0;

	for (i = 0; i < KCQ_PAGE_CNT; i++) {
		struct bnx2x_bd_chain_next *next =
			(struct bnx2x_bd_chain_next *) &kcq[i][MAX_KCQE_CNT];
		int j = i + 1;

		if (j >= KCQ_PAGE_CNT)
			j = 0;
		next->addr_hi = (u64) info->dma.pg_map_arr[j] >> 32;
		next->addr_lo = info->dma.pg_map_arr[j] & 0xffffffff;
	}
	return 0;
}

static int cnic_alloc_uio_rings(struct cnic_dev *dev, int pages)
{
	struct cnic_local *cp = dev->cnic_priv;
	struct cnic_uio_dev *udev;

	read_lock(&cnic_dev_lock);
	list_for_each_entry(udev, &cnic_udev_list, list) {
		if (udev->pdev == dev->pcidev) {
			udev->dev = dev;
			cp->udev = udev;
			read_unlock(&cnic_dev_lock);
			return 0;
		}
	}
	read_unlock(&cnic_dev_lock);

	udev = kzalloc(sizeof(struct cnic_uio_dev), GFP_ATOMIC);
	if (!udev)
		return -ENOMEM;

	udev->uio_dev = -1;

	udev->dev = dev;
	udev->pdev = dev->pcidev;
	udev->l2_ring_size = pages * BCM_PAGE_SIZE;
	udev->l2_ring = dma_alloc_coherent(&udev->pdev->dev, udev->l2_ring_size,
					   &udev->l2_ring_map,
					   GFP_KERNEL | __GFP_COMP);
	if (!udev->l2_ring)
		goto err_udev;

	udev->l2_buf_size = (cp->l2_rx_ring_size + 1) * cp->l2_single_buf_size;
	udev->l2_buf_size = PAGE_ALIGN(udev->l2_buf_size);
	udev->l2_buf = dma_alloc_coherent(&udev->pdev->dev, udev->l2_buf_size,
					  &udev->l2_buf_map,
					  GFP_KERNEL | __GFP_COMP);
	if (!udev->l2_buf)
		goto err_dma;

	write_lock(&cnic_dev_lock);
	list_add(&udev->list, &cnic_udev_list);
	write_unlock(&cnic_dev_lock);

	pci_dev_get(udev->pdev);

	cp->udev = udev;

	return 0;
 err_dma:
	dma_free_coherent(&udev->pdev->dev, udev->l2_ring_size,
			  udev->l2_ring, udev->l2_ring_map);
 err_udev:
	kfree(udev);
	return -ENOMEM;
}

static int cnic_init_uio(struct cnic_dev *dev)
{
	struct cnic_local *cp = dev->cnic_priv;
	struct cnic_uio_dev *udev = cp->udev;
	struct uio_info *uinfo;
	int ret = 0;

	if (!udev)
		return -ENOMEM;

	uinfo = &udev->cnic_uinfo;

	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;

	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(*cp->bnx2x_def_status_blk);

		uinfo->name = "bnx2x_cnic";
	}

	uinfo->mem[1].memtype = UIO_MEM_LOGICAL;

	uinfo->mem[2].addr = (unsigned long) udev->l2_ring;
	uinfo->mem[2].size = udev->l2_ring_size;
	uinfo->mem[2].memtype = UIO_MEM_LOGICAL;

	uinfo->mem[3].addr = (unsigned long) udev->l2_buf;
	uinfo->mem[3].size = udev->l2_buf_size;
	uinfo->mem[3].memtype = UIO_MEM_LOGICAL;

	uinfo->version = CNIC_MODULE_VERSION;
	uinfo->irq = UIO_IRQ_CUSTOM;

	uinfo->open = cnic_uio_open;
	uinfo->release = cnic_uio_close;

	if (udev->uio_dev == -1) {
		if (!uinfo->priv) {
			uinfo->priv = udev;

			ret = uio_register_device(&udev->pdev->dev, uinfo);
		}
	} else {
		cnic_init_rings(dev);
	}

	return ret;
}

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_kcq(dev, &cp->kcq1);
	if (ret)
		goto error;

	ret = cnic_alloc_context(dev);
	if (ret)
		goto error;

	ret = cnic_alloc_uio_rings(dev, 2);
	if (ret)
		goto error;

	ret = cnic_init_uio(dev);
	if (ret)
		goto error;

	return 0;

error:
	cnic_free_resc(dev);
	return ret;
}

static int cnic_alloc_bnx2x_context(struct cnic_dev *dev)
{
	struct cnic_local *cp = dev->cnic_priv;
	int ctx_blk_size = cp->ethdev->ctx_blk_size;
	int total_mem, blks, i;

	total_mem = BNX2X_CONTEXT_MEM_SIZE * cp->max_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 = kcalloc(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_57710(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;
	struct cnic_eth_dev *ethdev = cp->ethdev;
	u32 start_cid = ethdev->starting_cid;
	int i, j, n, ret, pages;
	struct cnic_dma *kwq_16_dma = &cp->kwq_16_data_info;

	cp->iro_arr = ethdev->iro_arr;

	cp->max_cid_space = MAX_ISCSI_TBL_SZ;
	cp->iscsi_start_cid = start_cid;
	if (start_cid < BNX2X_ISCSI_START_CID) {
		u32 delta = BNX2X_ISCSI_START_CID - start_cid;

		cp->iscsi_start_cid = BNX2X_ISCSI_START_CID;
		cp->max_cid_space += delta;
	}

	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) *
				cp->max_cid_space, 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(cp->max_cid_space * 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 < cp->max_cid_space; 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_kcq(dev, &cp->kcq1);
	if (ret)
		goto error;

