[mirror_ubuntu-zesty-kernel.git] / drivers / infiniband / hw / ocrdma / ocrdma_main.c

/*******************************************************************
 * This file is part of the Emulex RoCE Device Driver for          *
 * RoCE (RDMA over Converged Ethernet) adapters.                   *
 * Copyright (C) 2008-2012 Emulex. All rights reserved.            *
 * EMULEX and SLI are trademarks of Emulex.                        *
 * www.emulex.com                                                  *
 *                                                                 *
 * This program is free software; you can redistribute it and/or   *
 * modify it under the terms of version 2 of the GNU General       *
 * Public License as published by the Free Software Foundation.    *
 * This program is distributed in the hope that it will be useful. *
 * ALL EXPRESS OR IMPLIED CONDITIONS, REPRESENTATIONS AND          *
 * WARRANTIES, INCLUDING ANY IMPLIED WARRANTY OF MERCHANTABILITY,  *
 * FITNESS FOR A PARTICULAR PURPOSE, OR NON-INFRINGEMENT, ARE      *
 * DISCLAIMED, EXCEPT TO THE EXTENT THAT SUCH DISCLAIMERS ARE HELD *
 * TO BE LEGALLY INVALID.  See the GNU General Public License for  *
 * more details, a copy of which can be found in the file COPYING  *
 * included with this package.                                     *
 *
 * Contact Information:
 * linux-drivers@emulex.com
 *
 * Emulex
 * 3333 Susan Street
 * Costa Mesa, CA 92626
 *******************************************************************/

#include <linux/module.h>
#include <linux/version.h>
#include <linux/idr.h>
#include <rdma/ib_verbs.h>
#include <rdma/ib_user_verbs.h>
#include <rdma/ib_addr.h>

#include <linux/netdevice.h>
#include <net/addrconf.h>

#include "ocrdma.h"
#include "ocrdma_verbs.h"
#include "ocrdma_ah.h"
#include "be_roce.h"
#include "ocrdma_hw.h"

MODULE_VERSION(OCRDMA_ROCE_DEV_VERSION);
MODULE_DESCRIPTION("Emulex RoCE HCA Driver");
MODULE_AUTHOR("Emulex Corporation");
MODULE_LICENSE("GPL");

static LIST_HEAD(ocrdma_dev_list);
static DEFINE_SPINLOCK(ocrdma_devlist_lock);
static DEFINE_IDR(ocrdma_dev_id);

static union ib_gid ocrdma_zero_sgid;
static int ocrdma_inet6addr_event(struct notifier_block *,
				  unsigned long, void *);

static struct notifier_block ocrdma_inet6addr_notifier = {
	.notifier_call = ocrdma_inet6addr_event
};

static int ocrdma_get_instance(void)
{
	int instance = 0;

	/* Assign an unused number */
	if (!idr_pre_get(&ocrdma_dev_id, GFP_KERNEL))
		return -1;
	if (idr_get_new(&ocrdma_dev_id, NULL, &instance))
		return -1;
	return instance;
}

void ocrdma_get_guid(struct ocrdma_dev *dev, u8 *guid)
{
	u8 mac_addr[6];

	memcpy(&mac_addr[0], &dev->nic_info.mac_addr[0], ETH_ALEN);
	guid[0] = mac_addr[0] ^ 2;
	guid[1] = mac_addr[1];
	guid[2] = mac_addr[2];
	guid[3] = 0xff;
	guid[4] = 0xfe;
	guid[5] = mac_addr[3];
	guid[6] = mac_addr[4];
	guid[7] = mac_addr[5];
}

static void ocrdma_build_sgid_mac(union ib_gid *sgid, unsigned char *mac_addr,
				  bool is_vlan, u16 vlan_id)
{
	sgid->global.subnet_prefix = cpu_to_be64(0xfe80000000000000LL);
	sgid->raw[8] = mac_addr[0] ^ 2;
	sgid->raw[9] = mac_addr[1];
	sgid->raw[10] = mac_addr[2];
	if (is_vlan) {
		sgid->raw[11] = vlan_id >> 8;
		sgid->raw[12] = vlan_id & 0xff;
	} else {
		sgid->raw[11] = 0xff;
		sgid->raw[12] = 0xfe;
	}
	sgid->raw[13] = mac_addr[3];
	sgid->raw[14] = mac_addr[4];
	sgid->raw[15] = mac_addr[5];
}

static void ocrdma_add_sgid(struct ocrdma_dev *dev, unsigned char *mac_addr,
			    bool is_vlan, u16 vlan_id)
{
	int i;
	bool found = false;
	union ib_gid new_sgid;
	int free_idx = OCRDMA_MAX_SGID;
	unsigned long flags;

	memset(&ocrdma_zero_sgid, 0, sizeof(union ib_gid));

	ocrdma_build_sgid_mac(&new_sgid, mac_addr, is_vlan, vlan_id);

	spin_lock_irqsave(&dev->sgid_lock, flags);
	for (i = 0; i < OCRDMA_MAX_SGID; i++) {
		if (!memcmp(&dev->sgid_tbl[i], &ocrdma_zero_sgid,
			    sizeof(union ib_gid))) {
			/* found free entry */
			if (!found) {
				free_idx = i;
				found = true;
				break;
			}
		} else if (!memcmp(&dev->sgid_tbl[i], &new_sgid,
				   sizeof(union ib_gid))) {
			/* entry already present, no addition is required. */
			spin_unlock_irqrestore(&dev->sgid_lock, flags);
			return;
		}
	}
	/* if entry doesn't exist and if table has some space, add entry */
	if (found)
		memcpy(&dev->sgid_tbl[free_idx], &new_sgid,
		       sizeof(union ib_gid));
	spin_unlock_irqrestore(&dev->sgid_lock, flags);
}

