[linux-block.git] / net / smc / smc_ib.c

/*
 *  Shared Memory Communications over RDMA (SMC-R) and RoCE
 *
 *  IB infrastructure:
 *  Establish SMC-R as an Infiniband Client to be notified about added and
 *  removed IB devices of type RDMA.
 *  Determine device and port characteristics for these IB devices.
 *
 *  Copyright IBM Corp. 2016
 *
 *  Author(s):  Ursula Braun <ubraun@linux.vnet.ibm.com>
 */

#include <linux/random.h>
#include <linux/workqueue.h>
#include <rdma/ib_verbs.h>

#include "smc_pnet.h"
#include "smc_ib.h"
#include "smc_core.h"
#include "smc_wr.h"
#include "smc.h"

#define SMC_QP_MIN_RNR_TIMER		5
#define SMC_QP_TIMEOUT			15 /* 4096 * 2 ** timeout usec */
#define SMC_QP_RETRY_CNT			7 /* 7: infinite */
#define SMC_QP_RNR_RETRY			7 /* 7: infinite */

struct smc_ib_devices smc_ib_devices = {	/* smc-registered ib devices */
	.lock = __SPIN_LOCK_UNLOCKED(smc_ib_devices.lock),
	.list = LIST_HEAD_INIT(smc_ib_devices.list),
};

#define SMC_LOCAL_SYSTEMID_RESET	"%%%%%%%"

u8 local_systemid[SMC_SYSTEMID_LEN] = SMC_LOCAL_SYSTEMID_RESET;	/* unique system
								 * identifier
								 */

static int smc_ib_modify_qp_init(struct smc_link *lnk)
{
	struct ib_qp_attr qp_attr;

	memset(&qp_attr, 0, sizeof(qp_attr));
	qp_attr.qp_state = IB_QPS_INIT;
	qp_attr.pkey_index = 0;
	qp_attr.port_num = lnk->ibport;
	qp_attr.qp_access_flags = IB_ACCESS_LOCAL_WRITE
				| IB_ACCESS_REMOTE_WRITE;
	return ib_modify_qp(lnk->roce_qp, &qp_attr,
			    IB_QP_STATE | IB_QP_PKEY_INDEX |
			    IB_QP_ACCESS_FLAGS | IB_QP_PORT);
}

static int smc_ib_modify_qp_rtr(struct smc_link *lnk)
{
	enum ib_qp_attr_mask qp_attr_mask =
		IB_QP_STATE | IB_QP_AV | IB_QP_PATH_MTU | IB_QP_DEST_QPN |
		IB_QP_RQ_PSN | IB_QP_MAX_DEST_RD_ATOMIC | IB_QP_MIN_RNR_TIMER;
	struct ib_qp_attr qp_attr;

	memset(&qp_attr, 0, sizeof(qp_attr));
	qp_attr.qp_state = IB_QPS_RTR;
	qp_attr.path_mtu = min(lnk->path_mtu, lnk->peer_mtu);
	qp_attr.ah_attr.type = RDMA_AH_ATTR_TYPE_ROCE;
	rdma_ah_set_port_num(&qp_attr.ah_attr, lnk->ibport);
	rdma_ah_set_grh(&qp_attr.ah_attr, NULL, 0, 0, 1, 0);
	rdma_ah_set_dgid_raw(&qp_attr.ah_attr, lnk->peer_gid);
	memcpy(&qp_attr.ah_attr.roce.dmac, lnk->peer_mac,
	       sizeof(lnk->peer_mac));
	qp_attr.dest_qp_num = lnk->peer_qpn;
	qp_attr.rq_psn = lnk->peer_psn; /* starting receive packet seq # */
	qp_attr.max_dest_rd_atomic = 1; /* max # of resources for incoming
					 * requests
					 */
	qp_attr.min_rnr_timer = SMC_QP_MIN_RNR_TIMER;