	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;

	cp->l2_rx_ring_size = 15;

	ret = cnic_alloc_uio_rings(dev, 4);
	if (ret)
		goto error;

	ret = cnic_init_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 -
		((cp->kwq_prod_idx - cp->kwq_con_idx) & cp->max_kwq_idx);
}

static int cnic_submit_bnx2_kwqes(struct cnic_dev *dev, struct kwqe *wqes[],
				  u32 num_wqes)
{
	struct cnic_local *cp = dev->cnic_priv;
	struct kwqe *prod_qe;
	u16 prod, sw_prod, i;

	if (!test_bit(CNIC_F_CNIC_UP, &dev->flags))
		return -EAGAIN;		/* bnx2 is down */

	spin_lock_bh(&cp->cnic_ulp_lock);
	if (num_wqes > cnic_kwq_avail(cp) &&
	    !test_bit(CNIC_LCL_FL_KWQ_INIT, &cp->cnic_local_flags)) {
		spin_unlock_bh(&cp->cnic_ulp_lock);
		return -EAGAIN;
	}

	clear_bit(CNIC_LCL_FL_KWQ_INIT, &cp->cnic_local_flags);

	prod = cp->kwq_prod_idx;
	sw_prod = prod & MAX_KWQ_IDX;
	for (i = 0; i < num_wqes; i++) {
		prod_qe = &cp->kwq[KWQ_PG(sw_prod)][KWQ_IDX(sw_prod)];
		memcpy(prod_qe, wqes[i], sizeof(struct kwqe));
		prod++;
		sw_prod = prod & MAX_KWQ_IDX;
	}
	cp->kwq_prod_idx = prod;

	CNIC_WR16(dev, cp->kwq_io_addr, cp->kwq_prod_idx);

	spin_unlock_bh(&cp->cnic_ulp_lock);
	return 0;
}

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;
	struct l5cm_spe kwqe;
	struct kwqe_16 *kwq[1];
	int ret;

	kwqe.hdr.conn_and_cmd_data =
		cpu_to_le32(((cmd << SPE_HDR_CMD_ID_SHIFT) |
			     BNX2X_HW_CID(cp, cid)));
	kwqe.hdr.type = cpu_to_le16(type);
	kwqe.hdr.reserved1 = 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);

	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 hq_bds, pages;
	u32 pfid = cp->pfid;

	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(pfid),
		  req1->rq_num_wqes);
	CNIC_WR16(dev, BAR_TSTRORM_INTMEM + TSTORM_ISCSI_PAGE_SIZE_OFFSET(pfid),
		  PAGE_SIZE);
	CNIC_WR8(dev, BAR_TSTRORM_INTMEM +
		 TSTORM_ISCSI_PAGE_SIZE_LOG_OFFSET(pfid), PAGE_SHIFT);
	CNIC_WR16(dev, BAR_TSTRORM_INTMEM +
		  TSTORM_ISCSI_NUM_OF_TASKS_OFFSET(pfid),
		  req1->num_tasks_per_conn);

	/* init Ustorm RAM */
	CNIC_WR16(dev, BAR_USTRORM_INTMEM +
		  USTORM_ISCSI_RQ_BUFFER_SIZE_OFFSET(pfid),
		  req1->rq_buffer_size);
	CNIC_WR16(dev, BAR_USTRORM_INTMEM + USTORM_ISCSI_PAGE_SIZE_OFFSET(pfid),
		  PAGE_SIZE);
	CNIC_WR8(dev, BAR_USTRORM_INTMEM +
		 USTORM_ISCSI_PAGE_SIZE_LOG_OFFSET(pfid), PAGE_SHIFT);
	CNIC_WR16(dev, BAR_USTRORM_INTMEM +
		  USTORM_ISCSI_NUM_OF_TASKS_OFFSET(pfid),
		  req1->num_tasks_per_conn);
	CNIC_WR16(dev, BAR_USTRORM_INTMEM + USTORM_ISCSI_RQ_SIZE_OFFSET(pfid),
		  req1->rq_num_wqes);
	CNIC_WR16(dev, BAR_USTRORM_INTMEM + USTORM_ISCSI_CQ_SIZE_OFFSET(pfid),
		  req1->cq_num_wqes);
	CNIC_WR16(dev, BAR_USTRORM_INTMEM + USTORM_ISCSI_R2TQ_SIZE_OFFSET(pfid),
		  cp->num_iscsi_tasks * BNX2X_ISCSI_MAX_PENDING_R2TS);

	/* init Xstorm RAM */
	CNIC_WR16(dev, BAR_XSTRORM_INTMEM + XSTORM_ISCSI_PAGE_SIZE_OFFSET(pfid),
		  PAGE_SIZE);
	CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
		 XSTORM_ISCSI_PAGE_SIZE_LOG_OFFSET(pfid), PAGE_SHIFT);
	CNIC_WR16(dev, BAR_XSTRORM_INTMEM +
		  XSTORM_ISCSI_NUM_OF_TASKS_OFFSET(pfid),
		  req1->num_tasks_per_conn);
	CNIC_WR16(dev, BAR_XSTRORM_INTMEM + XSTORM_ISCSI_HQ_SIZE_OFFSET(pfid),
		  hq_bds);
	CNIC_WR16(dev, BAR_XSTRORM_INTMEM + XSTORM_ISCSI_SQ_SIZE_OFFSET(pfid),
		  req1->num_tasks_per_conn);
	CNIC_WR16(dev, BAR_XSTRORM_INTMEM + XSTORM_ISCSI_R2TQ_SIZE_OFFSET(pfid),
		  cp->num_iscsi_tasks * BNX2X_ISCSI_MAX_PENDING_R2TS);

