/*
 * This file is part of the Chelsio T4 Ethernet driver for Linux.
 *
 * Copyright (c) 2003-2016 Chelsio Communications, Inc. All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */

#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

#include <linux/bitmap.h>
#include <linux/crc32.h>
#include <linux/ctype.h>
#include <linux/debugfs.h>
#include <linux/err.h>
#include <linux/etherdevice.h>
#include <linux/firmware.h>
#include <linux/if.h>
#include <linux/if_vlan.h>
#include <linux/init.h>
#include <linux/log2.h>
#include <linux/mdio.h>
#include <linux/module.h>
#include <linux/moduleparam.h>
#include <linux/mutex.h>
#include <linux/netdevice.h>
#include <linux/pci.h>
#include <linux/aer.h>
#include <linux/rtnetlink.h>
#include <linux/sched.h>
#include <linux/seq_file.h>
#include <linux/sockios.h>
#include <linux/vmalloc.h>
#include <linux/workqueue.h>
#include <net/neighbour.h>
#include <net/netevent.h>
#include <net/addrconf.h>
#include <net/bonding.h>
#include <linux/uaccess.h>
#include <linux/crash_dump.h>
#include <net/udp_tunnel.h>
#include <net/xfrm.h>
#if IS_ENABLED(CONFIG_CHELSIO_TLS_DEVICE)
#include <net/tls.h>
#endif

#include "cxgb4.h"
#include "cxgb4_filter.h"
#include "t4_regs.h"
#include "t4_values.h"
#include "t4_msg.h"
#include "t4fw_api.h"
#include "t4fw_version.h"
#include "cxgb4_dcb.h"
#include "srq.h"
#include "cxgb4_debugfs.h"
#include "clip_tbl.h"
#include "l2t.h"
#include "smt.h"
#include "sched.h"
#include "cxgb4_tc_u32.h"
#include "cxgb4_tc_flower.h"
#include "cxgb4_tc_mqprio.h"
#include "cxgb4_tc_matchall.h"
#include "cxgb4_ptp.h"
#include "cxgb4_cudbg.h"

char cxgb4_driver_name[] = KBUILD_MODNAME;

#define DRV_DESC "Chelsio T4/T5/T6 Network Driver"

#define DFLT_MSG_ENABLE (NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_LINK | \
                         NETIF_MSG_TIMER | NETIF_MSG_IFDOWN | NETIF_MSG_IFUP |\
                         NETIF_MSG_RX_ERR | NETIF_MSG_TX_ERR)

/* Macros needed to support the PCI Device ID Table ...
 */
#define CH_PCI_DEVICE_ID_TABLE_DEFINE_BEGIN \
        static const struct pci_device_id cxgb4_pci_tbl[] = {
#define CXGB4_UNIFIED_PF 0x4

#define CH_PCI_DEVICE_ID_FUNCTION CXGB4_UNIFIED_PF

/* Include PCI Device IDs for both PF4 and PF0-3 so our PCI probe() routine is
 * called for both.
 */
#define CH_PCI_DEVICE_ID_FUNCTION2 0x0

#define CH_PCI_ID_TABLE_ENTRY(devid) \
                {PCI_VDEVICE(CHELSIO, (devid)), CXGB4_UNIFIED_PF}

#define CH_PCI_DEVICE_ID_TABLE_DEFINE_END \
                { 0, } \
        }

#include "t4_pci_id_tbl.h"

#define FW4_FNAME "cxgb4/t4fw.bin"
#define FW5_FNAME "cxgb4/t5fw.bin"
#define FW6_FNAME "cxgb4/t6fw.bin"
#define FW4_CFNAME "cxgb4/t4-config.txt"
#define FW5_CFNAME "cxgb4/t5-config.txt"
#define FW6_CFNAME "cxgb4/t6-config.txt"
#define PHY_AQ1202_FIRMWARE "cxgb4/aq1202_fw.cld"
#define PHY_BCM84834_FIRMWARE "cxgb4/bcm8483.bin"
#define PHY_AQ1202_DEVICEID 0x4409
#define PHY_BCM84834_DEVICEID 0x4486

MODULE_DESCRIPTION(DRV_DESC);
MODULE_AUTHOR("Chelsio Communications");
MODULE_LICENSE("Dual BSD/GPL");
MODULE_DEVICE_TABLE(pci, cxgb4_pci_tbl);
MODULE_FIRMWARE(FW4_FNAME);
MODULE_FIRMWARE(FW5_FNAME);
MODULE_FIRMWARE(FW6_FNAME);

/*
 * The driver uses the best interrupt scheme available on a platform in the
 * order MSI-X, MSI, legacy INTx interrupts.  This parameter determines which
 * of these schemes the driver may consider as follows:
 *
 * msi = 2: choose from among all three options
 * msi = 1: only consider MSI and INTx interrupts
 * msi = 0: force INTx interrupts
 */
static int msi = 2;

module_param(msi, int, 0644);
MODULE_PARM_DESC(msi, "whether to use INTx (0), MSI (1) or MSI-X (2)");

/*
 * Normally we tell the chip to deliver Ingress Packets into our DMA buffers
 * offset by 2 bytes in order to have the IP headers line up on 4-byte
 * boundaries.  This is a requirement for many architectures which will throw
 * a machine check fault if an attempt is made to access one of the 4-byte IP
 * header fields on a non-4-byte boundary.  And it's a major performance issue
 * even on some architectures which allow it like some implementations of the
 * x86 ISA.  However, some architectures don't mind this and for some very
 * edge-case performance sensitive applications (like forwarding large volumes
 * of small packets), setting this DMA offset to 0 will decrease the number of
 * PCI-E Bus transfers enough to measurably affect performance.
 */
static int rx_dma_offset = 2;

/* TX Queue select used to determine what algorithm to use for selecting TX
 * queue. Select between the kernel provided function (select_queue=0) or user
 * cxgb_select_queue function (select_queue=1)
 *
 * Default: select_queue=0
 */
static int select_queue;
module_param(select_queue, int, 0644);
MODULE_PARM_DESC(select_queue,
                 "Select between kernel provided method of selecting or driver method of selecting TX queue. Default is kernel method.");

static struct dentry *cxgb4_debugfs_root;

LIST_HEAD(adapter_list);
DEFINE_MUTEX(uld_mutex);
LIST_HEAD(uld_list);

static int cfg_queues(struct adapter *adap);

static void link_report(struct net_device *dev)
{
        if (!netif_carrier_ok(dev))
                netdev_info(dev, "link down\n");
        else {
                static const char *fc[] = { "no", "Rx", "Tx", "Tx/Rx" };

                const char *s;
                const struct port_info *p = netdev_priv(dev);

                switch (p->link_cfg.speed) {
                case 100:
                        s = "100Mbps";
                        break;
                case 1000:
                        s = "1Gbps";
                        break;
                case 10000:
                        s = "10Gbps";
                        break;
                case 25000:
                        s = "25Gbps";
                        break;
                case 40000:
                        s = "40Gbps";
                        break;
                case 50000:
                        s = "50Gbps";
                        break;
                case 100000:
                        s = "100Gbps";
                        break;
                default:
                        pr_info("%s: unsupported speed: %d\n",
                                dev->name, p->link_cfg.speed);
                        return;
                }

                netdev_info(dev, "link up, %s, full-duplex, %s PAUSE\n", s,
                            fc[p->link_cfg.fc]);
        }
}

#ifdef CONFIG_CHELSIO_T4_DCB
/* Set up/tear down Data Center Bridging Priority mapping for a net device. */
static void dcb_tx_queue_prio_enable(struct net_device *dev, int enable)
{
        struct port_info *pi = netdev_priv(dev);
        struct adapter *adap = pi->adapter;
        struct sge_eth_txq *txq = &adap->sge.ethtxq[pi->first_qset];
        int i;

        /* We use a simple mapping of Port TX Queue Index to DCB
         * Priority when we're enabling DCB.
         */
        for (i = 0; i < pi->nqsets; i++, txq++) {
                u32 name, value;
                int err;

                name = (FW_PARAMS_MNEM_V(FW_PARAMS_MNEM_DMAQ) |
                        FW_PARAMS_PARAM_X_V(
                                FW_PARAMS_PARAM_DMAQ_EQ_DCBPRIO_ETH) |
                        FW_PARAMS_PARAM_YZ_V(txq->q.cntxt_id));
                value = enable ? i : 0xffffffff;

                /* Since we can be called while atomic (from "interrupt
                 * level") we need to issue the Set Parameters Commannd
                 * without sleeping (timeout < 0).
                 */
                err = t4_set_params_timeout(adap, adap->mbox, adap->pf, 0, 1,
                                            &name, &value,
                                            -FW_CMD_MAX_TIMEOUT);

                if (err)
                        dev_err(adap->pdev_dev,
                                "Can't %s DCB Priority on port %d, TX Queue %d: err=%d\n",
                                enable ? "set" : "unset", pi->port_id, i, -err);
                else
                        txq->dcb_prio = enable ? value : 0;
        }
}

int cxgb4_dcb_enabled(const struct net_device *dev)
{
        struct port_info *pi = netdev_priv(dev);

        if (!pi->dcb.enabled)
                return 0;

        return ((pi->dcb.state == CXGB4_DCB_STATE_FW_ALLSYNCED) ||
                (pi->dcb.state == CXGB4_DCB_STATE_HOST));
}
#endif /* CONFIG_CHELSIO_T4_DCB */

void t4_os_link_changed(struct adapter *adapter, int port_id, int link_stat)
{
        struct net_device *dev = adapter->port[port_id];