	return ib_modify_qp(lnk->roce_qp, &qp_attr, qp_attr_mask);
}

int smc_ib_modify_qp_rts(struct smc_link *lnk)
{
	struct ib_qp_attr qp_attr;

	memset(&qp_attr, 0, sizeof(qp_attr));
	qp_attr.qp_state = IB_QPS_RTS;
	qp_attr.timeout = SMC_QP_TIMEOUT;	/* local ack timeout */
	qp_attr.retry_cnt = SMC_QP_RETRY_CNT;	/* retry count */
	qp_attr.rnr_retry = SMC_QP_RNR_RETRY;	/* RNR retries, 7=infinite */
	qp_attr.sq_psn = lnk->psn_initial;	/* starting send packet seq # */
	qp_attr.max_rd_atomic = 1;	/* # of outstanding RDMA reads and
					 * atomic ops allowed
					 */
	return ib_modify_qp(lnk->roce_qp, &qp_attr,
			    IB_QP_STATE | IB_QP_TIMEOUT | IB_QP_RETRY_CNT |
			    IB_QP_SQ_PSN | IB_QP_RNR_RETRY |
			    IB_QP_MAX_QP_RD_ATOMIC);
}

int smc_ib_modify_qp_reset(struct smc_link *lnk)
{
	struct ib_qp_attr qp_attr;

	memset(&qp_attr, 0, sizeof(qp_attr));
	qp_attr.qp_state = IB_QPS_RESET;
	return ib_modify_qp(lnk->roce_qp, &qp_attr, IB_QP_STATE);
}

int smc_ib_ready_link(struct smc_link *lnk)
{
	struct smc_link_group *lgr =
		container_of(lnk, struct smc_link_group, lnk[0]);
	int rc = 0;

	rc = smc_ib_modify_qp_init(lnk);
	if (rc)
		goto out;

	rc = smc_ib_modify_qp_rtr(lnk);
	if (rc)
		goto out;
	smc_wr_remember_qp_attr(lnk);
	rc = ib_req_notify_cq(lnk->smcibdev->roce_cq_recv,
			      IB_CQ_SOLICITED_MASK);
	if (rc)
		goto out;
	rc = smc_wr_rx_post_init(lnk);
	if (rc)
		goto out;
	smc_wr_remember_qp_attr(lnk);

	if (lgr->role == SMC_SERV) {
		rc = smc_ib_modify_qp_rts(lnk);
		if (rc)
			goto out;
		smc_wr_remember_qp_attr(lnk);
	}
out:
	return rc;
}

/* process context wrapper for might_sleep smc_ib_remember_port_attr */
static void smc_ib_port_event_work(struct work_struct *work)
{
	struct smc_ib_device *smcibdev = container_of(
		work, struct smc_ib_device, port_event_work);
	u8 port_idx;

	for_each_set_bit(port_idx, &smcibdev->port_event_mask, SMC_MAX_PORTS) {
		smc_ib_remember_port_attr(smcibdev, port_idx + 1);
		clear_bit(port_idx, &smcibdev->port_event_mask);
	}
}

/* can be called in IRQ context */
static void smc_ib_global_event_handler(struct ib_event_handler *handler,
					struct ib_event *ibevent)
{
	struct smc_ib_device *smcibdev;
	u8 port_idx;

	smcibdev = container_of(handler, struct smc_ib_device, event_handler);

	switch (ibevent->event) {
	case IB_EVENT_PORT_ERR:
		port_idx = ibevent->element.port_num - 1;
		set_bit(port_idx, &smcibdev->port_event_mask);
		schedule_work(&smcibdev->port_event_work);
		/* fall through */
	case IB_EVENT_DEVICE_FATAL:
		/* tbd in follow-on patch:
		 * abnormal close of corresponding connections
		 */
		break;
	case IB_EVENT_PORT_ACTIVE:
		port_idx = ibevent->element.port_num - 1;
		set_bit(port_idx, &smcibdev->port_event_mask);
		schedule_work(&smcibdev->port_event_work);
		break;
	default:
		break;
	}
}