static bool ocrdma_del_sgid(struct ocrdma_dev *dev, unsigned char *mac_addr,
			    bool is_vlan, u16 vlan_id)
{
	int found = false;
	int i;
	union ib_gid sgid;
	unsigned long flags;

	ocrdma_build_sgid_mac(&sgid, mac_addr, is_vlan, vlan_id);

	spin_lock_irqsave(&dev->sgid_lock, flags);
	/* first is default sgid, which cannot be deleted. */
	for (i = 1; i < OCRDMA_MAX_SGID; i++) {
		if (!memcmp(&dev->sgid_tbl[i], &sgid, sizeof(union ib_gid))) {
			/* found matching entry */
			memset(&dev->sgid_tbl[i], 0, sizeof(union ib_gid));
			found = true;
			break;
		}
	}
	spin_unlock_irqrestore(&dev->sgid_lock, flags);
	return found;
}

static void ocrdma_add_default_sgid(struct ocrdma_dev *dev)
{
	/* GID Index 0 - Invariant manufacturer-assigned EUI-64 */
	union ib_gid *sgid = &dev->sgid_tbl[0];

	sgid->global.subnet_prefix = cpu_to_be64(0xfe80000000000000LL);
	ocrdma_get_guid(dev, &sgid->raw[8]);
}

static int ocrdma_build_sgid_tbl(struct ocrdma_dev *dev)
{
	struct net_device *netdev, *tmp;
	u16 vlan_id;
	bool is_vlan;

	netdev = dev->nic_info.netdev;

	ocrdma_add_default_sgid(dev);

	rcu_read_lock();
	for_each_netdev_rcu(&init_net, tmp) {
		if (netdev == tmp || vlan_dev_real_dev(tmp) == netdev) {
			if (!netif_running(tmp) || !netif_oper_up(tmp))
				continue;
			if (netdev != tmp) {
				vlan_id = vlan_dev_vlan_id(tmp);
				is_vlan = true;
			} else {
				is_vlan = false;
				vlan_id = 0;
				tmp = netdev;
			}
			ocrdma_add_sgid(dev, tmp->dev_addr, is_vlan, vlan_id);
		}
	}
	rcu_read_unlock();
	return 0;
}

static int ocrdma_inet6addr_event(struct notifier_block *notifier,
				  unsigned long event, void *ptr)
{
	struct inet6_ifaddr *ifa = (struct inet6_ifaddr *)ptr;
	struct net_device *event_netdev = ifa->idev->dev;
	struct net_device *netdev = NULL;
	struct ib_event gid_event;
	struct ocrdma_dev *dev;
	bool found = false;
	bool is_vlan = false;
	u16 vid = 0;

	netdev = vlan_dev_real_dev(event_netdev);
	if (netdev != event_netdev) {
		is_vlan = true;
		vid = vlan_dev_vlan_id(event_netdev);
	}
	rcu_read_lock();
	list_for_each_entry_rcu(dev, &ocrdma_dev_list, entry) {
		if (dev->nic_info.netdev == netdev) {
			found = true;
			break;
		}
	}
	rcu_read_unlock();

	if (!found)
		return NOTIFY_DONE;
	if (!rdma_link_local_addr((struct in6_addr *)&ifa->addr))
		return NOTIFY_DONE;

	mutex_lock(&dev->dev_lock);
	switch (event) {
	case NETDEV_UP:
		ocrdma_add_sgid(dev, netdev->dev_addr, is_vlan, vid);
		break;
	case NETDEV_DOWN:
		found = ocrdma_del_sgid(dev, netdev->dev_addr, is_vlan, vid);
		if (found) {
			/* found the matching entry, notify
			 * the consumers about it
			 */
			gid_event.device = &dev->ibdev;
			gid_event.element.port_num = 1;
			gid_event.event = IB_EVENT_GID_CHANGE;
			ib_dispatch_event(&gid_event);
		}
		break;
	default:
		break;
	}
	mutex_unlock(&dev->dev_lock);
	return NOTIFY_OK;
}

static enum rdma_link_layer ocrdma_link_layer(struct ib_device *device,
					      u8 port_num)
{
	return IB_LINK_LAYER_ETHERNET;
}

static int ocrdma_register_device(struct ocrdma_dev *dev)
{
	strlcpy(dev->ibdev.name, "ocrdma%d", IB_DEVICE_NAME_MAX);
	ocrdma_get_guid(dev, (u8 *)&dev->ibdev.node_guid);
	memcpy(dev->ibdev.node_desc, OCRDMA_NODE_DESC,
	       sizeof(OCRDMA_NODE_DESC));
	dev->ibdev.owner = THIS_MODULE;
	dev->ibdev.uverbs_cmd_mask =
	    OCRDMA_UVERBS(GET_CONTEXT) |
	    OCRDMA_UVERBS(QUERY_DEVICE) |
	    OCRDMA_UVERBS(QUERY_PORT) |
	    OCRDMA_UVERBS(ALLOC_PD) |
	    OCRDMA_UVERBS(DEALLOC_PD) |
	    OCRDMA_UVERBS(REG_MR) |
	    OCRDMA_UVERBS(DEREG_MR) |
	    OCRDMA_UVERBS(CREATE_COMP_CHANNEL) |
	    OCRDMA_UVERBS(CREATE_CQ) |
	    OCRDMA_UVERBS(RESIZE_CQ) |
	    OCRDMA_UVERBS(DESTROY_CQ) |
	    OCRDMA_UVERBS(REQ_NOTIFY_CQ) |
	    OCRDMA_UVERBS(CREATE_QP) |
	    OCRDMA_UVERBS(MODIFY_QP) |
	    OCRDMA_UVERBS(QUERY_QP) |
	    OCRDMA_UVERBS(DESTROY_QP) |
	    OCRDMA_UVERBS(POLL_CQ) |
	    OCRDMA_UVERBS(POST_SEND) |
	    OCRDMA_UVERBS(POST_RECV);

	dev->ibdev.uverbs_cmd_mask |=
	    OCRDMA_UVERBS(CREATE_AH) |
	     OCRDMA_UVERBS(MODIFY_AH) |
	     OCRDMA_UVERBS(QUERY_AH) |
	     OCRDMA_UVERBS(DESTROY_AH);

	dev->ibdev.node_type = RDMA_NODE_IB_CA;
	dev->ibdev.phys_port_cnt = 1;
	dev->ibdev.num_comp_vectors = 1;