        /* Skip changes from disabled ports. */
        if (netif_running(dev) && link_stat != netif_carrier_ok(dev)) {
                if (link_stat)
                        netif_carrier_on(dev);
                else {
#ifdef CONFIG_CHELSIO_T4_DCB
                        if (cxgb4_dcb_enabled(dev)) {
                                cxgb4_dcb_reset(dev);
                                dcb_tx_queue_prio_enable(dev, false);
                        }
#endif /* CONFIG_CHELSIO_T4_DCB */
                        netif_carrier_off(dev);
                }

                link_report(dev);
        }
}

void t4_os_portmod_changed(struct adapter *adap, int port_id)
{
        static const char *mod_str[] = {
                NULL, "LR", "SR", "ER", "passive DA", "active DA", "LRM"
        };

        struct net_device *dev = adap->port[port_id];
        struct port_info *pi = netdev_priv(dev);

        if (pi->mod_type == FW_PORT_MOD_TYPE_NONE)
                netdev_info(dev, "port module unplugged\n");
        else if (pi->mod_type < ARRAY_SIZE(mod_str))
                netdev_info(dev, "%s module inserted\n", mod_str[pi->mod_type]);
        else if (pi->mod_type == FW_PORT_MOD_TYPE_NOTSUPPORTED)
                netdev_info(dev, "%s: unsupported port module inserted\n",
                            dev->name);
        else if (pi->mod_type == FW_PORT_MOD_TYPE_UNKNOWN)
                netdev_info(dev, "%s: unknown port module inserted\n",
                            dev->name);
        else if (pi->mod_type == FW_PORT_MOD_TYPE_ERROR)
                netdev_info(dev, "%s: transceiver module error\n", dev->name);
        else
                netdev_info(dev, "%s: unknown module type %d inserted\n",
                            dev->name, pi->mod_type);

        /* If the interface is running, then we'll need any "sticky" Link
         * Parameters redone with a new Transceiver Module.
         */
        pi->link_cfg.redo_l1cfg = netif_running(dev);
}

int dbfifo_int_thresh = 10; /* 10 == 640 entry threshold */
module_param(dbfifo_int_thresh, int, 0644);
MODULE_PARM_DESC(dbfifo_int_thresh, "doorbell fifo interrupt threshold");

/*
 * usecs to sleep while draining the dbfifo
 */
static int dbfifo_drain_delay = 1000;
module_param(dbfifo_drain_delay, int, 0644);
MODULE_PARM_DESC(dbfifo_drain_delay,
                 "usecs to sleep while draining the dbfifo");

static inline int cxgb4_set_addr_hash(struct port_info *pi)
{
        struct adapter *adap = pi->adapter;
        u64 vec = 0;
        bool ucast = false;
        struct hash_mac_addr *entry;

        /* Calculate the hash vector for the updated list and program it */
        list_for_each_entry(entry, &adap->mac_hlist, list) {
                ucast |= is_unicast_ether_addr(entry->addr);
                vec |= (1ULL << hash_mac_addr(entry->addr));
        }
        return t4_set_addr_hash(adap, adap->mbox, pi->viid, ucast,
                                vec, false);
}

static int cxgb4_mac_sync(struct net_device *netdev, const u8 *mac_addr)
{
        struct port_info *pi = netdev_priv(netdev);
        struct adapter *adap = pi->adapter;
        int ret;
        u64 mhash = 0;
        u64 uhash = 0;
        /* idx stores the index of allocated filters,
         * its size should be modified based on the number of
         * MAC addresses that we allocate filters for
         */

        u16 idx[1] = {};
        bool free = false;
        bool ucast = is_unicast_ether_addr(mac_addr);
        const u8 *maclist[1] = {mac_addr};
        struct hash_mac_addr *new_entry;

        ret = cxgb4_alloc_mac_filt(adap, pi->viid, free, 1, maclist,
                                   idx, ucast ? &uhash : &mhash, false);
        if (ret < 0)
                goto out;
        /* if hash != 0, then add the addr to hash addr list
         * so on the end we will calculate the hash for the
         * list and program it
         */
        if (uhash || mhash) {
                new_entry = kzalloc(sizeof(*new_entry), GFP_ATOMIC);
                if (!new_entry)
                        return -ENOMEM;
                ether_addr_copy(new_entry->addr, mac_addr);
                list_add_tail(&new_entry->list, &adap->mac_hlist);
                ret = cxgb4_set_addr_hash(pi);
        }
out:
        return ret < 0 ? ret : 0;
}

static int cxgb4_mac_unsync(struct net_device *netdev, const u8 *mac_addr)
{
        struct port_info *pi = netdev_priv(netdev);
        struct adapter *adap = pi->adapter;
        int ret;
        const u8 *maclist[1] = {mac_addr};
        struct hash_mac_addr *entry, *tmp;

        /* If the MAC address to be removed is in the hash addr
         * list, delete it from the list and update hash vector
         */
        list_for_each_entry_safe(entry, tmp, &adap->mac_hlist, list) {
                if (ether_addr_equal(entry->addr, mac_addr)) {
                        list_del(&entry->list);
                        kfree(entry);
                        return cxgb4_set_addr_hash(pi);
                }
        }

        ret = cxgb4_free_mac_filt(adap, pi->viid, 1, maclist, false);
        return ret < 0 ? -EINVAL : 0;
}

/*
 * Set Rx properties of a port, such as promiscruity, address filters, and MTU.
 * If @mtu is -1 it is left unchanged.
 */
static int set_rxmode(struct net_device *dev, int mtu, bool sleep_ok)
{
        struct port_info *pi = netdev_priv(dev);
        struct adapter *adapter = pi->adapter;

        __dev_uc_sync(dev, cxgb4_mac_sync, cxgb4_mac_unsync);
        __dev_mc_sync(dev, cxgb4_mac_sync, cxgb4_mac_unsync);

        return t4_set_rxmode(adapter, adapter->mbox, pi->viid, pi->viid_mirror,
                             mtu, (dev->flags & IFF_PROMISC) ? 1 : 0,
                             (dev->flags & IFF_ALLMULTI) ? 1 : 0, 1, -1,
                             sleep_ok);
}

/**
 *      cxgb4_change_mac - Update match filter for a MAC address.
 *      @pi: the port_info
 *      @viid: the VI id
 *      @tcam_idx: TCAM index of existing filter for old value of MAC address,
 *                 or -1
 *      @addr: the new MAC address value
 *      @persist: whether a new MAC allocation should be persistent
 *      @smt_idx: the destination to store the new SMT index.
 *
 *      Modifies an MPS filter and sets it to the new MAC address if
 *      @tcam_idx >= 0, or adds the MAC address to a new filter if
 *      @tcam_idx < 0. In the latter case the address is added persistently
 *      if @persist is %true.
 *      Addresses are programmed to hash region, if tcam runs out of entries.
 *
 */
int cxgb4_change_mac(struct port_info *pi, unsigned int viid,
                     int *tcam_idx, const u8 *addr, bool persist,
                     u8 *smt_idx)
{
        struct adapter *adapter = pi->adapter;
        struct hash_mac_addr *entry, *new_entry;
        int ret;

        ret = t4_change_mac(adapter, adapter->mbox, viid,
                            *tcam_idx, addr, persist, smt_idx);
        /* We ran out of TCAM entries. try programming hash region. */
        if (ret == -ENOMEM) {
                /* If the MAC address to be updated is in the hash addr
                 * list, update it from the list
                 */
                list_for_each_entry(entry, &adapter->mac_hlist, list) {
                        if (entry->iface_mac) {
                                ether_addr_copy(entry->addr, addr);
                                goto set_hash;
                        }
                }
                new_entry = kzalloc(sizeof(*new_entry), GFP_KERNEL);
                if (!new_entry)
                        return -ENOMEM;
                ether_addr_copy(new_entry->addr, addr);
                new_entry->iface_mac = true;
                list_add_tail(&new_entry->list, &adapter->mac_hlist);
set_hash:
                ret = cxgb4_set_addr_hash(pi);
        } else if (ret >= 0) {
                *tcam_idx = ret;
                ret = 0;
        }

        return ret;
}

/*
 *      link_start - enable a port
 *      @dev: the port to enable
 *
 *      Performs the MAC and PHY actions needed to enable a port.
 */
static int link_start(struct net_device *dev)
{
        struct port_info *pi = netdev_priv(dev);
        unsigned int mb = pi->adapter->mbox;
        int ret;