	/* init Cstorm RAM */
	CNIC_WR16(dev, BAR_CSTRORM_INTMEM + CSTORM_ISCSI_PAGE_SIZE_OFFSET(pfid),
		  PAGE_SIZE);
	CNIC_WR8(dev, BAR_CSTRORM_INTMEM +
		 CSTORM_ISCSI_PAGE_SIZE_LOG_OFFSET(pfid), PAGE_SHIFT);
	CNIC_WR16(dev, BAR_CSTRORM_INTMEM +
		  CSTORM_ISCSI_NUM_OF_TASKS_OFFSET(pfid),
		  req1->num_tasks_per_conn);
	CNIC_WR16(dev, BAR_CSTRORM_INTMEM + CSTORM_ISCSI_CQ_SIZE_OFFSET(pfid),
		  req1->cq_num_wqes);
	CNIC_WR16(dev, BAR_CSTRORM_INTMEM + CSTORM_ISCSI_HQ_SIZE_OFFSET(pfid),
		  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;
	u32 pfid = cp->pfid;
	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(pfid), req2->error_bit_map[0]);
	CNIC_WR(dev, BAR_TSTRORM_INTMEM +
		TSTORM_ISCSI_ERROR_BITMAP_OFFSET(pfid) + 4,
		req2->error_bit_map[1]);

	CNIC_WR16(dev, BAR_USTRORM_INTMEM +
		  USTORM_ISCSI_CQ_SQN_SIZE_OFFSET(pfid), req2->max_cq_sqn);
	CNIC_WR(dev, BAR_USTRORM_INTMEM +
		USTORM_ISCSI_ERROR_BITMAP_OFFSET(pfid), req2->error_bit_map[0]);
	CNIC_WR(dev, BAR_USTRORM_INTMEM +
		USTORM_ISCSI_ERROR_BITMAP_OFFSET(pfid) + 4,
		req2->error_bit_map[1]);

	CNIC_WR16(dev, BAR_CSTRORM_INTMEM +
		  CSTORM_ISCSI_CQ_SQN_SIZE_OFFSET(pfid), 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(cp, cid);
	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->tstorm_st_context.tcp.ooo_support_mode =
		TCP_TSTORM_OOO_DROP_AND_PROC_ACK;

	ictx->timers_context.flags |= 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 cnic_context *ctx;
	struct iscsi_kcqe kcqe;
	struct kcqe *cqes[1];
	u32 l5_cid;
	int ret = 0;

	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;

	ctx = &cp->ctx_tbl[l5_cid];
	if (test_bit(CTX_FL_OFFLD_START, &ctx->ctx_flags)) {
		kcqe.completion_status =
			ISCSI_KCQE_COMPLETION_STATUS_CID_BUSY;
		goto done;
	}

	if (atomic_inc_return(&cp->iscsi_conn) > dev->max_iscsi_conn) {
		atomic_dec(&cp->iscsi_conn);
		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, cp->ctx_tbl[l5_cid].cid);

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_destroy_ramrod(struct cnic_dev *dev, u32 l5_cid)
{
	struct cnic_local *cp = dev->cnic_priv;
	struct cnic_context *ctx = &cp->ctx_tbl[l5_cid];
	union l5cm_specific_data l5_data;
	int ret;
	u32 hw_cid, type;

	init_waitqueue_head(&ctx->waitq);
	ctx->wait_cond = 0;
	memset(&l5_data, 0, sizeof(l5_data));
	hw_cid = BNX2X_HW_CID(cp, ctx->cid);
	type = (NONE_CONNECTION_TYPE << SPE_HDR_CONN_TYPE_SHIFT)
		& SPE_HDR_CONN_TYPE;
	type |= ((cp->pfid << SPE_HDR_FUNCTION_ID_SHIFT) &
		 SPE_HDR_FUNCTION_ID);

	ret = cnic_submit_kwqe_16(dev, RAMROD_CMD_ID_COMMON_CFC_DEL,
				  hw_cid, type, &l5_data);

	if (ret == 0)
		wait_event(ctx->waitq, ctx->wait_cond);

	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;
	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 (!test_bit(CTX_FL_OFFLD_START, &ctx->ctx_flags))
		goto skip_cfc_delete;

	if (!time_after(jiffies, ctx->timestamp + (2 * HZ))) {
		unsigned long delta = ctx->timestamp + (2 * HZ) - jiffies;

		if (delta > (2 * HZ))
			delta = 0;

		set_bit(CTX_FL_DELETE_WAIT, &ctx->ctx_flags);
		queue_delayed_work(cnic_wq, &cp->delete_task, delta);
		goto destroy_reply;
	}

	ret = cnic_bnx2x_destroy_ramrod(dev, l5_cid);

skip_cfc_delete:
	cnic_free_bnx2x_conn_resc(dev, l5_cid);

	atomic_dec(&cp->iscsi_conn);
	clear_bit(CTX_FL_OFFLD_START, &ctx->ctx_flags);

destroy_reply:
	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;
	u32 pfid = cp->pfid;
	u8 *mac = dev->mac_addr;

	CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
		 XSTORM_ISCSI_LOCAL_MAC_ADDR0_OFFSET(pfid), mac[0]);
	CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
		 XSTORM_ISCSI_LOCAL_MAC_ADDR1_OFFSET(pfid), mac[1]);
	CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
		 XSTORM_ISCSI_LOCAL_MAC_ADDR2_OFFSET(pfid), mac[2]);
	CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
		 XSTORM_ISCSI_LOCAL_MAC_ADDR3_OFFSET(pfid), mac[3]);
	CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
		 XSTORM_ISCSI_LOCAL_MAC_ADDR4_OFFSET(pfid), mac[4]);
	CNIC_WR8(dev, BAR_XSTRORM_INTMEM +
		 XSTORM_ISCSI_LOCAL_MAC_ADDR5_OFFSET(pfid), mac[5]);

	CNIC_WR8(dev, BAR_TSTRORM_INTMEM +
		 TSTORM_ISCSI_TCP_VARS_LSB_LOCAL_MAC_ADDR_OFFSET(pfid), mac[5]);
	CNIC_WR8(dev, BAR_TSTRORM_INTMEM +
		 TSTORM_ISCSI_TCP_VARS_LSB_LOCAL_MAC_ADDR_OFFSET(pfid) + 1,
		 mac[4]);
	CNIC_WR8(dev, BAR_TSTRORM_INTMEM +
		 TSTORM_ISCSI_TCP_VARS_MSB_LOCAL_MAC_ADDR_OFFSET(pfid), mac[3]);
	CNIC_WR8(dev, BAR_TSTRORM_INTMEM +
		 TSTORM_ISCSI_TCP_VARS_MSB_LOCAL_MAC_ADDR_OFFSET(pfid) + 1,
		 mac[2]);
	CNIC_WR8(dev, BAR_TSTRORM_INTMEM +
		 TSTORM_ISCSI_TCP_VARS_MSB_LOCAL_MAC_ADDR_OFFSET(pfid) + 2,
		 mac[1]);
	CNIC_WR8(dev, BAR_TSTRORM_INTMEM +
		 TSTORM_ISCSI_TCP_VARS_MSB_LOCAL_MAC_ADDR_OFFSET(pfid) + 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->pfid), xstorm_flags);

	CNIC_WR16(dev, BAR_TSTRORM_INTMEM +
		  TSTORM_ISCSI_TCP_VARS_FLAGS_OFFSET(cp->pfid), 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->pfid), 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)
		set_bit(CTX_FL_OFFLD_START, &ctx->ctx_flags);

	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, comp = 0;

	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))
			comp++;

		while (j < num_cqes) {
			u32 next_op = cp->completed_kcq[i + j]->kcqe_op_flag;

			if ((next_op & KCQE_FLAGS_LAYER_MASK) != kcqe_layer)
				break;

			if (unlikely(next_op & KCQE_RAMROD_COMPLETION))
				comp++;
			j++;
		}

		if (kcqe_layer == KCQE_FLAGS_LAYER_MASK_L5_RDMA)
			ulp_type = CNIC_ULP_RDMA;
		else if (kcqe_layer == KCQE_FLAGS_LAYER_MASK_L5_ISCSI)
			ulp_type = CNIC_ULP_ISCSI;
		else if (kcqe_layer == KCQE_FLAGS_LAYER_MASK_L4)
			ulp_type = CNIC_ULP_L4;
		else if (kcqe_layer == KCQE_FLAGS_LAYER_MASK_L2)
			goto end;
		else {
			netdev_err(dev->netdev, "Unknown type of KCQE(0x%x)\n",
				   kcqe_op_flag);
			goto end;
		}

		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],
						  cp->completed_kcq + i, j);
		}
		rcu_read_unlock();
end:
		num_cqes -= j;
		i += j;
		j = 1;
	}
	if (unlikely(comp))
		cnic_spq_completion(dev, DRV_CTL_RET_L5_SPQ_CREDIT_CMD, comp);
}

static u16 cnic_bnx2_next_idx(u16 idx)
{
	return idx + 1;
}

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, struct kcq_info *info)
{
	struct cnic_local *cp = dev->cnic_priv;
	u16 i, ri, hw_prod, last;
	struct kcqe *kcqe;
	int kcqe_cnt = 0, last_cnt = 0;

	i = ri = last = info->sw_prod_idx;
	ri &= MAX_KCQ_IDX;
	hw_prod = *info->hw_prod_idx_ptr;
	hw_prod = cp->hw_idx(hw_prod);

	while ((i != hw_prod) && (kcqe_cnt < MAX_COMPLETED_KCQE)) {
		kcqe = &info->kcq[KCQ_PG(ri)][KCQ_IDX(ri)];
		cp->completed_kcq[kcqe_cnt++] = kcqe;
		i = cp->next_idx(i);
		ri = i & MAX_KCQ_IDX;
		if (likely(!(kcqe->kcqe_op_flag & KCQE_FLAGS_NEXT))) {
			last_cnt = kcqe_cnt;
			last = i;
		}
	}

	info->sw_prod_idx = last;
	return last_cnt;
}

static int cnic_l2_completion(struct cnic_local *cp)
{
	u16 hw_cons, sw_cons;
	struct cnic_uio_dev *udev = cp->udev;
	union eth_rx_cqe *cqe, *cqe_ring = (union eth_rx_cqe *)
					(udev->l2_ring + (2 * BCM_PAGE_SIZE));
	u32 cmd;
	int comp = 0;

	if (!test_bit(CNIC_F_BNX2X_CLASS, &cp->dev->flags))
		return 0;