void smc_ib_dealloc_protection_domain(struct smc_link *lnk)
{
	ib_dealloc_pd(lnk->roce_pd);
	lnk->roce_pd = NULL;
}

int smc_ib_create_protection_domain(struct smc_link *lnk)
{
	int rc;

	lnk->roce_pd = ib_alloc_pd(lnk->smcibdev->ibdev, 0);
	rc = PTR_ERR_OR_ZERO(lnk->roce_pd);
	if (IS_ERR(lnk->roce_pd))
		lnk->roce_pd = NULL;
	return rc;
}

static void smc_ib_qp_event_handler(struct ib_event *ibevent, void *priv)
{
	switch (ibevent->event) {
	case IB_EVENT_DEVICE_FATAL:
	case IB_EVENT_GID_CHANGE:
	case IB_EVENT_PORT_ERR:
	case IB_EVENT_QP_ACCESS_ERR:
		/* tbd in follow-on patch:
		 * abnormal close of corresponding connections
		 */
		break;
	default:
		break;
	}
}

void smc_ib_destroy_queue_pair(struct smc_link *lnk)
{
	ib_destroy_qp(lnk->roce_qp);
	lnk->roce_qp = NULL;
}

/* create a queue pair within the protection domain for a link */
int smc_ib_create_queue_pair(struct smc_link *lnk)
{
	struct ib_qp_init_attr qp_attr = {
		.event_handler = smc_ib_qp_event_handler,
		.qp_context = lnk,
		.send_cq = lnk->smcibdev->roce_cq_send,
		.recv_cq = lnk->smcibdev->roce_cq_recv,
		.srq = NULL,
		.cap = {
				/* include unsolicited rdma_writes as well,
				 * there are max. 2 RDMA_WRITE per 1 WR_SEND
				 */
			.max_send_wr = SMC_WR_BUF_CNT * 3,
			.max_recv_wr = SMC_WR_BUF_CNT * 3,
			.max_send_sge = SMC_IB_MAX_SEND_SGE,
			.max_recv_sge = 1,
		},
		.sq_sig_type = IB_SIGNAL_REQ_WR,
		.qp_type = IB_QPT_RC,
	};
	int rc;

	lnk->roce_qp = ib_create_qp(lnk->roce_pd, &qp_attr);
	rc = PTR_ERR_OR_ZERO(lnk->roce_qp);
	if (IS_ERR(lnk->roce_qp))
		lnk->roce_qp = NULL;
	else
		smc_wr_remember_qp_attr(lnk);
	return rc;
}

void smc_ib_put_memory_region(struct ib_mr *mr)
{
	ib_dereg_mr(mr);
}

static int smc_ib_map_mr_sg(struct smc_buf_desc *buf_slot)
{
	unsigned int offset = 0;
	int sg_num;

	/* map the largest prefix of a dma mapped SG list */
	sg_num = ib_map_mr_sg(buf_slot->mr_rx[SMC_SINGLE_LINK],
			      buf_slot->sgt[SMC_SINGLE_LINK].sgl,
			      buf_slot->sgt[SMC_SINGLE_LINK].orig_nents,
			      &offset, PAGE_SIZE);

	return sg_num;
}

/* Allocate a memory region and map the dma mapped SG list of buf_slot */
int smc_ib_get_memory_region(struct ib_pd *pd, int access_flags,
			     struct smc_buf_desc *buf_slot)
{
	if (buf_slot->mr_rx[SMC_SINGLE_LINK])
		return 0; /* already done */

	buf_slot->mr_rx[SMC_SINGLE_LINK] =
		ib_alloc_mr(pd, IB_MR_TYPE_MEM_REG, 1 << buf_slot->order);
	if (IS_ERR(buf_slot->mr_rx[SMC_SINGLE_LINK])) {
		int rc;

		rc = PTR_ERR(buf_slot->mr_rx[SMC_SINGLE_LINK]);
		buf_slot->mr_rx[SMC_SINGLE_LINK] = NULL;
		return rc;
	}