        /*
         * We do not set address filters and promiscuity here, the stack does
         * that step explicitly.
         */
        ret = t4_set_rxmode(pi->adapter, mb, pi->viid, pi->viid_mirror,
                            dev->mtu, -1, -1, -1,
                            !!(dev->features & NETIF_F_HW_VLAN_CTAG_RX), true);
        if (ret == 0)
                ret = cxgb4_update_mac_filt(pi, pi->viid, &pi->xact_addr_filt,
                                            dev->dev_addr, true, &pi->smt_idx);
        if (ret == 0)
                ret = t4_link_l1cfg(pi->adapter, mb, pi->tx_chan,
                                    &pi->link_cfg);
        if (ret == 0) {
                local_bh_disable();
                ret = t4_enable_pi_params(pi->adapter, mb, pi, true,
                                          true, CXGB4_DCB_ENABLED);
                local_bh_enable();
        }

        return ret;
}

#ifdef CONFIG_CHELSIO_T4_DCB
/* Handle a Data Center Bridging update message from the firmware. */
static void dcb_rpl(struct adapter *adap, const struct fw_port_cmd *pcmd)
{
        int port = FW_PORT_CMD_PORTID_G(ntohl(pcmd->op_to_portid));
        struct net_device *dev = adap->port[adap->chan_map[port]];
        int old_dcb_enabled = cxgb4_dcb_enabled(dev);
        int new_dcb_enabled;

        cxgb4_dcb_handle_fw_update(adap, pcmd);
        new_dcb_enabled = cxgb4_dcb_enabled(dev);

        /* If the DCB has become enabled or disabled on the port then we're
         * going to need to set up/tear down DCB Priority parameters for the
         * TX Queues associated with the port.
         */
        if (new_dcb_enabled != old_dcb_enabled)
                dcb_tx_queue_prio_enable(dev, new_dcb_enabled);
}
#endif /* CONFIG_CHELSIO_T4_DCB */

/* Response queue handler for the FW event queue.
 */
static int fwevtq_handler(struct sge_rspq *q, const __be64 *rsp,
                          const struct pkt_gl *gl)
{
        u8 opcode = ((const struct rss_header *)rsp)->opcode;

        rsp++;                                          /* skip RSS header */

        /* FW can send EGR_UPDATEs encapsulated in a CPL_FW4_MSG.
         */
        if (unlikely(opcode == CPL_FW4_MSG &&
           ((const struct cpl_fw4_msg *)rsp)->type == FW_TYPE_RSSCPL)) {
                rsp++;
                opcode = ((const struct rss_header *)rsp)->opcode;
                rsp++;
                if (opcode != CPL_SGE_EGR_UPDATE) {
                        dev_err(q->adap->pdev_dev, "unexpected FW4/CPL %#x on FW event queue\n"
                                , opcode);
                        goto out;
                }
        }

        if (likely(opcode == CPL_SGE_EGR_UPDATE)) {
                const struct cpl_sge_egr_update *p = (void *)rsp;
                unsigned int qid = EGR_QID_G(ntohl(p->opcode_qid));
                struct sge_txq *txq;

                txq = q->adap->sge.egr_map[qid - q->adap->sge.egr_start];
                txq->restarts++;
                if (txq->q_type == CXGB4_TXQ_ETH) {
                        struct sge_eth_txq *eq;

                        eq = container_of(txq, struct sge_eth_txq, q);
                        t4_sge_eth_txq_egress_update(q->adap, eq, -1);
                } else {
                        struct sge_uld_txq *oq;

                        oq = container_of(txq, struct sge_uld_txq, q);
                        tasklet_schedule(&oq->qresume_tsk);
                }
        } else if (opcode == CPL_FW6_MSG || opcode == CPL_FW4_MSG) {
                const struct cpl_fw6_msg *p = (void *)rsp;

#ifdef CONFIG_CHELSIO_T4_DCB
                const struct fw_port_cmd *pcmd = (const void *)p->data;
                unsigned int cmd = FW_CMD_OP_G(ntohl(pcmd->op_to_portid));
                unsigned int action =
                        FW_PORT_CMD_ACTION_G(ntohl(pcmd->action_to_len16));

                if (cmd == FW_PORT_CMD &&
                    (action == FW_PORT_ACTION_GET_PORT_INFO ||
                     action == FW_PORT_ACTION_GET_PORT_INFO32)) {
                        int port = FW_PORT_CMD_PORTID_G(
                                        be32_to_cpu(pcmd->op_to_portid));
                        struct net_device *dev;
                        int dcbxdis, state_input;

                        dev = q->adap->port[q->adap->chan_map[port]];
                        dcbxdis = (action == FW_PORT_ACTION_GET_PORT_INFO
                          ? !!(pcmd->u.info.dcbxdis_pkd & FW_PORT_CMD_DCBXDIS_F)
                          : !!(be32_to_cpu(pcmd->u.info32.lstatus32_to_cbllen32)
                               & FW_PORT_CMD_DCBXDIS32_F));
                        state_input = (dcbxdis
                                       ? CXGB4_DCB_INPUT_FW_DISABLED
                                       : CXGB4_DCB_INPUT_FW_ENABLED);

                        cxgb4_dcb_state_fsm(dev, state_input);
                }

                if (cmd == FW_PORT_CMD &&
                    action == FW_PORT_ACTION_L2_DCB_CFG)
                        dcb_rpl(q->adap, pcmd);
                else
#endif
                        if (p->type == 0)
                                t4_handle_fw_rpl(q->adap, p->data);
        } else if (opcode == CPL_L2T_WRITE_RPL) {
                const struct cpl_l2t_write_rpl *p = (void *)rsp;

                do_l2t_write_rpl(q->adap, p);
        } else if (opcode == CPL_SMT_WRITE_RPL) {
                const struct cpl_smt_write_rpl *p = (void *)rsp;

                do_smt_write_rpl(q->adap, p);
        } else if (opcode == CPL_SET_TCB_RPL) {
                const struct cpl_set_tcb_rpl *p = (void *)rsp;

                filter_rpl(q->adap, p);
        } else if (opcode == CPL_ACT_OPEN_RPL) {
                const struct cpl_act_open_rpl *p = (void *)rsp;

                hash_filter_rpl(q->adap, p);
        } else if (opcode == CPL_ABORT_RPL_RSS) {
                const struct cpl_abort_rpl_rss *p = (void *)rsp;

                hash_del_filter_rpl(q->adap, p);
        } else if (opcode == CPL_SRQ_TABLE_RPL) {
                const struct cpl_srq_table_rpl *p = (void *)rsp;

                do_srq_table_rpl(q->adap, p);
        } else
                dev_err(q->adap->pdev_dev,
                        "unexpected CPL %#x on FW event queue\n", opcode);
out:
        return 0;
}

static void disable_msi(struct adapter *adapter)
{
        if (adapter->flags & CXGB4_USING_MSIX) {
                pci_disable_msix(adapter->pdev);
                adapter->flags &= ~CXGB4_USING_MSIX;
        } else if (adapter->flags & CXGB4_USING_MSI) {
                pci_disable_msi(adapter->pdev);
                adapter->flags &= ~CXGB4_USING_MSI;
        }
}

/*
 * Interrupt handler for non-data events used with MSI-X.
 */
static irqreturn_t t4_nondata_intr(int irq, void *cookie)
{
        struct adapter *adap = cookie;
        u32 v = t4_read_reg(adap, MYPF_REG(PL_PF_INT_CAUSE_A));

        if (v & PFSW_F) {
                adap->swintr = 1;
                t4_write_reg(adap, MYPF_REG(PL_PF_INT_CAUSE_A), v);
        }
        if (adap->flags & CXGB4_MASTER_PF)
                t4_slow_intr_handler(adap);
        return IRQ_HANDLED;
}

int cxgb4_set_msix_aff(struct adapter *adap, unsigned short vec,
                       cpumask_var_t *aff_mask, int idx)
{
        int rv;

        if (!zalloc_cpumask_var(aff_mask, GFP_KERNEL)) {
                dev_err(adap->pdev_dev, "alloc_cpumask_var failed\n");
                return -ENOMEM;
        }

        cpumask_set_cpu(cpumask_local_spread(idx, dev_to_node(adap->pdev_dev)),
                        *aff_mask);

        rv = irq_set_affinity_hint(vec, *aff_mask);
        if (rv)
                dev_warn(adap->pdev_dev,
                         "irq_set_affinity_hint %u failed %d\n",
                         vec, rv);

        return 0;
}

void cxgb4_clear_msix_aff(unsigned short vec, cpumask_var_t aff_mask)
{
        irq_set_affinity_hint(vec, NULL);
        free_cpumask_var(aff_mask);
}

static int request_msix_queue_irqs(struct adapter *adap)
{
        struct sge *s = &adap->sge;
        struct msix_info *minfo;
        int err, ethqidx;

        if (s->fwevtq_msix_idx < 0)
                return -ENOMEM;