	/* mandatory verbs. */
	dev->ibdev.query_device = ocrdma_query_device;
	dev->ibdev.query_port = ocrdma_query_port;
	dev->ibdev.modify_port = ocrdma_modify_port;
	dev->ibdev.query_gid = ocrdma_query_gid;
	dev->ibdev.get_link_layer = ocrdma_link_layer;
	dev->ibdev.alloc_pd = ocrdma_alloc_pd;
	dev->ibdev.dealloc_pd = ocrdma_dealloc_pd;

	dev->ibdev.create_cq = ocrdma_create_cq;
	dev->ibdev.destroy_cq = ocrdma_destroy_cq;
	dev->ibdev.resize_cq = ocrdma_resize_cq;

	dev->ibdev.create_qp = ocrdma_create_qp;
	dev->ibdev.modify_qp = ocrdma_modify_qp;
	dev->ibdev.query_qp = ocrdma_query_qp;
	dev->ibdev.destroy_qp = ocrdma_destroy_qp;

	dev->ibdev.query_pkey = ocrdma_query_pkey;
	dev->ibdev.create_ah = ocrdma_create_ah;
	dev->ibdev.destroy_ah = ocrdma_destroy_ah;
	dev->ibdev.query_ah = ocrdma_query_ah;
	dev->ibdev.modify_ah = ocrdma_modify_ah;

	dev->ibdev.poll_cq = ocrdma_poll_cq;
	dev->ibdev.post_send = ocrdma_post_send;
	dev->ibdev.post_recv = ocrdma_post_recv;
	dev->ibdev.req_notify_cq = ocrdma_arm_cq;

	dev->ibdev.get_dma_mr = ocrdma_get_dma_mr;
	dev->ibdev.dereg_mr = ocrdma_dereg_mr;
	dev->ibdev.reg_user_mr = ocrdma_reg_user_mr;

	/* mandatory to support user space verbs consumer. */
	dev->ibdev.alloc_ucontext = ocrdma_alloc_ucontext;
	dev->ibdev.dealloc_ucontext = ocrdma_dealloc_ucontext;
	dev->ibdev.mmap = ocrdma_mmap;
	dev->ibdev.dma_device = &dev->nic_info.pdev->dev;

	dev->ibdev.process_mad = ocrdma_process_mad;

	if (dev->nic_info.dev_family == OCRDMA_GEN2_FAMILY) {
		dev->ibdev.uverbs_cmd_mask |=
		     OCRDMA_UVERBS(CREATE_SRQ) |
		     OCRDMA_UVERBS(MODIFY_SRQ) |
		     OCRDMA_UVERBS(QUERY_SRQ) |
		     OCRDMA_UVERBS(DESTROY_SRQ) |
		     OCRDMA_UVERBS(POST_SRQ_RECV);

		dev->ibdev.create_srq = ocrdma_create_srq;
		dev->ibdev.modify_srq = ocrdma_modify_srq;
		dev->ibdev.query_srq = ocrdma_query_srq;
		dev->ibdev.destroy_srq = ocrdma_destroy_srq;
		dev->ibdev.post_srq_recv = ocrdma_post_srq_recv;
	}
	return ib_register_device(&dev->ibdev, NULL);
}

static int ocrdma_alloc_resources(struct ocrdma_dev *dev)
{
	mutex_init(&dev->dev_lock);
	dev->sgid_tbl = kzalloc(sizeof(union ib_gid) *
				OCRDMA_MAX_SGID, GFP_KERNEL);
	if (!dev->sgid_tbl)
		goto alloc_err;
	spin_lock_init(&dev->sgid_lock);

	dev->cq_tbl = kzalloc(sizeof(struct ocrdma_cq *) *
			      OCRDMA_MAX_CQ, GFP_KERNEL);
	if (!dev->cq_tbl)
		goto alloc_err;

	if (dev->attr.max_qp) {
		dev->qp_tbl = kzalloc(sizeof(struct ocrdma_qp *) *
				      OCRDMA_MAX_QP, GFP_KERNEL);
		if (!dev->qp_tbl)
			goto alloc_err;
	}
	spin_lock_init(&dev->av_tbl.lock);
	spin_lock_init(&dev->flush_q_lock);
	return 0;
alloc_err:
	ocrdma_err("%s(%d) error.\n", __func__, dev->id);
	return -ENOMEM;
}

static void ocrdma_free_resources(struct ocrdma_dev *dev)
{
	kfree(dev->qp_tbl);
	kfree(dev->cq_tbl);
	kfree(dev->sgid_tbl);
}

static struct ocrdma_dev *ocrdma_add(struct be_dev_info *dev_info)
{
	int status = 0;
	struct ocrdma_dev *dev;

	dev = (struct ocrdma_dev *)ib_alloc_device(sizeof(struct ocrdma_dev));
	if (!dev) {
		ocrdma_err("Unable to allocate ib device\n");
		return NULL;
	}
	dev->mbx_cmd = kzalloc(sizeof(struct ocrdma_mqe_emb_cmd), GFP_KERNEL);
	if (!dev->mbx_cmd)
		goto idr_err;

	memcpy(&dev->nic_info, dev_info, sizeof(*dev_info));
	dev->id = ocrdma_get_instance();
	if (dev->id < 0)
		goto idr_err;

	status = ocrdma_init_hw(dev);
	if (status)
		goto init_err;