	hw_cons = *cp->rx_cons_ptr;
	if ((hw_cons & BNX2X_MAX_RCQ_DESC_CNT) == BNX2X_MAX_RCQ_DESC_CNT)
		hw_cons++;

	sw_cons = cp->rx_cons;
	while (sw_cons != hw_cons) {
		u8 cqe_fp_flags;

		cqe = &cqe_ring[sw_cons & BNX2X_MAX_RCQ_DESC_CNT];
		cqe_fp_flags = cqe->fast_path_cqe.type_error_flags;
		if (cqe_fp_flags & ETH_FAST_PATH_RX_CQE_TYPE) {
			cmd = le32_to_cpu(cqe->ramrod_cqe.conn_and_cmd_data);
			cmd >>= COMMON_RAMROD_ETH_RX_CQE_CMD_ID_SHIFT;
			if (cmd == RAMROD_CMD_ID_ETH_CLIENT_SETUP ||
			    cmd == RAMROD_CMD_ID_ETH_HALT)
				comp++;
		}
		sw_cons = BNX2X_NEXT_RCQE(sw_cons);
	}
	return comp;
}

static void cnic_chk_pkt_rings(struct cnic_local *cp)
{
	u16 rx_cons, tx_cons;
	int comp = 0;

	if (!test_bit(CNIC_LCL_FL_RINGS_INITED, &cp->cnic_local_flags))
		return;

	rx_cons = *cp->rx_cons_ptr;
	tx_cons = *cp->tx_cons_ptr;
	if (cp->tx_cons != tx_cons || cp->rx_cons != rx_cons) {
		if (test_bit(CNIC_LCL_FL_L2_WAIT, &cp->cnic_local_flags))
			comp = cnic_l2_completion(cp);

		cp->tx_cons = tx_cons;
		cp->rx_cons = rx_cons;

		if (cp->udev)
			uio_event_notify(&cp->udev->cnic_uinfo);
	}
	if (comp)
		clear_bit(CNIC_LCL_FL_L2_WAIT, &cp->cnic_local_flags);
}

static u32 cnic_service_bnx2_queues(struct cnic_dev *dev)
{
	struct cnic_local *cp = dev->cnic_priv;
	u32 status_idx = (u16) *cp->kcq1.status_idx_ptr;
	int kcqe_cnt;

	cp->kwq_con_idx = *cp->kwq_con_idx_ptr;

	while ((kcqe_cnt = cnic_get_kcqes(dev, &cp->kcq1))) {

		service_kcqes(dev, kcqe_cnt);

		/* Tell compiler that status_blk fields can change. */
		barrier();
		if (status_idx != *cp->kcq1.status_idx_ptr) {
			status_idx = (u16) *cp->kcq1.status_idx_ptr;
			cp->kwq_con_idx = *cp->kwq_con_idx_ptr;
		} else
			break;
	}

	CNIC_WR16(dev, cp->kcq1.io_addr, cp->kcq1.sw_prod_idx);

	cnic_chk_pkt_rings(cp);

	return status_idx;
}

static int cnic_service_bnx2(void *data, void *status_blk)
{
	struct cnic_dev *dev = data;
	struct cnic_local *cp = dev->cnic_priv;
	u32 status_idx = *cp->kcq1.status_idx_ptr;

	if (unlikely(!test_bit(CNIC_F_CNIC_UP, &dev->flags)))
		return status_idx;

	return cnic_service_bnx2_queues(dev);
}

static void cnic_service_bnx2_msix(unsigned long data)
{
	struct cnic_dev *dev = (struct cnic_dev *) data;
	struct cnic_local *cp = dev->cnic_priv;

	cp->last_status_idx = cnic_service_bnx2_queues(dev);

	CNIC_WR(dev, BNX2_PCICFG_INT_ACK_CMD, cp->int_num |
		BNX2_PCICFG_INT_ACK_CMD_INDEX_VALID | cp->last_status_idx);
}

static void cnic_doirq(struct cnic_dev *dev)
{
	struct cnic_local *cp = dev->cnic_priv;
	u16 prod = cp->kcq1.sw_prod_idx & MAX_KCQ_IDX;

	if (likely(test_bit(CNIC_F_CNIC_UP, &dev->flags))) {
		prefetch(cp->status_blk.gen);
		prefetch(&cp->kcq1.kcq[KCQ_PG(prod)][KCQ_IDX(prod)]);

		tasklet_schedule(&cp->cnic_irq_task);
	}
}

static irqreturn_t cnic_irq(int irq, void *dev_instance)
{
	struct cnic_dev *dev = dev_instance;
	struct cnic_local *cp = dev->cnic_priv;

	if (cp->ack_int)
		cp->ack_int(dev);

	cnic_doirq(dev);

	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_igu_sb(struct cnic_dev *dev, u8 igu_sb_id, u8 segment,
			    u16 index, u8 op, u8 update)
{
	struct igu_regular cmd_data;
	u32 igu_addr = BAR_IGU_INTMEM + (IGU_CMD_INT_ACK_BASE + igu_sb_id) * 8;

	cmd_data.sb_id_and_flags =
		(index << IGU_REGULAR_SB_INDEX_SHIFT) |
		(segment << IGU_REGULAR_SEGMENT_ACCESS_SHIFT) |
		(update << IGU_REGULAR_BUPDATE_SHIFT) |
		(op << IGU_REGULAR_ENABLE_INT_SHIFT);