        err = request_irq(adap->msix_info[s->fwevtq_msix_idx].vec,
                          t4_sge_intr_msix, 0,
                          adap->msix_info[s->fwevtq_msix_idx].desc,
                          &s->fw_evtq);
        if (err)
                return err;

        for_each_ethrxq(s, ethqidx) {
                minfo = s->ethrxq[ethqidx].msix;
                err = request_irq(minfo->vec,
                                  t4_sge_intr_msix, 0,
                                  minfo->desc,
                                  &s->ethrxq[ethqidx].rspq);
                if (err)
                        goto unwind;

                cxgb4_set_msix_aff(adap, minfo->vec,
                                   &minfo->aff_mask, ethqidx);
        }
        return 0;

unwind:
        while (--ethqidx >= 0) {
                minfo = s->ethrxq[ethqidx].msix;
                cxgb4_clear_msix_aff(minfo->vec, minfo->aff_mask);
                free_irq(minfo->vec, &s->ethrxq[ethqidx].rspq);
        }
        free_irq(adap->msix_info[s->fwevtq_msix_idx].vec, &s->fw_evtq);
        return err;
}

static void free_msix_queue_irqs(struct adapter *adap)
{
        struct sge *s = &adap->sge;
        struct msix_info *minfo;
        int i;

        free_irq(adap->msix_info[s->fwevtq_msix_idx].vec, &s->fw_evtq);
        for_each_ethrxq(s, i) {
                minfo = s->ethrxq[i].msix;
                cxgb4_clear_msix_aff(minfo->vec, minfo->aff_mask);
                free_irq(minfo->vec, &s->ethrxq[i].rspq);
        }
}

static int setup_ppod_edram(struct adapter *adap)
{
        unsigned int param, val;
        int ret;

        /* Driver sends FW_PARAMS_PARAM_DEV_PPOD_EDRAM read command to check
         * if firmware supports ppod edram feature or not. If firmware
         * returns 1, then driver can enable this feature by sending
         * FW_PARAMS_PARAM_DEV_PPOD_EDRAM write command with value 1 to
         * enable ppod edram feature.
         */
        param = (FW_PARAMS_MNEM_V(FW_PARAMS_MNEM_DEV) |
                FW_PARAMS_PARAM_X_V(FW_PARAMS_PARAM_DEV_PPOD_EDRAM));

        ret = t4_query_params(adap, adap->mbox, adap->pf, 0, 1, &param, &val);
        if (ret < 0) {
                dev_warn(adap->pdev_dev,
                         "querying PPOD_EDRAM support failed: %d\n",
                         ret);
                return -1;
        }

        if (val != 1)
                return -1;

        ret = t4_set_params(adap, adap->mbox, adap->pf, 0, 1, &param, &val);
        if (ret < 0) {
                dev_err(adap->pdev_dev,
                        "setting PPOD_EDRAM failed: %d\n", ret);
                return -1;
        }
        return 0;
}

static void adap_config_hpfilter(struct adapter *adapter)
{
        u32 param, val = 0;
        int ret;

        /* Enable HP filter region. Older fw will fail this request and
         * it is fine.
         */
        param = FW_PARAM_DEV(HPFILTER_REGION_SUPPORT);
        ret = t4_set_params(adapter, adapter->mbox, adapter->pf, 0,
                            1, &param, &val);

        /* An error means FW doesn't know about HP filter support,
         * it's not a problem, don't return an error.
         */
        if (ret < 0)
                dev_err(adapter->pdev_dev,
                        "HP filter region isn't supported by FW\n");
}

static int cxgb4_config_rss(const struct port_info *pi, u16 *rss,
                            u16 rss_size, u16 viid)
{
        struct adapter *adap = pi->adapter;
        int ret;

        ret = t4_config_rss_range(adap, adap->mbox, viid, 0, rss_size, rss,
                                  rss_size);
        if (ret)
                return ret;

        /* If Tunnel All Lookup isn't specified in the global RSS
         * Configuration, then we need to specify a default Ingress
         * Queue for any ingress packets which aren't hashed.  We'll
         * use our first ingress queue ...
         */
        return t4_config_vi_rss(adap, adap->mbox, viid,
                                FW_RSS_VI_CONFIG_CMD_IP6FOURTUPEN_F |
                                FW_RSS_VI_CONFIG_CMD_IP6TWOTUPEN_F |
                                FW_RSS_VI_CONFIG_CMD_IP4FOURTUPEN_F |
                                FW_RSS_VI_CONFIG_CMD_IP4TWOTUPEN_F |
                                FW_RSS_VI_CONFIG_CMD_UDPEN_F,
                                rss[0]);
}

/**
 *      cxgb4_write_rss - write the RSS table for a given port
 *      @pi: the port
 *      @queues: array of queue indices for RSS
 *
 *      Sets up the portion of the HW RSS table for the port's VI to distribute
 *      packets to the Rx queues in @queues.
 *      Should never be called before setting up sge eth rx queues
 */
int cxgb4_write_rss(const struct port_info *pi, const u16 *queues)
{
        struct adapter *adapter = pi->adapter;
        const struct sge_eth_rxq *rxq;
        int i, err;
        u16 *rss;

        rxq = &adapter->sge.ethrxq[pi->first_qset];
        rss = kmalloc_array(pi->rss_size, sizeof(u16), GFP_KERNEL);
        if (!rss)
                return -ENOMEM;

        /* map the queue indices to queue ids */
        for (i = 0; i < pi->rss_size; i++, queues++)
                rss[i] = rxq[*queues].rspq.abs_id;

        err = cxgb4_config_rss(pi, rss, pi->rss_size, pi->viid);
        kfree(rss);
        return err;
}

/**
 *      setup_rss - configure RSS
 *      @adap: the adapter
 *
 *      Sets up RSS for each port.
 */
static int setup_rss(struct adapter *adap)
{
        int i, j, err;

        for_each_port(adap, i) {
                const struct port_info *pi = adap2pinfo(adap, i);

                /* Fill default values with equal distribution */
                for (j = 0; j < pi->rss_size; j++)
                        pi->rss[j] = j % pi->nqsets;

                err = cxgb4_write_rss(pi, pi->rss);
                if (err)
                        return err;
        }
        return 0;
}

/*
 * Return the channel of the ingress queue with the given qid.
 */
static unsigned int rxq_to_chan(const struct sge *p, unsigned int qid)
{
        qid -= p->ingr_start;
        return netdev2pinfo(p->ingr_map[qid]->netdev)->tx_chan;
}

void cxgb4_quiesce_rx(struct sge_rspq *q)
{
        if (q->handler)
                napi_disable(&q->napi);
}

/*
 * Wait until all NAPI handlers are descheduled.
 */
static void quiesce_rx(struct adapter *adap)
{
        int i;

        for (i = 0; i < adap->sge.ingr_sz; i++) {
                struct sge_rspq *q = adap->sge.ingr_map[i];

                if (!q)
                        continue;

                cxgb4_quiesce_rx(q);
        }
}

/* Disable interrupt and napi handler */
static void disable_interrupts(struct adapter *adap)
{
        struct sge *s = &adap->sge;

        if (adap->flags & CXGB4_FULL_INIT_DONE) {
                t4_intr_disable(adap);
                if (adap->flags & CXGB4_USING_MSIX) {
                        free_msix_queue_irqs(adap);
                        free_irq(adap->msix_info[s->nd_msix_idx].vec,
                                 adap);
                } else {
                        free_irq(adap->pdev->irq, adap);
                }
                quiesce_rx(adap);
        }
}

void cxgb4_enable_rx(struct adapter *adap, struct sge_rspq *q)
{
        if (q->handler)
                napi_enable(&q->napi);

        /* 0-increment GTS to start the timer and enable interrupts */
        t4_write_reg(adap, MYPF_REG(SGE_PF_GTS_A),
                     SEINTARM_V(q->intr_params) |
                     INGRESSQID_V(q->cntxt_id));
}

/*
 * Enable NAPI scheduling and interrupt generation for all Rx queues.
 */
static void enable_rx(struct adapter *adap)
{
        int i;

        for (i = 0; i < adap->sge.ingr_sz; i++) {
                struct sge_rspq *q = adap->sge.ingr_map[i];

                if (!q)
                        continue;

                cxgb4_enable_rx(adap, q);
        }
}

static int setup_non_data_intr(struct adapter *adap)
{
        int msix;

        adap->sge.nd_msix_idx = -1;
        if (!(adap->flags & CXGB4_USING_MSIX))
                return 0;