	if (smc_ib_map_mr_sg(buf_slot) != 1)
		return -EINVAL;

	return 0;
}

/* Map a new TX or RX buffer SG-table to DMA */
int smc_ib_buf_map_sg(struct smc_ib_device *smcibdev,
		      struct smc_buf_desc *buf_slot,
		      enum dma_data_direction data_direction)
{
	int mapped_nents;

	mapped_nents = ib_dma_map_sg(smcibdev->ibdev,
				     buf_slot->sgt[SMC_SINGLE_LINK].sgl,
				     buf_slot->sgt[SMC_SINGLE_LINK].orig_nents,
				     data_direction);
	if (!mapped_nents)
		return -ENOMEM;

	return mapped_nents;
}

void smc_ib_buf_unmap_sg(struct smc_ib_device *smcibdev,
			 struct smc_buf_desc *buf_slot,
			 enum dma_data_direction data_direction)
{
	if (!buf_slot->sgt[SMC_SINGLE_LINK].sgl->dma_address)
		return; /* already unmapped */

	ib_dma_unmap_sg(smcibdev->ibdev,
			buf_slot->sgt[SMC_SINGLE_LINK].sgl,
			buf_slot->sgt[SMC_SINGLE_LINK].orig_nents,
			data_direction);
	buf_slot->sgt[SMC_SINGLE_LINK].sgl->dma_address = 0;
}

static int smc_ib_fill_gid_and_mac(struct smc_ib_device *smcibdev, u8 ibport)
{
	struct net_device *ndev;
	int rc;

	rc = ib_query_gid(smcibdev->ibdev, ibport, 0,
			  &smcibdev->gid[ibport - 1], NULL);
	/* the SMC protocol requires specification of the roce MAC address;
	 * if net_device cannot be determined, it can be derived from gid 0
	 */
	ndev = smcibdev->ibdev->get_netdev(smcibdev->ibdev, ibport);
	if (ndev) {
		memcpy(&smcibdev->mac, ndev->dev_addr, ETH_ALEN);
	} else if (!rc) {
		memcpy(&smcibdev->mac[ibport - 1][0],
		       &smcibdev->gid[ibport - 1].raw[8], 3);
		memcpy(&smcibdev->mac[ibport - 1][3],
		       &smcibdev->gid[ibport - 1].raw[13], 3);
		smcibdev->mac[ibport - 1][0] &= ~0x02;
	}
	return rc;
}

/* Create an identifier unique for this instance of SMC-R.
 * The MAC-address of the first active registered IB device
 * plus a random 2-byte number is used to create this identifier.
 * This name is delivered to the peer during connection initialization.
 */
static inline void smc_ib_define_local_systemid(struct smc_ib_device *smcibdev,
						u8 ibport)
{
	memcpy(&local_systemid[2], &smcibdev->mac[ibport - 1],
	       sizeof(smcibdev->mac[ibport - 1]));
	get_random_bytes(&local_systemid[0], 2);
}

bool smc_ib_port_active(struct smc_ib_device *smcibdev, u8 ibport)
{
	return smcibdev->pattr[ibport - 1].state == IB_PORT_ACTIVE;
}

int smc_ib_remember_port_attr(struct smc_ib_device *smcibdev, u8 ibport)
{
	int rc;

	memset(&smcibdev->pattr[ibport - 1], 0,
	       sizeof(smcibdev->pattr[ibport - 1]));
	rc = ib_query_port(smcibdev->ibdev, ibport,
			   &smcibdev->pattr[ibport - 1]);
	if (rc)
		goto out;
	rc = smc_ib_fill_gid_and_mac(smcibdev, ibport);
	if (rc)
		goto out;
	if (!strncmp(local_systemid, SMC_LOCAL_SYSTEMID_RESET,
		     sizeof(local_systemid)) &&
	    smc_ib_port_active(smcibdev, ibport))
		/* create unique system identifier */
		smc_ib_define_local_systemid(smcibdev, ibport);
out:
	return rc;
}