	status = ocrdma_alloc_resources(dev);
	if (status)
		goto alloc_err;

	status = ocrdma_build_sgid_tbl(dev);
	if (status)
		goto alloc_err;

	status = ocrdma_register_device(dev);
	if (status)
		goto alloc_err;

	spin_lock(&ocrdma_devlist_lock);
	list_add_tail_rcu(&dev->entry, &ocrdma_dev_list);
	spin_unlock(&ocrdma_devlist_lock);
	return dev;

alloc_err:
	ocrdma_free_resources(dev);
	ocrdma_cleanup_hw(dev);
init_err:
	idr_remove(&ocrdma_dev_id, dev->id);
idr_err:
	kfree(dev->mbx_cmd);
	ib_dealloc_device(&dev->ibdev);
	ocrdma_err("%s() leaving. ret=%d\n", __func__, status);
	return NULL;
}

static void ocrdma_remove_free(struct rcu_head *rcu)
{
	struct ocrdma_dev *dev = container_of(rcu, struct ocrdma_dev, rcu);

	ocrdma_free_resources(dev);
	ocrdma_cleanup_hw(dev);

	idr_remove(&ocrdma_dev_id, dev->id);
	kfree(dev->mbx_cmd);
	ib_dealloc_device(&dev->ibdev);
}

static void ocrdma_remove(struct ocrdma_dev *dev)
{
	/* first unregister with stack to stop all the active traffic
	 * of the registered clients.
	 */
	ib_unregister_device(&dev->ibdev);

	spin_lock(&ocrdma_devlist_lock);
	list_del_rcu(&dev->entry);
	spin_unlock(&ocrdma_devlist_lock);
	call_rcu(&dev->rcu, ocrdma_remove_free);
}

static int ocrdma_open(struct ocrdma_dev *dev)
{
	struct ib_event port_event;

	port_event.event = IB_EVENT_PORT_ACTIVE;
	port_event.element.port_num = 1;
	port_event.device = &dev->ibdev;
	ib_dispatch_event(&port_event);
	return 0;
}

static int ocrdma_close(struct ocrdma_dev *dev)
{
	int i;
	struct ocrdma_qp *qp, **cur_qp;
	struct ib_event err_event;
	struct ib_qp_attr attrs;
	int attr_mask = IB_QP_STATE;

	attrs.qp_state = IB_QPS_ERR;
	mutex_lock(&dev->dev_lock);
	if (dev->qp_tbl) {
		cur_qp = dev->qp_tbl;
		for (i = 0; i < OCRDMA_MAX_QP; i++) {
			qp = cur_qp[i];
			if (qp) {
				/* change the QP state to ERROR */
				_ocrdma_modify_qp(&qp->ibqp, &attrs, attr_mask);

				err_event.event = IB_EVENT_QP_FATAL;
				err_event.element.qp = &qp->ibqp;
				err_event.device = &dev->ibdev;
				ib_dispatch_event(&err_event);
			}
		}
	}
	mutex_unlock(&dev->dev_lock);

	err_event.event = IB_EVENT_PORT_ERR;
	err_event.element.port_num = 1;
	err_event.device = &dev->ibdev;
	ib_dispatch_event(&err_event);
	return 0;
}

/* event handling via NIC driver ensures that all the NIC specific
 * initialization done before RoCE driver notifies
 * event to stack.
 */
static void ocrdma_event_handler(struct ocrdma_dev *dev, u32 event)
{
	switch (event) {
	case BE_DEV_UP:
		ocrdma_open(dev);
		break;
	case BE_DEV_DOWN:
		ocrdma_close(dev);
		break;
	};
}

static struct ocrdma_driver ocrdma_drv = {
	.name			= "ocrdma_driver",
	.add			= ocrdma_add,
	.remove			= ocrdma_remove,
	.state_change_handler	= ocrdma_event_handler,
};

static int __init ocrdma_init_module(void)
{
	int status;

	status = register_inet6addr_notifier(&ocrdma_inet6addr_notifier);
	if (status)
		return status;
	status = be_roce_register_driver(&ocrdma_drv);
	if (status)
		unregister_inet6addr_notifier(&ocrdma_inet6addr_notifier);
	return status;
}

static void __exit ocrdma_exit_module(void)
{
	be_roce_unregister_driver(&ocrdma_drv);
	unregister_inet6addr_notifier(&ocrdma_inet6addr_notifier);
}