        /* Request MSI-X vector for non-data interrupt */
        msix = cxgb4_get_msix_idx_from_bmap(adap);
        if (msix < 0)
                return -ENOMEM;

        snprintf(adap->msix_info[msix].desc,
                 sizeof(adap->msix_info[msix].desc),
                 "%s", adap->port[0]->name);

        adap->sge.nd_msix_idx = msix;

        return 0;
}

static int setup_fw_sge_queues(struct adapter *adap)
{
        struct sge *s = &adap->sge;
        int msix, err = 0;

        bitmap_zero(s->starving_fl, s->egr_sz);
        bitmap_zero(s->txq_maperr, s->egr_sz);

        if (adap->flags & CXGB4_USING_MSIX) {
                s->fwevtq_msix_idx = -1;
                msix = cxgb4_get_msix_idx_from_bmap(adap);
                if (msix < 0)
                        return -ENOMEM;

                snprintf(adap->msix_info[msix].desc,
                         sizeof(adap->msix_info[msix].desc),
                         "%s-FWeventq", adap->port[0]->name);
        } else {
                err = t4_sge_alloc_rxq(adap, &s->intrq, false, adap->port[0], 0,
                                       NULL, NULL, NULL, -1);
                if (err)
                        return err;
                msix = -((int)s->intrq.abs_id + 1);
        }

        err = t4_sge_alloc_rxq(adap, &s->fw_evtq, true, adap->port[0],
                               msix, NULL, fwevtq_handler, NULL, -1);
        if (err && msix >= 0)
                cxgb4_free_msix_idx_in_bmap(adap, msix);

        s->fwevtq_msix_idx = msix;
        return err;
}

/**
 *      setup_sge_queues - configure SGE Tx/Rx/response queues
 *      @adap: the adapter
 *
 *      Determines how many sets of SGE queues to use and initializes them.
 *      We support multiple queue sets per port if we have MSI-X, otherwise
 *      just one queue set per port.
 */
static int setup_sge_queues(struct adapter *adap)
{
        struct sge_uld_rxq_info *rxq_info = NULL;
        struct sge *s = &adap->sge;
        unsigned int cmplqid = 0;
        int err, i, j, msix = 0;

        if (is_uld(adap))
                rxq_info = s->uld_rxq_info[CXGB4_ULD_RDMA];

        if (!(adap->flags & CXGB4_USING_MSIX))
                msix = -((int)s->intrq.abs_id + 1);

        for_each_port(adap, i) {
                struct net_device *dev = adap->port[i];
                struct port_info *pi = netdev_priv(dev);
                struct sge_eth_rxq *q = &s->ethrxq[pi->first_qset];
                struct sge_eth_txq *t = &s->ethtxq[pi->first_qset];

                for (j = 0; j < pi->nqsets; j++, q++) {
                        if (msix >= 0) {
                                msix = cxgb4_get_msix_idx_from_bmap(adap);
                                if (msix < 0) {
                                        err = msix;
                                        goto freeout;
                                }

                                snprintf(adap->msix_info[msix].desc,
                                         sizeof(adap->msix_info[msix].desc),
                                         "%s-Rx%d", dev->name, j);
                                q->msix = &adap->msix_info[msix];
                        }

                        err = t4_sge_alloc_rxq(adap, &q->rspq, false, dev,
                                               msix, &q->fl,
                                               t4_ethrx_handler,
                                               NULL,
                                               t4_get_tp_ch_map(adap,
                                                                pi->tx_chan));
                        if (err)
                                goto freeout;
                        q->rspq.idx = j;
                        memset(&q->stats, 0, sizeof(q->stats));
                }

                q = &s->ethrxq[pi->first_qset];
                for (j = 0; j < pi->nqsets; j++, t++, q++) {
                        err = t4_sge_alloc_eth_txq(adap, t, dev,
                                        netdev_get_tx_queue(dev, j),
                                        q->rspq.cntxt_id,
                                        !!(adap->flags & CXGB4_SGE_DBQ_TIMER));
                        if (err)
                                goto freeout;
                }
        }

        for_each_port(adap, i) {
                /* Note that cmplqid below is 0 if we don't
                 * have RDMA queues, and that's the right value.
                 */
                if (rxq_info)
                        cmplqid = rxq_info->uldrxq[i].rspq.cntxt_id;

                err = t4_sge_alloc_ctrl_txq(adap, &s->ctrlq[i], adap->port[i],
                                            s->fw_evtq.cntxt_id, cmplqid);
                if (err)
                        goto freeout;
        }

        if (!is_t4(adap->params.chip)) {
                err = t4_sge_alloc_eth_txq(adap, &s->ptptxq, adap->port[0],
                                           netdev_get_tx_queue(adap->port[0], 0)
                                           , s->fw_evtq.cntxt_id, false);
                if (err)
                        goto freeout;
        }

        t4_write_reg(adap, is_t4(adap->params.chip) ?
                                MPS_TRC_RSS_CONTROL_A :
                                MPS_T5_TRC_RSS_CONTROL_A,
                     RSSCONTROL_V(netdev2pinfo(adap->port[0])->tx_chan) |
                     QUEUENUMBER_V(s->ethrxq[0].rspq.abs_id));
        return 0;
freeout:
        dev_err(adap->pdev_dev, "Can't allocate queues, err=%d\n", -err);
        t4_free_sge_resources(adap);
        return err;
}

static u16 cxgb_select_queue(struct net_device *dev, struct sk_buff *skb,
                             struct net_device *sb_dev)
{
        int txq;

#ifdef CONFIG_CHELSIO_T4_DCB
        /* If a Data Center Bridging has been successfully negotiated on this
         * link then we'll use the skb's priority to map it to a TX Queue.
         * The skb's priority is determined via the VLAN Tag Priority Code
         * Point field.
         */
        if (cxgb4_dcb_enabled(dev) && !is_kdump_kernel()) {
                u16 vlan_tci;
                int err;

                err = vlan_get_tag(skb, &vlan_tci);
                if (unlikely(err)) {
                        if (net_ratelimit())
                                netdev_warn(dev,
                                            "TX Packet without VLAN Tag on DCB Link\n");
                        txq = 0;
                } else {
                        txq = (vlan_tci & VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
#ifdef CONFIG_CHELSIO_T4_FCOE
                        if (skb->protocol == htons(ETH_P_FCOE))
                                txq = skb->priority & 0x7;
#endif /* CONFIG_CHELSIO_T4_FCOE */
                }
                return txq;
        }
#endif /* CONFIG_CHELSIO_T4_DCB */

        if (dev->num_tc) {
                struct port_info *pi = netdev2pinfo(dev);
                u8 ver, proto;

                ver = ip_hdr(skb)->version;
                proto = (ver == 6) ? ipv6_hdr(skb)->nexthdr :
                                     ip_hdr(skb)->protocol;

                /* Send unsupported traffic pattern to normal NIC queues. */
                txq = netdev_pick_tx(dev, skb, sb_dev);
                if (xfrm_offload(skb) || is_ptp_enabled(skb, dev) ||
                    skb->encapsulation ||
                    cxgb4_is_ktls_skb(skb) ||
                    (proto != IPPROTO_TCP && proto != IPPROTO_UDP))
                        txq = txq % pi->nqsets;

                return txq;
        }

        if (select_queue) {
                txq = (skb_rx_queue_recorded(skb)
                        ? skb_get_rx_queue(skb)
                        : smp_processor_id());

                while (unlikely(txq >= dev->real_num_tx_queues))
                        txq -= dev->real_num_tx_queues;

                return txq;
        }

        return netdev_pick_tx(dev, skb, NULL) % dev->real_num_tx_queues;
}

static int closest_timer(const struct sge *s, int time)
{
        int i, delta, match = 0, min_delta = INT_MAX;

        for (i = 0; i < ARRAY_SIZE(s->timer_val); i++) {
                delta = time - s->timer_val[i];
                if (delta < 0)
                        delta = -delta;
                if (delta < min_delta) {
                        min_delta = delta;
                        match = i;
                }
        }
        return match;
}

static int closest_thres(const struct sge *s, int thres)
{
        int i, delta, match = 0, min_delta = INT_MAX;

        for (i = 0; i < ARRAY_SIZE(s->counter_val); i++) {
                delta = thres - s->counter_val[i];
                if (delta < 0)
                        delta = -delta;
                if (delta < min_delta) {
                        min_delta = delta;
                        match = i;
                }
        }
        return match;
}