	CNIC_WR(dev, igu_addr, cmd_data.sb_id_and_flags);
}

static void cnic_ack_bnx2x_msix(struct cnic_dev *dev)
{
	struct cnic_local *cp = dev->cnic_priv;

	cnic_ack_bnx2x_int(dev, cp->bnx2x_igu_sb_id, CSTORM_ID, 0,
			   IGU_INT_DISABLE, 0);
}

static void cnic_ack_bnx2x_e2_msix(struct cnic_dev *dev)
{
	struct cnic_local *cp = dev->cnic_priv;

	cnic_ack_igu_sb(dev, cp->bnx2x_igu_sb_id, IGU_SEG_ACCESS_DEF, 0,
			IGU_INT_DISABLE, 0);
}

static u32 cnic_service_bnx2x_kcq(struct cnic_dev *dev, struct kcq_info *info)
{
	u32 last_status = *info->status_idx_ptr;
	int kcqe_cnt;

	while ((kcqe_cnt = cnic_get_kcqes(dev, info))) {

		service_kcqes(dev, kcqe_cnt);

		/* Tell compiler that sblk fields can change. */
		barrier();
		if (last_status == *info->status_idx_ptr)
			break;

		last_status = *info->status_idx_ptr;
	}
	return last_status;
}

static void cnic_service_bnx2x_bh(unsigned long data)
{
	struct cnic_dev *dev = (struct cnic_dev *) data;
	struct cnic_local *cp = dev->cnic_priv;
	u32 status_idx;

	if (unlikely(!test_bit(CNIC_F_CNIC_UP, &dev->flags)))
		return;

	status_idx = cnic_service_bnx2x_kcq(dev, &cp->kcq1);

	CNIC_WR16(dev, cp->kcq1.io_addr, cp->kcq1.sw_prod_idx + MAX_KCQ_IDX);
	if (BNX2X_CHIP_IS_E2(cp->chip_id))
		cnic_ack_igu_sb(dev, cp->bnx2x_igu_sb_id, IGU_SEG_ACCESS_DEF,
				status_idx, IGU_INT_ENABLE, 1);
	else
		cnic_ack_bnx2x_int(dev, cp->bnx2x_igu_sb_id, USTORM_ID,
				   status_idx, IGU_INT_ENABLE, 1);
}

static int cnic_service_bnx2x(void *data, void *status_blk)
{
	struct cnic_dev *dev = data;
	struct cnic_local *cp = dev->cnic_priv;

	if (!(cp->ethdev->drv_state & CNIC_DRV_STATE_USING_MSIX))
		cnic_doirq(dev);

	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;

	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;

		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]);
	}
}

static void cnic_ulp_start(struct cnic_dev *dev)
{
	struct cnic_local *cp = dev->cnic_priv;
	int if_type;

	for (if_type = 0; if_type < MAX_CNIC_ULP_TYPE; if_type++) {
		struct cnic_ulp_ops *ulp_ops;

		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]);
	}
}

static int cnic_ctl(void *data, struct cnic_ctl_info *info)
{
	struct cnic_dev *dev = data;

	switch (info->cmd) {
	case CNIC_CTL_STOP_CMD:
		cnic_hold(dev);

		cnic_ulp_stop(dev);
		cnic_stop_hw(dev);

		cnic_put(dev);
		break;
	case CNIC_CTL_START_CMD:
		cnic_hold(dev);

		if (!cnic_start_hw(dev))
			cnic_ulp_start(dev);

		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;
	}
	return 0;
}

static void cnic_ulp_init(struct cnic_dev *dev)
{
	int i;
	struct cnic_local *cp = dev->cnic_priv;

	for (i = 0; i < MAX_CNIC_ULP_TYPE_EXT; i++) {
		struct cnic_ulp_ops *ulp_ops;

		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);
	}
}

static void cnic_ulp_exit(struct cnic_dev *dev)
{
	int i;
	struct cnic_local *cp = dev->cnic_priv;

	for (i = 0; i < MAX_CNIC_ULP_TYPE_EXT; i++) {
		struct cnic_ulp_ops *ulp_ops;

		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);
	}
}

static int cnic_cm_offload_pg(struct cnic_sock *csk)
{
	struct cnic_dev *dev = csk->dev;
	struct l4_kwq_offload_pg *l4kwqe;
	struct kwqe *wqes[1];

	l4kwqe = (struct l4_kwq_offload_pg *) &csk->kwqe1;
	memset(l4kwqe, 0, sizeof(*l4kwqe));
	wqes[0] = (struct kwqe *) l4kwqe;

	l4kwqe->op_code = L4_KWQE_OPCODE_VALUE_OFFLOAD_PG;
	l4kwqe->flags =
		L4_LAYER_CODE << L4_KWQ_OFFLOAD_PG_LAYER_CODE_SHIFT;
	l4kwqe->l2hdr_nbytes = ETH_HLEN;

	l4kwqe->da0 = csk->ha[0];
	l4kwqe->da1 = csk->ha[1];
	l4kwqe->da2 = csk->ha[2];
	l4kwqe->da3 = csk->ha[3];
	l4kwqe->da4 = csk->ha[4];
	l4kwqe->da5 = csk->ha[5];

	l4kwqe->sa0 = dev->mac_addr[0];
	l4kwqe->sa1 = dev->mac_addr[1];
	l4kwqe->sa2 = dev->mac_addr[2];
	l4kwqe->sa3 = dev->mac_addr[3];
	l4kwqe->sa4 = dev->mac_addr[4];
	l4kwqe->sa5 = dev->mac_addr[5];