module_init(ocrdma_init_module);
module_exit(ocrdma_exit_module);
Commit	Line	Data
fe2caefc PP	1	/*******************************************************************
	2	* This file is part of the Emulex RoCE Device Driver for *
	3	* RoCE (RDMA over Converged Ethernet) adapters. *
	4	* Copyright (C) 2008-2012 Emulex. All rights reserved. *
	5	* EMULEX and SLI are trademarks of Emulex. *
	6	* www.emulex.com *
	7	* *
	8	* This program is free software; you can redistribute it and/or *
	9	* modify it under the terms of version 2 of the GNU General *
	10	* Public License as published by the Free Software Foundation. *
	11	* This program is distributed in the hope that it will be useful. *
	12	* ALL EXPRESS OR IMPLIED CONDITIONS, REPRESENTATIONS AND *
	13	* WARRANTIES, INCLUDING ANY IMPLIED WARRANTY OF MERCHANTABILITY, *
	14	* FITNESS FOR A PARTICULAR PURPOSE, OR NON-INFRINGEMENT, ARE *
	15	* DISCLAIMED, EXCEPT TO THE EXTENT THAT SUCH DISCLAIMERS ARE HELD *
	16	* TO BE LEGALLY INVALID. See the GNU General Public License for *
	17	* more details, a copy of which can be found in the file COPYING *
	18	* included with this package. *
	19	*
	20	* Contact Information:
	21	* linux-drivers@emulex.com
	22	*
	23	* Emulex
	24	* 3333 Susan Street
	25	* Costa Mesa, CA 92626
	26	*******************************************************************/
	27
	28	#include <linux/module.h>
	29	#include <linux/version.h>
	30	#include <linux/idr.h>
	31	#include <rdma/ib_verbs.h>
	32	#include <rdma/ib_user_verbs.h>
	33	#include <rdma/ib_addr.h>
	34
	35	#include <linux/netdevice.h>
	36	#include <net/addrconf.h>
	37
	38	#include "ocrdma.h"
	39	#include "ocrdma_verbs.h"
	40	#include "ocrdma_ah.h"
	41	#include "be_roce.h"
	42	#include "ocrdma_hw.h"
	43
	44	MODULE_VERSION(OCRDMA_ROCE_DEV_VERSION);
	45	MODULE_DESCRIPTION("Emulex RoCE HCA Driver");
	46	MODULE_AUTHOR("Emulex Corporation");
	47	MODULE_LICENSE("GPL");
	48
	49	static LIST_HEAD(ocrdma_dev_list);
3e4d60a8	50	static DEFINE_SPINLOCK(ocrdma_devlist_lock);
fe2caefc PP	51	static DEFINE_IDR(ocrdma_dev_id);
	52
	53	static union ib_gid ocrdma_zero_sgid;
	54	static int ocrdma_inet6addr_event(struct notifier_block *,
	55	unsigned long, void *);
	56
	57	static struct notifier_block ocrdma_inet6addr_notifier = {
	58	.notifier_call = ocrdma_inet6addr_event
	59	};
	60
abe3afac	61	static int ocrdma_get_instance(void)
fe2caefc PP	62	{
	63	int instance = 0;
	64
	65	/* Assign an unused number */
	66	if (!idr_pre_get(&ocrdma_dev_id, GFP_KERNEL))
	67	return -1;
	68	if (idr_get_new(&ocrdma_dev_id, NULL, &instance))
	69	return -1;
	70	return instance;
	71	}
	72
	73	void ocrdma_get_guid(struct ocrdma_dev dev, u8 guid)
	74	{
	75	u8 mac_addr[6];
	76
	77	memcpy(&mac_addr[0], &dev->nic_info.mac_addr[0], ETH_ALEN);
	78	guid[0] = mac_addr[0] ^ 2;
	79	guid[1] = mac_addr[1];
	80	guid[2] = mac_addr[2];
	81	guid[3] = 0xff;
	82	guid[4] = 0xfe;
	83	guid[5] = mac_addr[3];
	84	guid[6] = mac_addr[4];
	85	guid[7] = mac_addr[5];
	86	}
	87
	88	static void ocrdma_build_sgid_mac(union ib_gid sgid, unsigned char mac_addr,
	89	bool is_vlan, u16 vlan_id)
	90	{
	91	sgid->global.subnet_prefix = cpu_to_be64(0xfe80000000000000LL);
	92	sgid->raw[8] = mac_addr[0] ^ 2;
	93	sgid->raw[9] = mac_addr[1];
	94	sgid->raw[10] = mac_addr[2];
	95	if (is_vlan) {
	96	sgid->raw[11] = vlan_id >> 8;
	97	sgid->raw[12] = vlan_id & 0xff;
	98	} else {
	99	sgid->raw[11] = 0xff;
	100	sgid->raw[12] = 0xfe;
	101	}
	102	sgid->raw[13] = mac_addr[3];
	103	sgid->raw[14] = mac_addr[4];
	104	sgid->raw[15] = mac_addr[5];
	105	}
	106
	107	static void ocrdma_add_sgid(struct ocrdma_dev dev, unsigned char mac_addr,
	108	bool is_vlan, u16 vlan_id)
	109	{
	110	int i;
	111	bool found = false;
	112	union ib_gid new_sgid;
	113	int free_idx = OCRDMA_MAX_SGID;
	114	unsigned long flags;
	115
	116	memset(&ocrdma_zero_sgid, 0, sizeof(union ib_gid));
	117
	118	ocrdma_build_sgid_mac(&new_sgid, mac_addr, is_vlan, vlan_id);
	119
	120	spin_lock_irqsave(&dev->sgid_lock, flags);
	121	for (i = 0; i < OCRDMA_MAX_SGID; i++) {
	122	if (!memcmp(&dev->sgid_tbl[i], &ocrdma_zero_sgid,
	123	sizeof(union ib_gid))) {
	124	/* found free entry */
	125	if (!found) {
126	free_idx = i;
127	found = true;
128	break;
129	}