/**
 *      cxgb4_set_rspq_intr_params - set a queue's interrupt holdoff parameters
 *      @q: the Rx queue
 *      @us: the hold-off time in us, or 0 to disable timer
 *      @cnt: the hold-off packet count, or 0 to disable counter
 *
 *      Sets an Rx queue's interrupt hold-off time and packet count.  At least
 *      one of the two needs to be enabled for the queue to generate interrupts.
 */
int cxgb4_set_rspq_intr_params(struct sge_rspq *q,
                               unsigned int us, unsigned int cnt)
{
        struct adapter *adap = q->adap;

        if ((us | cnt) == 0)
                cnt = 1;

        if (cnt) {
                int err;
                u32 v, new_idx;

                new_idx = closest_thres(&adap->sge, cnt);
                if (q->desc && q->pktcnt_idx != new_idx) {
                        /* the queue has already been created, update it */
                        v = FW_PARAMS_MNEM_V(FW_PARAMS_MNEM_DMAQ) |
                            FW_PARAMS_PARAM_X_V(
                                        FW_PARAMS_PARAM_DMAQ_IQ_INTCNTTHRESH) |
                            FW_PARAMS_PARAM_YZ_V(q->cntxt_id);
                        err = t4_set_params(adap, adap->mbox, adap->pf, 0, 1,
                                            &v, &new_idx);
                        if (err)
                                return err;
                }
                q->pktcnt_idx = new_idx;
        }

        us = us == 0 ? 6 : closest_timer(&adap->sge, us);
        q->intr_params = QINTR_TIMER_IDX_V(us) | QINTR_CNT_EN_V(cnt > 0);
        return 0;
}

static int cxgb_set_features(struct net_device *dev, netdev_features_t features)
{
        netdev_features_t changed = dev->features ^ features;
        const struct port_info *pi = netdev_priv(dev);
        int err;

        if (!(changed & NETIF_F_HW_VLAN_CTAG_RX))
                return 0;

        err = t4_set_rxmode(pi->adapter, pi->adapter->mbox, pi->viid,
                            pi->viid_mirror, -1, -1, -1, -1,
                            !!(features & NETIF_F_HW_VLAN_CTAG_RX), true);
        if (unlikely(err))
                dev->features = features ^ NETIF_F_HW_VLAN_CTAG_RX;
        return err;
}

static int setup_debugfs(struct adapter *adap)
{
        if (IS_ERR_OR_NULL(adap->debugfs_root))
                return -1;

#ifdef CONFIG_DEBUG_FS
        t4_setup_debugfs(adap);
#endif
        return 0;
}

static void cxgb4_port_mirror_free_rxq(struct adapter *adap,
                                       struct sge_eth_rxq *mirror_rxq)
{
        if ((adap->flags & CXGB4_FULL_INIT_DONE) &&
            !(adap->flags & CXGB4_SHUTTING_DOWN))
                cxgb4_quiesce_rx(&mirror_rxq->rspq);

        if (adap->flags & CXGB4_USING_MSIX) {
                cxgb4_clear_msix_aff(mirror_rxq->msix->vec,
                                     mirror_rxq->msix->aff_mask);
                free_irq(mirror_rxq->msix->vec, &mirror_rxq->rspq);
                cxgb4_free_msix_idx_in_bmap(adap, mirror_rxq->msix->idx);
        }

        free_rspq_fl(adap, &mirror_rxq->rspq, &mirror_rxq->fl);
}

static int cxgb4_port_mirror_alloc_queues(struct net_device *dev)
{
        struct port_info *pi = netdev2pinfo(dev);
        struct adapter *adap = netdev2adap(dev);
        struct sge_eth_rxq *mirror_rxq;
        struct sge *s = &adap->sge;
        int ret = 0, msix = 0;
        u16 i, rxqid;
        u16 *rss;

        if (!pi->vi_mirror_count)
                return 0;

        if (s->mirror_rxq[pi->port_id])
                return 0;

        mirror_rxq = kcalloc(pi->nmirrorqsets, sizeof(*mirror_rxq), GFP_KERNEL);
        if (!mirror_rxq)
                return -ENOMEM;

        s->mirror_rxq[pi->port_id] = mirror_rxq;

        if (!(adap->flags & CXGB4_USING_MSIX))
                msix = -((int)adap->sge.intrq.abs_id + 1);

        for (i = 0, rxqid = 0; i < pi->nmirrorqsets; i++, rxqid++) {
                mirror_rxq = &s->mirror_rxq[pi->port_id][i];

                /* Allocate Mirror Rxqs */
                if (msix >= 0) {
                        msix = cxgb4_get_msix_idx_from_bmap(adap);
                        if (msix < 0) {
                                ret = msix;
                                goto out_free_queues;
                        }

                        mirror_rxq->msix = &adap->msix_info[msix];
                        snprintf(mirror_rxq->msix->desc,
                                 sizeof(mirror_rxq->msix->desc),
                                 "%s-mirrorrxq%d", dev->name, i);
                }

                init_rspq(adap, &mirror_rxq->rspq,
                          CXGB4_MIRROR_RXQ_DEFAULT_INTR_USEC,
                          CXGB4_MIRROR_RXQ_DEFAULT_PKT_CNT,
                          CXGB4_MIRROR_RXQ_DEFAULT_DESC_NUM,
                          CXGB4_MIRROR_RXQ_DEFAULT_DESC_SIZE);

                mirror_rxq->fl.size = CXGB4_MIRROR_FLQ_DEFAULT_DESC_NUM;

                ret = t4_sge_alloc_rxq(adap, &mirror_rxq->rspq, false,
                                       dev, msix, &mirror_rxq->fl,
                                       t4_ethrx_handler, NULL, 0);
                if (ret)
                        goto out_free_msix_idx;

                /* Setup MSI-X vectors for Mirror Rxqs */
                if (adap->flags & CXGB4_USING_MSIX) {
                        ret = request_irq(mirror_rxq->msix->vec,
                                          t4_sge_intr_msix, 0,
                                          mirror_rxq->msix->desc,
                                          &mirror_rxq->rspq);
                        if (ret)
                                goto out_free_rxq;

                        cxgb4_set_msix_aff(adap, mirror_rxq->msix->vec,
                                           &mirror_rxq->msix->aff_mask, i);
                }

                /* Start NAPI for Mirror Rxqs */
                cxgb4_enable_rx(adap, &mirror_rxq->rspq);
        }

        /* Setup RSS for Mirror Rxqs */
        rss = kcalloc(pi->rss_size, sizeof(u16), GFP_KERNEL);
        if (!rss) {
                ret = -ENOMEM;
                goto out_free_queues;
        }

        mirror_rxq = &s->mirror_rxq[pi->port_id][0];
        for (i = 0; i < pi->rss_size; i++)
                rss[i] = mirror_rxq[i % pi->nmirrorqsets].rspq.abs_id;

        ret = cxgb4_config_rss(pi, rss, pi->rss_size, pi->viid_mirror);
        kfree(rss);
        if (ret)
                goto out_free_queues;

        return 0;

out_free_rxq:
        free_rspq_fl(adap, &mirror_rxq->rspq, &mirror_rxq->fl);

out_free_msix_idx:
        cxgb4_free_msix_idx_in_bmap(adap, mirror_rxq->msix->idx);

out_free_queues:
        while (rxqid-- > 0)
                cxgb4_port_mirror_free_rxq(adap,
                                           &s->mirror_rxq[pi->port_id][rxqid]);

        kfree(s->mirror_rxq[pi->port_id]);
        s->mirror_rxq[pi->port_id] = NULL;
        return ret;
}

static void cxgb4_port_mirror_free_queues(struct net_device *dev)
{
        struct port_info *pi = netdev2pinfo(dev);
        struct adapter *adap = netdev2adap(dev);
        struct sge *s = &adap->sge;
        u16 i;

        if (!pi->vi_mirror_count)
                return;

        if (!s->mirror_rxq[pi->port_id])
                return;

        for (i = 0; i < pi->nmirrorqsets; i++)
                cxgb4_port_mirror_free_rxq(adap,
                                           &s->mirror_rxq[pi->port_id][i]);

        kfree(s->mirror_rxq[pi->port_id]);
        s->mirror_rxq[pi->port_id] = NULL;
}

static int cxgb4_port_mirror_start(struct net_device *dev)
{
        struct port_info *pi = netdev2pinfo(dev);
        struct adapter *adap = netdev2adap(dev);
        int ret, idx = -1;

        if (!pi->vi_mirror_count)
                return 0;

        /* Mirror VIs can be created dynamically after stack had
         * already setup Rx modes like MTU, promisc, allmulti, etc.
         * on main VI. So, parse what the stack had setup on the
         * main VI and update the same on the mirror VI.
         */
        ret = t4_set_rxmode(adap, adap->mbox, pi->viid, pi->viid_mirror,
                            dev->mtu, (dev->flags & IFF_PROMISC) ? 1 : 0,
                            (dev->flags & IFF_ALLMULTI) ? 1 : 0, 1,
                            !!(dev->features & NETIF_F_HW_VLAN_CTAG_RX), true);
        if (ret) {
                dev_err(adap->pdev_dev,
                        "Failed start up Rx mode for Mirror VI 0x%x, ret: %d\n",
                        pi->viid_mirror, ret);
                return ret;
        }

        /* Enable replication bit for the device's MAC address
         * in MPS TCAM, so that the packets for the main VI are
         * replicated to mirror VI.
         */
        ret = cxgb4_update_mac_filt(pi, pi->viid_mirror, &idx,
                                    dev->dev_addr, true, NULL);
        if (ret) {
                dev_err(adap->pdev_dev,
                        "Failed updating MAC filter for Mirror VI 0x%x, ret: %d\n",
                        pi->viid_mirror, ret);
                return ret;
        }