	l4kwqe->etype = ETH_P_IP;
	l4kwqe->ipid_start = DEF_IPID_START;
	l4kwqe->host_opaque = csk->l5_cid;

	if (csk->vlan_id) {
		l4kwqe->pg_flags |= L4_KWQ_OFFLOAD_PG_VLAN_TAGGING;
		l4kwqe->vlan_tag = csk->vlan_id;
		l4kwqe->l2hdr_nbytes += 4;
	}

	return dev->submit_kwqes(dev, wqes, 1);
}

static int cnic_cm_update_pg(struct cnic_sock *csk)
{
	struct cnic_dev *dev = csk->dev;
	struct l4_kwq_update_pg *l4kwqe;
	struct kwqe *wqes[1];

	l4kwqe = (struct l4_kwq_update_pg *) &csk->kwqe1;
	memset(l4kwqe, 0, sizeof(*l4kwqe));
	wqes[0] = (struct kwqe *) l4kwqe;

	l4kwqe->opcode = L4_KWQE_OPCODE_VALUE_UPDATE_PG;
	l4kwqe->flags =
		L4_LAYER_CODE << L4_KWQ_UPDATE_PG_LAYER_CODE_SHIFT;
	l4kwqe->pg_cid = csk->pg_cid;

	l4kwqe->da0 = csk->ha[0];
	l4kwqe->da1 = csk->ha[1];
	l4kwqe->da2 = csk->ha[2];
	l4kwqe->da3 = csk->ha[3];
	l4kwqe->da4 = csk->ha[4];
	l4kwqe->da5 = csk->ha[5];

	l4kwqe->pg_host_opaque = csk->l5_cid;
	l4kwqe->pg_valids = L4_KWQ_UPDATE_PG_VALIDS_DA;

	return dev->submit_kwqes(dev, wqes, 1);
}

static int cnic_cm_upload_pg(struct cnic_sock *csk)
{
	struct cnic_dev *dev = csk->dev;
	struct l4_kwq_upload *l4kwqe;
	struct kwqe *wqes[1];

	l4kwqe = (struct l4_kwq_upload *) &csk->kwqe1;
	memset(l4kwqe, 0, sizeof(*l4kwqe));
	wqes[0] = (struct kwqe *) l4kwqe;

	l4kwqe->opcode = L4_KWQE_OPCODE_VALUE_UPLOAD_PG;
	l4kwqe->flags =
		L4_LAYER_CODE << L4_KWQ_UPLOAD_LAYER_CODE_SHIFT;
	l4kwqe->cid = csk->pg_cid;

	return dev->submit_kwqes(dev, wqes, 1);
}

static int cnic_cm_conn_req(struct cnic_sock *csk)
{
	struct cnic_dev *dev = csk->dev;
	struct l4_kwq_connect_req1 *l4kwqe1;
	struct l4_kwq_connect_req2 *l4kwqe2;
	struct l4_kwq_connect_req3 *l4kwqe3;
	struct kwqe *wqes[3];
	u8 tcp_flags = 0;
	int num_wqes = 2;

	l4kwqe1 = (struct l4_kwq_connect_req1 *) &csk->kwqe1;
	l4kwqe2 = (struct l4_kwq_connect_req2 *) &csk->kwqe2;
	l4kwqe3 = (struct l4_kwq_connect_req3 *) &csk->kwqe3;
	memset(l4kwqe1, 0, sizeof(*l4kwqe1));
	memset(l4kwqe2, 0, sizeof(*l4kwqe2));
	memset(l4kwqe3, 0, sizeof(*l4kwqe3));

	l4kwqe3->op_code = L4_KWQE_OPCODE_VALUE_CONNECT3;
	l4kwqe3->flags =
		L4_LAYER_CODE << L4_KWQ_CONNECT_REQ3_LAYER_CODE_SHIFT;
	l4kwqe3->ka_timeout = csk->ka_timeout;
	l4kwqe3->ka_interval = csk->ka_interval;
	l4kwqe3->ka_max_probe_count = csk->ka_max_probe_count;
	l4kwqe3->tos = csk->tos;
	l4kwqe3->ttl = csk->ttl;
	l4kwqe3->snd_seq_scale = csk->snd_seq_scale;
	l4kwqe3->pmtu = csk->mtu;
	l4kwqe3->rcv_buf = csk->rcv_buf;
	l4kwqe3->snd_buf = csk->snd_buf;
	l4kwqe3->seed = csk->seed;

	wqes[0] = (struct kwqe *) l4kwqe1;
	if (test_bit(SK_F_IPV6, &csk->flags)) {
		wqes[1] = (struct kwqe *) l4kwqe2;
		wqes[2] = (struct kwqe *) l4kwqe3;
		num_wqes = 3;

		l4kwqe1->conn_flags = L4_KWQ_CONNECT_REQ1_IP_V6;
		l4kwqe2->op_code = L4_KWQE_OPCODE_VALUE_CONNECT2;
		l4kwqe2->flags =
			L4_KWQ_CONNECT_REQ2_LINKED_WITH_NEXT |
			L4_LAYER_CODE << L4_KWQ_CONNECT_REQ2_LAYER_CODE_SHIFT;
		l4kwqe2->src_ip_v6_2 = be32_to_cpu(csk->src_ip[1]);
		l4kwqe2->src_ip_v6_3 = be32_to_cpu(csk->src_ip[2]);
		l4kwqe2->src_ip_v6_4 = be32_to_cpu(csk->src_ip[3]);
		l4kwqe2->dst_ip_v6_2 = be32_to_cpu(csk->dst_ip[1]);
		l4kwqe2->dst_ip_v6_3 = be32_to_cpu(csk->dst_ip[2]);
		l4kwqe2->dst_ip_v6_4 = be32_to_cpu(csk->dst_ip[3]);
		l4kwqe3->mss = l4kwqe3->pmtu - sizeof(struct ipv6hdr) -
			       sizeof(struct tcphdr);
	} else {
		wqes[1] = (struct kwqe *) l4kwqe3;
		l4kwqe3->mss = l4kwqe3->pmtu - sizeof(struct iphdr) -
			       sizeof(struct tcphdr);
	}