130	} else if (!memcmp(&dev->sgid_tbl[i], &new_sgid,
131	sizeof(union ib_gid))) {
132	/* entry already present, no addition is required. */
133	spin_unlock_irqrestore(&dev->sgid_lock, flags);
134	return;
135	}
136	}
137	/* if entry doesn't exist and if table has some space, add entry */
138	if (found)
139	memcpy(&dev->sgid_tbl[free_idx], &new_sgid,
140	sizeof(union ib_gid));
141	spin_unlock_irqrestore(&dev->sgid_lock, flags);
142	}
143
144	static bool ocrdma_del_sgid(struct ocrdma_dev dev, unsigned char mac_addr,
145	bool is_vlan, u16 vlan_id)
146	{
147	int found = false;
148	int i;
149	union ib_gid sgid;
150	unsigned long flags;
151
152	ocrdma_build_sgid_mac(&sgid, mac_addr, is_vlan, vlan_id);
153
154	spin_lock_irqsave(&dev->sgid_lock, flags);
155	/* first is default sgid, which cannot be deleted. */
156	for (i = 1; i < OCRDMA_MAX_SGID; i++) {
157	if (!memcmp(&dev->sgid_tbl[i], &sgid, sizeof(union ib_gid))) {
158	/* found matching entry */
159	memset(&dev->sgid_tbl[i], 0, sizeof(union ib_gid));
160	found = true;
161	break;
162	}
163	}
164	spin_unlock_irqrestore(&dev->sgid_lock, flags);
165	return found;
166	}
167
168	static void ocrdma_add_default_sgid(struct ocrdma_dev *dev)
169	{
170	/* GID Index 0 - Invariant manufacturer-assigned EUI-64 */
171	union ib_gid *sgid = &dev->sgid_tbl[0];
172
173	sgid->global.subnet_prefix = cpu_to_be64(0xfe80000000000000LL);
174	ocrdma_get_guid(dev, &sgid->raw[8]);
175	}
176
177	static int ocrdma_build_sgid_tbl(struct ocrdma_dev *dev)
178	{
179	struct net_device netdev, tmp;
180	u16 vlan_id;
181	bool is_vlan;
182
183	netdev = dev->nic_info.netdev;
184
185	ocrdma_add_default_sgid(dev);
186
187	rcu_read_lock();
188	for_each_netdev_rcu(&init_net, tmp) {
189	if (netdev == tmp \|\| vlan_dev_real_dev(tmp) == netdev) {
190	if (!netif_running(tmp) \|\| !netif_oper_up(tmp))
191	continue;
192	if (netdev != tmp) {
193	vlan_id = vlan_dev_vlan_id(tmp);
194	is_vlan = true;
195	} else {
196	is_vlan = false;
197	vlan_id = 0;
198	tmp = netdev;
199	}
200	ocrdma_add_sgid(dev, tmp->dev_addr, is_vlan, vlan_id);
201	}
202	}
203	rcu_read_unlock();
204	return 0;
205	}
206
207	static int ocrdma_inet6addr_event(struct notifier_block *notifier,
208	unsigned long event, void *ptr)
209	{
210	struct inet6_ifaddr ifa = (struct inet6_ifaddr )ptr;
211	struct net_device *event_netdev = ifa->idev->dev;
212	struct net_device *netdev = NULL;
213	struct ib_event gid_event;
214	struct ocrdma_dev *dev;
215	bool found = false;
216	bool is_vlan = false;
217	u16 vid = 0;
218
219	netdev = vlan_dev_real_dev(event_netdev);
220	if (netdev != event_netdev) {
221	is_vlan = true;
222	vid = vlan_dev_vlan_id(event_netdev);
223	}
3e4d60a8 SL	224	rcu_read_lock();
3e4d60a8 SL	225	list_for_each_entry_rcu(dev, &ocrdma_dev_list, entry) {
fe2caefc PP	226	if (dev->nic_info.netdev == netdev) {
	227	found = true;
	228	break;
	229	}
	230	}
3e4d60a8	231	rcu_read_unlock();
fe2caefc PP	232
	233	if (!found)
	234	return NOTIFY_DONE;
	235	if (!rdma_link_local_addr((struct in6_addr *)&ifa->addr))
	236	return NOTIFY_DONE;
	237
	238	mutex_lock(&dev->dev_lock);
	239	switch (event) {
	240	case NETDEV_UP:
	241	ocrdma_add_sgid(dev, netdev->dev_addr, is_vlan, vid);
	242	break;
	243	case NETDEV_DOWN:
	244	found = ocrdma_del_sgid(dev, netdev->dev_addr, is_vlan, vid);
	245	if (found) {
	246	/* found the matching entry, notify
	247	* the consumers about it
	248	*/
	249	gid_event.device = &dev->ibdev;
	250	gid_event.element.port_num = 1;
	251	gid_event.event = IB_EVENT_GID_CHANGE;
	252	ib_dispatch_event(&gid_event);
	253	}
	254	break;
	255	default:
	256	break;
	257	}
	258	mutex_unlock(&dev->dev_lock);
	259	return NOTIFY_OK;
	260	}
	261
	262	static enum rdma_link_layer ocrdma_link_layer(struct ib_device *device,
	263	u8 port_num)
	264	{
	265	return IB_LINK_LAYER_ETHERNET;
	266	}
	267
abe3afac	268	static int ocrdma_register_device(struct ocrdma_dev *dev)
fe2caefc PP	269	{
	270	strlcpy(dev->ibdev.name, "ocrdma%d", IB_DEVICE_NAME_MAX);
	271	ocrdma_get_guid(dev, (u8 *)&dev->ibdev.node_guid);
	272	memcpy(dev->ibdev.node_desc, OCRDMA_NODE_DESC,
	273	sizeof(OCRDMA_NODE_DESC));
	274	dev->ibdev.owner = THIS_MODULE;
	275	dev->ibdev.uverbs_cmd_mask =
	276	OCRDMA_UVERBS(GET_CONTEXT) \|
	277	OCRDMA_UVERBS(QUERY_DEVICE) \|