        /* Enabling a Virtual Interface can result in an interrupt
         * during the processing of the VI Enable command and, in some
         * paths, result in an attempt to issue another command in the
         * interrupt context. Thus, we disable interrupts during the
         * course of the VI Enable command ...
         */
        local_bh_disable();
        ret = t4_enable_vi_params(adap, adap->mbox, pi->viid_mirror, true, true,
                                  false);
        local_bh_enable();
        if (ret)
                dev_err(adap->pdev_dev,
                        "Failed starting Mirror VI 0x%x, ret: %d\n",
                        pi->viid_mirror, ret);

        return ret;
}

static void cxgb4_port_mirror_stop(struct net_device *dev)
{
        struct port_info *pi = netdev2pinfo(dev);
        struct adapter *adap = netdev2adap(dev);

        if (!pi->vi_mirror_count)
                return;

        t4_enable_vi_params(adap, adap->mbox, pi->viid_mirror, false, false,
                            false);
}

int cxgb4_port_mirror_alloc(struct net_device *dev)
{
        struct port_info *pi = netdev2pinfo(dev);
        struct adapter *adap = netdev2adap(dev);
        int ret = 0;

        if (!pi->nmirrorqsets)
                return -EOPNOTSUPP;

        mutex_lock(&pi->vi_mirror_mutex);
        if (pi->viid_mirror) {
                pi->vi_mirror_count++;
                goto out_unlock;
        }

        ret = t4_init_port_mirror(pi, adap->mbox, pi->port_id, adap->pf, 0,
                                  &pi->viid_mirror);
        if (ret)
                goto out_unlock;

        pi->vi_mirror_count = 1;

        if (adap->flags & CXGB4_FULL_INIT_DONE) {
                ret = cxgb4_port_mirror_alloc_queues(dev);
                if (ret)
                        goto out_free_vi;

                ret = cxgb4_port_mirror_start(dev);
                if (ret)
                        goto out_free_queues;
        }

        mutex_unlock(&pi->vi_mirror_mutex);
        return 0;

out_free_queues:
        cxgb4_port_mirror_free_queues(dev);

out_free_vi:
        pi->vi_mirror_count = 0;
        t4_free_vi(adap, adap->mbox, adap->pf, 0, pi->viid_mirror);
        pi->viid_mirror = 0;

out_unlock:
        mutex_unlock(&pi->vi_mirror_mutex);
        return ret;
}

void cxgb4_port_mirror_free(struct net_device *dev)
{
        struct port_info *pi = netdev2pinfo(dev);
        struct adapter *adap = netdev2adap(dev);

        mutex_lock(&pi->vi_mirror_mutex);
        if (!pi->viid_mirror)
                goto out_unlock;

        if (pi->vi_mirror_count > 1) {
                pi->vi_mirror_count--;
                goto out_unlock;
        }

        cxgb4_port_mirror_stop(dev);
        cxgb4_port_mirror_free_queues(dev);

        pi->vi_mirror_count = 0;
        t4_free_vi(adap, adap->mbox, adap->pf, 0, pi->viid_mirror);
        pi->viid_mirror = 0;

out_unlock:
        mutex_unlock(&pi->vi_mirror_mutex);
}

/*
 * upper-layer driver support
 */

/*
 * Allocate an active-open TID and set it to the supplied value.
 */
int cxgb4_alloc_atid(struct tid_info *t, void *data)
{
        int atid = -1;

        spin_lock_bh(&t->atid_lock);
        if (t->afree) {
                union aopen_entry *p = t->afree;

                atid = (p - t->atid_tab) + t->atid_base;
                t->afree = p->next;
                p->data = data;
                t->atids_in_use++;
        }
        spin_unlock_bh(&t->atid_lock);
        return atid;
}
EXPORT_SYMBOL(cxgb4_alloc_atid);

/*
 * Release an active-open TID.
 */
void cxgb4_free_atid(struct tid_info *t, unsigned int atid)
{
        union aopen_entry *p = &t->atid_tab[atid - t->atid_base];

        spin_lock_bh(&t->atid_lock);
        p->next = t->afree;
        t->afree = p;
        t->atids_in_use--;
        spin_unlock_bh(&t->atid_lock);
}
EXPORT_SYMBOL(cxgb4_free_atid);

/*
 * Allocate a server TID and set it to the supplied value.
 */
int cxgb4_alloc_stid(struct tid_info *t, int family, void *data)
{
        int stid;

        spin_lock_bh(&t->stid_lock);
        if (family == PF_INET) {
                stid = find_first_zero_bit(t->stid_bmap, t->nstids);
                if (stid < t->nstids)
                        __set_bit(stid, t->stid_bmap);
                else
                        stid = -1;
        } else {
                stid = bitmap_find_free_region(t->stid_bmap, t->nstids, 1);
                if (stid < 0)
                        stid = -1;
        }
        if (stid >= 0) {
                t->stid_tab[stid].data = data;
                stid += t->stid_base;
                /* IPv6 requires max of 520 bits or 16 cells in TCAM
                 * This is equivalent to 4 TIDs. With CLIP enabled it
                 * needs 2 TIDs.
                 */
                if (family == PF_INET6) {
                        t->stids_in_use += 2;
                        t->v6_stids_in_use += 2;
                } else {
                        t->stids_in_use++;
                }
        }
        spin_unlock_bh(&t->stid_lock);
        return stid;
}
EXPORT_SYMBOL(cxgb4_alloc_stid);

/* Allocate a server filter TID and set it to the supplied value.
 */
int cxgb4_alloc_sftid(struct tid_info *t, int family, void *data)
{
        int stid;

        spin_lock_bh(&t->stid_lock);
        if (family == PF_INET) {
                stid = find_next_zero_bit(t->stid_bmap,
                                t->nstids + t->nsftids, t->nstids);
                if (stid < (t->nstids + t->nsftids))
                        __set_bit(stid, t->stid_bmap);
                else
                        stid = -1;
        } else {
                stid = -1;
        }
        if (stid >= 0) {
                t->stid_tab[stid].data = data;
                stid -= t->nstids;
                stid += t->sftid_base;
                t->sftids_in_use++;
        }
        spin_unlock_bh(&t->stid_lock);
        return stid;
}
EXPORT_SYMBOL(cxgb4_alloc_sftid);

/* Release a server TID.
 */
void cxgb4_free_stid(struct tid_info *t, unsigned int stid, int family)
{
        /* Is it a server filter TID? */
        if (t->nsftids && (stid >= t->sftid_base)) {
                stid -= t->sftid_base;
                stid += t->nstids;
        } else {
                stid -= t->stid_base;
        }

        spin_lock_bh(&t->stid_lock);
        if (family == PF_INET)
                __clear_bit(stid, t->stid_bmap);
        else
                bitmap_release_region(t->stid_bmap, stid, 1);
        t->stid_tab[stid].data = NULL;
        if (stid < t->nstids) {
                if (family == PF_INET6) {
                        t->stids_in_use -= 2;
                        t->v6_stids_in_use -= 2;
                } else {
                        t->stids_in_use--;
                }
        } else {
                t->sftids_in_use--;
        }

        spin_unlock_bh(&t->stid_lock);
}
EXPORT_SYMBOL(cxgb4_free_stid);

/*
 * Populate a TID_RELEASE WR.  Caller must properly size the skb.
 */
static void mk_tid_release(struct sk_buff *skb, unsigned int chan,
                           unsigned int tid)
{
        struct cpl_tid_release *req;

        set_wr_txq(skb, CPL_PRIORITY_SETUP, chan);
        req = __skb_put(skb, sizeof(*req));
        INIT_TP_WR(req, tid);
        OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_TID_RELEASE, tid));
}

/*
 * Queue a TID release request and if necessary schedule a work queue to
 * process it.
 */
static void cxgb4_queue_tid_release(struct tid_info *t, unsigned int chan,
                                    unsigned int tid)
{
        struct adapter *adap = container_of(t, struct adapter, tids);
        void **p = &t->tid_tab[tid - t->tid_base];

        spin_lock_bh(&adap->tid_release_lock);
        *p = adap->tid_release_head;
        /* Low 2 bits encode the Tx channel number */
        adap->tid_release_head = (void **)((uintptr_t)p | chan);
        if (!adap->tid_release_task_busy) {
                adap->tid_release_task_busy = true;
                queue_work(adap->workq, &adap->tid_release_task);
        }
        spin_unlock_bh(&adap->tid_release_lock);
}

/*
 * Process the list of pending TID release requests.
 */
static void process_tid_release_list(struct work_struct *work)
{
        struct sk_buff *skb;
        struct adapter *adap;

        adap = container_of(work, struct adapter, tid_release_task);

        spin_lock_bh(&adap->tid_release_lock);
        while (adap->tid_release_head) {
                void **p = adap->tid_release_head;
                unsigned int chan = (uintptr_t)p & 3;
                p = (void *)p - chan;

                adap->tid_release_head = *p;
                *p = NULL;
                spin_unlock_bh(&adap->tid_release_lock);

                while (!(skb = alloc_skb(sizeof(struct cpl_tid_release),
                                         GFP_KERNEL)))
                        schedule_timeout_uninterruptible(1);

                mk_tid_release(skb, chan, p - adap->tids.tid_tab);
                t4_ofld_send(adap, skb);
                spin_lock_bh(&adap->tid_release_lock);
        }
        adap->tid_release_task_busy = false;
        spin_unlock_bh(&adap->tid_release_lock);
}