	l4kwqe1->op_code = L4_KWQE_OPCODE_VALUE_CONNECT1;
	l4kwqe1->flags =
		(L4_LAYER_CODE << L4_KWQ_CONNECT_REQ1_LAYER_CODE_SHIFT) |
		 L4_KWQ_CONNECT_REQ3_LINKED_WITH_NEXT;
	l4kwqe1->cid = csk->cid;
	l4kwqe1->pg_cid = csk->pg_cid;
	l4kwqe1->src_ip = be32_to_cpu(csk->src_ip[0]);
	l4kwqe1->dst_ip = be32_to_cpu(csk->dst_ip[0]);
	l4kwqe1->src_port = be16_to_cpu(csk->src_port);
	l4kwqe1->dst_port = be16_to_cpu(csk->dst_port);
	if (csk->tcp_flags & SK_TCP_NO_DELAY_ACK)
		tcp_flags |= L4_KWQ_CONNECT_REQ1_NO_DELAY_ACK;
	if (csk->tcp_flags & SK_TCP_KEEP_ALIVE)
		tcp_flags |= L4_KWQ_CONNECT_REQ1_KEEP_ALIVE;
	if (csk->tcp_flags & SK_TCP_NAGLE)
		tcp_flags |= L4_KWQ_CONNECT_REQ1_NAGLE_ENABLE;
	if (csk->tcp_flags & SK_TCP_TIMESTAMP)
		tcp_flags |= L4_KWQ_CONNECT_REQ1_TIME_STAMP;
	if (csk->tcp_flags & SK_TCP_SACK)
		tcp_flags |= L4_KWQ_CONNECT_REQ1_SACK;
	if (csk->tcp_flags & SK_TCP_SEG_SCALING)
		tcp_flags |= L4_KWQ_CONNECT_REQ1_SEG_SCALING;

	l4kwqe1->tcp_flags = tcp_flags;

	return dev->submit_kwqes(dev, wqes, num_wqes);
}

static int cnic_cm_close_req(struct cnic_sock *csk)
{
	struct cnic_dev *dev = csk->dev;
	struct l4_kwq_close_req *l4kwqe;
	struct kwqe *wqes[1];

	l4kwqe = (struct l4_kwq_close_req *) &csk->kwqe2;
	memset(l4kwqe, 0, sizeof(*l4kwqe));
	wqes[0] = (struct kwqe *) l4kwqe;

	l4kwqe->op_code = L4_KWQE_OPCODE_VALUE_CLOSE;
	l4kwqe->flags = L4_LAYER_CODE << L4_KWQ_CLOSE_REQ_LAYER_CODE_SHIFT;
	l4kwqe->cid = csk->cid;

	return dev->submit_kwqes(dev, wqes, 1);
}

static int cnic_cm_abort_req(struct cnic_sock *csk)
{
	struct cnic_dev *dev = csk->dev;
	struct l4_kwq_reset_req *l4kwqe;
	struct kwqe *wqes[1];

	l4kwqe = (struct l4_kwq_reset_req *) &csk->kwqe2;
	memset(l4kwqe, 0, sizeof(*l4kwqe));
	wqes[0] = (struct kwqe *) l4kwqe;

	l4kwqe->op_code = L4_KWQE_OPCODE_VALUE_RESET;
	l4kwqe->flags = L4_LAYER_CODE << L4_KWQ_RESET_REQ_LAYER_CODE_SHIFT;
	l4kwqe->cid = csk->cid;

	return dev->submit_kwqes(dev, wqes, 1);
}

static int cnic_cm_create(struct cnic_dev *dev, int ulp_type, u32 cid,
			  u32 l5_cid, struct cnic_sock **csk, void *context)
{
	struct cnic_local *cp = dev->cnic_priv;
	struct cnic_sock *csk1;

	if (l5_cid >= MAX_CM_SK_TBL_SZ)
		return -EINVAL;

	if (cp->ctx_tbl) {
		struct cnic_context *ctx = &cp->ctx_tbl[l5_cid];

		if (test_bit(CTX_FL_OFFLD_START, &ctx->ctx_flags))
			return -EAGAIN;
	}

	csk1 = &cp->csk_tbl[l5_cid];
	if (atomic_read(&csk1->ref_count))
		return -EAGAIN;

	if (test_and_set_bit(SK_F_INUSE, &csk1->flags))
		return -EBUSY;

	csk1->dev = dev;
	csk1->cid = cid;
	csk1->l5_cid = l5_cid;
	csk1->ulp_type = ulp_type;
	csk1->context = context;

	csk1->ka_timeout = DEF_KA_TIMEOUT;
	csk1->ka_interval = DEF_KA_INTERVAL;
	csk1->ka_max_probe_count = DEF_KA_MAX_PROBE_COUNT;
	csk1->tos = DEF_TOS;
	csk1->ttl = DEF_TTL;
	csk1->snd_seq_scale = DEF_SND_SEQ_SCALE;