	278	OCRDMA_UVERBS(QUERY_PORT) \|
	279	OCRDMA_UVERBS(ALLOC_PD) \|
	280	OCRDMA_UVERBS(DEALLOC_PD) \|
	281	OCRDMA_UVERBS(REG_MR) \|
	282	OCRDMA_UVERBS(DEREG_MR) \|
	283	OCRDMA_UVERBS(CREATE_COMP_CHANNEL) \|
	284	OCRDMA_UVERBS(CREATE_CQ) \|
	285	OCRDMA_UVERBS(RESIZE_CQ) \|
	286	OCRDMA_UVERBS(DESTROY_CQ) \|
	287	OCRDMA_UVERBS(REQ_NOTIFY_CQ) \|
	288	OCRDMA_UVERBS(CREATE_QP) \|
	289	OCRDMA_UVERBS(MODIFY_QP) \|
	290	OCRDMA_UVERBS(QUERY_QP) \|
	291	OCRDMA_UVERBS(DESTROY_QP) \|
	292	OCRDMA_UVERBS(POLL_CQ) \|
	293	OCRDMA_UVERBS(POST_SEND) \|
	294	OCRDMA_UVERBS(POST_RECV);
	295
	296	dev->ibdev.uverbs_cmd_mask \|=
	297	OCRDMA_UVERBS(CREATE_AH) \|
	298	OCRDMA_UVERBS(MODIFY_AH) \|
	299	OCRDMA_UVERBS(QUERY_AH) \|
	300	OCRDMA_UVERBS(DESTROY_AH);
	301
	302	dev->ibdev.node_type = RDMA_NODE_IB_CA;
	303	dev->ibdev.phys_port_cnt = 1;
	304	dev->ibdev.num_comp_vectors = 1;
	305
	306	/* mandatory verbs. */
	307	dev->ibdev.query_device = ocrdma_query_device;
	308	dev->ibdev.query_port = ocrdma_query_port;
	309	dev->ibdev.modify_port = ocrdma_modify_port;
	310	dev->ibdev.query_gid = ocrdma_query_gid;
	311	dev->ibdev.get_link_layer = ocrdma_link_layer;
	312	dev->ibdev.alloc_pd = ocrdma_alloc_pd;
	313	dev->ibdev.dealloc_pd = ocrdma_dealloc_pd;
	314
	315	dev->ibdev.create_cq = ocrdma_create_cq;
	316	dev->ibdev.destroy_cq = ocrdma_destroy_cq;
	317	dev->ibdev.resize_cq = ocrdma_resize_cq;
	318
	319	dev->ibdev.create_qp = ocrdma_create_qp;
	320	dev->ibdev.modify_qp = ocrdma_modify_qp;
	321	dev->ibdev.query_qp = ocrdma_query_qp;
	322	dev->ibdev.destroy_qp = ocrdma_destroy_qp;
	323
	324	dev->ibdev.query_pkey = ocrdma_query_pkey;
	325	dev->ibdev.create_ah = ocrdma_create_ah;
	326	dev->ibdev.destroy_ah = ocrdma_destroy_ah;
	327	dev->ibdev.query_ah = ocrdma_query_ah;
	328	dev->ibdev.modify_ah = ocrdma_modify_ah;
	329
	330	dev->ibdev.poll_cq = ocrdma_poll_cq;
	331	dev->ibdev.post_send = ocrdma_post_send;
	332	dev->ibdev.post_recv = ocrdma_post_recv;
333	dev->ibdev.req_notify_cq = ocrdma_arm_cq;
334
335	dev->ibdev.get_dma_mr = ocrdma_get_dma_mr;
336	dev->ibdev.dereg_mr = ocrdma_dereg_mr;
337	dev->ibdev.reg_user_mr = ocrdma_reg_user_mr;
338
339	/* mandatory to support user space verbs consumer. */
340	dev->ibdev.alloc_ucontext = ocrdma_alloc_ucontext;
341	dev->ibdev.dealloc_ucontext = ocrdma_dealloc_ucontext;
342	dev->ibdev.mmap = ocrdma_mmap;
343	dev->ibdev.dma_device = &dev->nic_info.pdev->dev;
344
345	dev->ibdev.process_mad = ocrdma_process_mad;
346
347	if (dev->nic_info.dev_family == OCRDMA_GEN2_FAMILY) {
348	dev->ibdev.uverbs_cmd_mask \|=
349	OCRDMA_UVERBS(CREATE_SRQ) \|
350	OCRDMA_UVERBS(MODIFY_SRQ) \|
351	OCRDMA_UVERBS(QUERY_SRQ) \|
352	OCRDMA_UVERBS(DESTROY_SRQ) \|
353	OCRDMA_UVERBS(POST_SRQ_RECV);
354
355	dev->ibdev.create_srq = ocrdma_create_srq;
356	dev->ibdev.modify_srq = ocrdma_modify_srq;
357	dev->ibdev.query_srq = ocrdma_query_srq;
358	dev->ibdev.destroy_srq = ocrdma_destroy_srq;
359	dev->ibdev.post_srq_recv = ocrdma_post_srq_recv;
360	}
361	return ib_register_device(&dev->ibdev, NULL);
362	}
363
364	static int ocrdma_alloc_resources(struct ocrdma_dev *dev)
365	{
366	mutex_init(&dev->dev_lock);
367	dev->sgid_tbl = kzalloc(sizeof(union ib_gid) *
368	OCRDMA_MAX_SGID, GFP_KERNEL);
369	if (!dev->sgid_tbl)
370	goto alloc_err;
371	spin_lock_init(&dev->sgid_lock);
372
373	dev->cq_tbl = kzalloc(sizeof(struct ocrdma_cq )
374	OCRDMA_MAX_CQ, GFP_KERNEL);
375	if (!dev->cq_tbl)
376	goto alloc_err;
377
378	if (dev->attr.max_qp) {
379	dev->qp_tbl = kzalloc(sizeof(struct ocrdma_qp )
380	OCRDMA_MAX_QP, GFP_KERNEL);
381	if (!dev->qp_tbl)
382	goto alloc_err;
383	}
384	spin_lock_init(&dev->av_tbl.lock);
385	spin_lock_init(&dev->flush_q_lock);
386	return 0;
387	alloc_err:
388	ocrdma_err("%s(%d) error.\n", __func__, dev->id);
389	return -ENOMEM;
390	}
391
392	static void ocrdma_free_resources(struct ocrdma_dev *dev)
393	{
394	kfree(dev->qp_tbl);
395	kfree(dev->cq_tbl);
396	kfree(dev->sgid_tbl);
397	}
398
399	static struct ocrdma_dev ocrdma_add(struct be_dev_info dev_info)
400	{
401	int status = 0;
402	struct ocrdma_dev *dev;
403
404	dev = (struct ocrdma_dev *)ib_alloc_device(sizeof(struct ocrdma_dev));
405	if (!dev) {
406	ocrdma_err("Unable to allocate ib device\n");
407	return NULL;