/*
 * Release a TID and inform HW.  If we are unable to allocate the release
 * message we defer to a work queue.
 */
void cxgb4_remove_tid(struct tid_info *t, unsigned int chan, unsigned int tid,
                      unsigned short family)
{
        struct adapter *adap = container_of(t, struct adapter, tids);
        struct sk_buff *skb;

        WARN_ON(tid_out_of_range(&adap->tids, tid));

        if (t->tid_tab[tid - adap->tids.tid_base]) {
                t->tid_tab[tid - adap->tids.tid_base] = NULL;
                atomic_dec(&t->conns_in_use);
                if (t->hash_base && (tid >= t->hash_base)) {
                        if (family == AF_INET6)
                                atomic_sub(2, &t->hash_tids_in_use);
                        else
                                atomic_dec(&t->hash_tids_in_use);
                } else {
                        if (family == AF_INET6)
                                atomic_sub(2, &t->tids_in_use);
                        else
                                atomic_dec(&t->tids_in_use);
                }
        }

        skb = alloc_skb(sizeof(struct cpl_tid_release), GFP_ATOMIC);
        if (likely(skb)) {
                mk_tid_release(skb, chan, tid);
                t4_ofld_send(adap, skb);
        } else
                cxgb4_queue_tid_release(t, chan, tid);
}
EXPORT_SYMBOL(cxgb4_remove_tid);

/*
 * Allocate and initialize the TID tables.  Returns 0 on success.
 */
static int tid_init(struct tid_info *t)
{
        struct adapter *adap = container_of(t, struct adapter, tids);
        unsigned int max_ftids = t->nftids + t->nsftids;
        unsigned int natids = t->natids;
        unsigned int hpftid_bmap_size;
        unsigned int eotid_bmap_size;
        unsigned int stid_bmap_size;
        unsigned int ftid_bmap_size;
        size_t size;

        stid_bmap_size = BITS_TO_LONGS(t->nstids + t->nsftids);
        ftid_bmap_size = BITS_TO_LONGS(t->nftids);
        hpftid_bmap_size = BITS_TO_LONGS(t->nhpftids);
        eotid_bmap_size = BITS_TO_LONGS(t->neotids);
        size = t->ntids * sizeof(*t->tid_tab) +
               natids * sizeof(*t->atid_tab) +
               t->nstids * sizeof(*t->stid_tab) +
               t->nsftids * sizeof(*t->stid_tab) +
               stid_bmap_size * sizeof(long) +
               t->nhpftids * sizeof(*t->hpftid_tab) +
               hpftid_bmap_size * sizeof(long) +
               max_ftids * sizeof(*t->ftid_tab) +
               ftid_bmap_size * sizeof(long) +
               t->neotids * sizeof(*t->eotid_tab) +
               eotid_bmap_size * sizeof(long);

        t->tid_tab = kvzalloc(size, GFP_KERNEL);
        if (!t->tid_tab)
                return -ENOMEM;

        t->atid_tab = (union aopen_entry *)&t->tid_tab[t->ntids];
        t->stid_tab = (struct serv_entry *)&t->atid_tab[natids];
        t->stid_bmap = (unsigned long *)&t->stid_tab[t->nstids + t->nsftids];
        t->hpftid_tab = (struct filter_entry *)&t->stid_bmap[stid_bmap_size];
        t->hpftid_bmap = (unsigned long *)&t->hpftid_tab[t->nhpftids];
        t->ftid_tab = (struct filter_entry *)&t->hpftid_bmap[hpftid_bmap_size];
        t->ftid_bmap = (unsigned long *)&t->ftid_tab[max_ftids];
        t->eotid_tab = (struct eotid_entry *)&t->ftid_bmap[ftid_bmap_size];
        t->eotid_bmap = (unsigned long *)&t->eotid_tab[t->neotids];
        spin_lock_init(&t->stid_lock);
        spin_lock_init(&t->atid_lock);
        spin_lock_init(&t->ftid_lock);

        t->stids_in_use = 0;
        t->v6_stids_in_use = 0;
        t->sftids_in_use = 0;
        t->afree = NULL;
        t->atids_in_use = 0;
        atomic_set(&t->tids_in_use, 0);
        atomic_set(&t->conns_in_use, 0);
        atomic_set(&t->hash_tids_in_use, 0);
        atomic_set(&t->eotids_in_use, 0);

        /* Setup the free list for atid_tab and clear the stid bitmap. */
        if (natids) {
                while (--natids)
                        t->atid_tab[natids - 1].next = &t->atid_tab[natids];
                t->afree = t->atid_tab;
        }

        if (is_offload(adap)) {
                bitmap_zero(t->stid_bmap, t->nstids + t->nsftids);
                /* Reserve stid 0 for T4/T5 adapters */
                if (!t->stid_base &&
                    CHELSIO_CHIP_VERSION(adap->params.chip) <= CHELSIO_T5)
                        __set_bit(0, t->stid_bmap);

                if (t->neotids)
                        bitmap_zero(t->eotid_bmap, t->neotids);
        }

        if (t->nhpftids)
                bitmap_zero(t->hpftid_bmap, t->nhpftids);
        bitmap_zero(t->ftid_bmap, t->nftids);
        return 0;
}

/**
 *      cxgb4_create_server - create an IP server
 *      @dev: the device
 *      @stid: the server TID
 *      @sip: local IP address to bind server to
 *      @sport: the server's TCP port
 *      @vlan: the VLAN header information
 *      @queue: queue to direct messages from this server to
 *
 *      Create an IP server for the given port and address.
 *      Returns <0 on error and one of the %NET_XMIT_* values on success.
 */
int cxgb4_create_server(const struct net_device *dev, unsigned int stid,
                        __be32 sip, __be16 sport, __be16 vlan,
                        unsigned int queue)
{
        unsigned int chan;
        struct sk_buff *skb;
        struct adapter *adap;
        struct cpl_pass_open_req *req;
        int ret;

        skb = alloc_skb(sizeof(*req), GFP_KERNEL);
        if (!skb)
                return -ENOMEM;

        adap = netdev2adap(dev);
        req = __skb_put(skb, sizeof(*req));
        INIT_TP_WR(req, 0);
        OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_PASS_OPEN_REQ, stid));
        req->local_port = sport;
        req->peer_port = htons(0);
        req->local_ip = sip;
        req->peer_ip = htonl(0);
        chan = rxq_to_chan(&adap->sge, queue);
        req->opt0 = cpu_to_be64(TX_CHAN_V(chan));
        req->opt1 = cpu_to_be64(CONN_POLICY_V(CPL_CONN_POLICY_ASK) |
                                SYN_RSS_ENABLE_F | SYN_RSS_QUEUE_V(queue));
        ret = t4_mgmt_tx(adap, skb);
        return net_xmit_eval(ret);
}
EXPORT_SYMBOL(cxgb4_create_server);

/*      cxgb4_create_server6 - create an IPv6 server
 *      @dev: the device
 *      @stid: the server TID
 *      @sip: local IPv6 address to bind server to
 *      @sport: the server's TCP port
 *      @queue: queue to direct messages from this server to
 *
 *      Create an IPv6 server for the given port and address.
 *      Returns <0 on error and one of the %NET_XMIT_* values on success.
 */
int cxgb4_create_server6(const struct net_device *dev, unsigned int stid,
                         const struct in6_addr *sip, __be16 sport,
                         unsigned int queue)
{
        unsigned int chan;
        struct sk_buff *skb;
        struct adapter *adap;
        struct cpl_pass_open_req6 *req;
        int ret;

        skb = alloc_skb(sizeof(*req), GFP_KERNEL);
        if (!skb)
                return -ENOMEM;

        adap = netdev2adap(dev);
        req = __skb_put(skb, sizeof(*req));
        INIT_TP_WR(req, 0);
        OPCODE_TID(req) = htonl(MK_OPCODE_TID(CPL_PASS_OPEN_REQ6, stid));
        req->local_port = sport;
        req->peer_port = htons(0);
        req->local_ip_hi = *(__be64 *)(sip->s6_addr);
        req->local_ip_lo = *(__be64 *)(sip->s6_addr + 8);
        req->peer_ip_hi = cpu_to_be64(0);
        req->peer_ip_lo = cpu_to_be64(0);
        chan = rxq_to_chan(&adap->sge, queue);
        req->opt0 = cpu_to_be64(TX_CHAN_V(chan));
        req->opt1 = cpu_to_be64(CONN_POLICY_V(CPL_CONN_POLICY_ASK) |
                                SYN_RSS_ENABLE_F | SYN_RSS_QUEUE_V(queue));
        ret = t4_mgmt_tx(adap, skb);
        return net_xmit_eval(ret);
}
EXPORT_SYMBOL(cxgb4_create_server6);

int cxgb4_remove_server(const struct net_device *dev, unsigned int stid,
                        unsigned int queue, bool ipv6)
{
        struct sk_buff *skb;