408	}
409	dev->mbx_cmd = kzalloc(sizeof(struct ocrdma_mqe_emb_cmd), GFP_KERNEL);
410	if (!dev->mbx_cmd)
411	goto idr_err;
412
413	memcpy(&dev->nic_info, dev_info, sizeof(*dev_info));
414	dev->id = ocrdma_get_instance();
415	if (dev->id < 0)
416	goto idr_err;
417
418	status = ocrdma_init_hw(dev);
419	if (status)
420	goto init_err;
421
422	status = ocrdma_alloc_resources(dev);
423	if (status)
424	goto alloc_err;
425
426	status = ocrdma_build_sgid_tbl(dev);
427	if (status)
428	goto alloc_err;
429
430	status = ocrdma_register_device(dev);
431	if (status)
432	goto alloc_err;
433
3e4d60a8 SL	434	spin_lock(&ocrdma_devlist_lock);
	435	list_add_tail_rcu(&dev->entry, &ocrdma_dev_list);
	436	spin_unlock(&ocrdma_devlist_lock);
fe2caefc PP	437	return dev;
	438
	439	alloc_err:
	440	ocrdma_free_resources(dev);
	441	ocrdma_cleanup_hw(dev);
	442	init_err:
	443	idr_remove(&ocrdma_dev_id, dev->id);
	444	idr_err:
	445	kfree(dev->mbx_cmd);
	446	ib_dealloc_device(&dev->ibdev);
	447	ocrdma_err("%s() leaving. ret=%d\n", __func__, status);
	448	return NULL;
	449	}
	450
3e4d60a8	451	static void ocrdma_remove_free(struct rcu_head *rcu)
fe2caefc	452	{
3e4d60a8	453	struct ocrdma_dev *dev = container_of(rcu, struct ocrdma_dev, rcu);
fe2caefc PP	454
	455	ocrdma_free_resources(dev);
	456	ocrdma_cleanup_hw(dev);
	457
	458	idr_remove(&ocrdma_dev_id, dev->id);
	459	kfree(dev->mbx_cmd);
	460	ib_dealloc_device(&dev->ibdev);
	461	}
	462
3e4d60a8 SL	463	static void ocrdma_remove(struct ocrdma_dev *dev)
	464	{
	465	/* first unregister with stack to stop all the active traffic
	466	* of the registered clients.
	467	*/
	468	ib_unregister_device(&dev->ibdev);
	469
	470	spin_lock(&ocrdma_devlist_lock);
	471	list_del_rcu(&dev->entry);
	472	spin_unlock(&ocrdma_devlist_lock);
	473	call_rcu(&dev->rcu, ocrdma_remove_free);
	474	}
	475
fe2caefc PP	476	static int ocrdma_open(struct ocrdma_dev *dev)
	477	{
	478	struct ib_event port_event;
	479
	480	port_event.event = IB_EVENT_PORT_ACTIVE;
	481	port_event.element.port_num = 1;
	482	port_event.device = &dev->ibdev;
	483	ib_dispatch_event(&port_event);
	484	return 0;
	485	}
	486
	487	static int ocrdma_close(struct ocrdma_dev *dev)
	488	{
	489	int i;
	490	struct ocrdma_qp qp, *cur_qp;
	491	struct ib_event err_event;
	492	struct ib_qp_attr attrs;
	493	int attr_mask = IB_QP_STATE;
	494
	495	attrs.qp_state = IB_QPS_ERR;
	496	mutex_lock(&dev->dev_lock);
	497	if (dev->qp_tbl) {
	498	cur_qp = dev->qp_tbl;
	499	for (i = 0; i < OCRDMA_MAX_QP; i++) {
	500	qp = cur_qp[i];
	501	if (qp) {
	502	/* change the QP state to ERROR */
	503	_ocrdma_modify_qp(&qp->ibqp, &attrs, attr_mask);
	504
	505	err_event.event = IB_EVENT_QP_FATAL;
	506	err_event.element.qp = &qp->ibqp;
	507	err_event.device = &dev->ibdev;
	508	ib_dispatch_event(&err_event);
	509	}
	510	}
	511	}
	512	mutex_unlock(&dev->dev_lock);
	513
	514	err_event.event = IB_EVENT_PORT_ERR;
	515	err_event.element.port_num = 1;
	516	err_event.device = &dev->ibdev;
	517	ib_dispatch_event(&err_event);
	518	return 0;
	519	}
	520
	521	/* event handling via NIC driver ensures that all the NIC specific
	522	* initialization done before RoCE driver notifies
	523	* event to stack.
	524	*/
	525	static void ocrdma_event_handler(struct ocrdma_dev *dev, u32 event)
	526	{
	527	switch (event) {
	528	case BE_DEV_UP:
	529	ocrdma_open(dev);
	530	break;
	531	case BE_DEV_DOWN:
	532	ocrdma_close(dev);
	533	break;
	534	};
	535	}
	536
abe3afac	537	static struct ocrdma_driver ocrdma_drv = {
fe2caefc PP	538	.name = "ocrdma_driver",
	539	.add = ocrdma_add,
	540	.remove = ocrdma_remove,
	541	.state_change_handler = ocrdma_event_handler,
	542	};
	543
	544	static int __init ocrdma_init_module(void)
	545	{
	546	int status;
	547
	548	status = register_inet6addr_notifier(&ocrdma_inet6addr_notifier);
	549	if (status)
	550	return status;
	551	status = be_roce_register_driver(&ocrdma_drv);
	552	if (status)
	553	unregister_inet6addr_notifier(&ocrdma_inet6addr_notifier);
	554	return status;
	555	}
	556
	557	static void __exit ocrdma_exit_module(void)
	558	{
	559	be_roce_unregister_driver(&ocrdma_drv);
	560	unregister_inet6addr_notifier(&ocrdma_inet6addr_notifier);
	561	}
	562
	563	module_init(ocrdma_init_module);
	564	module_exit(ocrdma_exit_module);