long smc_ib_setup_per_ibdev(struct smc_ib_device *smcibdev)
{
	struct ib_cq_init_attr cqattr =	{
		.cqe = SMC_WR_MAX_CQE, .comp_vector = 0 };
	long rc;

	smcibdev->roce_cq_send = ib_create_cq(smcibdev->ibdev,
					      smc_wr_tx_cq_handler, NULL,
					      smcibdev, &cqattr);
	rc = PTR_ERR_OR_ZERO(smcibdev->roce_cq_send);
	if (IS_ERR(smcibdev->roce_cq_send)) {
		smcibdev->roce_cq_send = NULL;
		return rc;
	}
	smcibdev->roce_cq_recv = ib_create_cq(smcibdev->ibdev,
					      smc_wr_rx_cq_handler, NULL,
					      smcibdev, &cqattr);
	rc = PTR_ERR_OR_ZERO(smcibdev->roce_cq_recv);
	if (IS_ERR(smcibdev->roce_cq_recv)) {
		smcibdev->roce_cq_recv = NULL;
		goto err;
	}
	INIT_IB_EVENT_HANDLER(&smcibdev->event_handler, smcibdev->ibdev,
			      smc_ib_global_event_handler);
	ib_register_event_handler(&smcibdev->event_handler);
	smc_wr_add_dev(smcibdev);
	smcibdev->initialized = 1;
	return rc;

err:
	ib_destroy_cq(smcibdev->roce_cq_send);
	return rc;
}

static void smc_ib_cleanup_per_ibdev(struct smc_ib_device *smcibdev)
{
	if (!smcibdev->initialized)
		return;
	smc_wr_remove_dev(smcibdev);
	ib_unregister_event_handler(&smcibdev->event_handler);
	ib_destroy_cq(smcibdev->roce_cq_recv);
	ib_destroy_cq(smcibdev->roce_cq_send);
}

static struct ib_client smc_ib_client;

/* callback function for ib_register_client() */
static void smc_ib_add_dev(struct ib_device *ibdev)
{
	struct smc_ib_device *smcibdev;

	if (ibdev->node_type != RDMA_NODE_IB_CA)
		return;

	smcibdev = kzalloc(sizeof(*smcibdev), GFP_KERNEL);
	if (!smcibdev)
		return;

	smcibdev->ibdev = ibdev;
	INIT_WORK(&smcibdev->port_event_work, smc_ib_port_event_work);

	spin_lock(&smc_ib_devices.lock);
	list_add_tail(&smcibdev->list, &smc_ib_devices.list);
	spin_unlock(&smc_ib_devices.lock);
	ib_set_client_data(ibdev, &smc_ib_client, smcibdev);
}

/* callback function for ib_register_client() */
static void smc_ib_remove_dev(struct ib_device *ibdev, void *client_data)
{
	struct smc_ib_device *smcibdev;

	smcibdev = ib_get_client_data(ibdev, &smc_ib_client);
	ib_set_client_data(ibdev, &smc_ib_client, NULL);
	spin_lock(&smc_ib_devices.lock);
	list_del_init(&smcibdev->list); /* remove from smc_ib_devices */
	spin_unlock(&smc_ib_devices.lock);
	smc_pnet_remove_by_ibdev(smcibdev);
	smc_ib_cleanup_per_ibdev(smcibdev);
	kfree(smcibdev);
}

static struct ib_client smc_ib_client = {
	.name	= "smc_ib",
	.add	= smc_ib_add_dev,
	.remove = smc_ib_remove_dev,
};

int __init smc_ib_register_client(void)
{
	return ib_register_client(&smc_ib_client);
}

void smc_ib_unregister_client(void)
{
	ib_unregister_client(&smc_ib_client);
}
Commit	Line	Data
a4cf0443 UB	1	/*
	2	* Shared Memory Communications over RDMA (SMC-R) and RoCE
	3	*
	4	* IB infrastructure:
	5	* Establish SMC-R as an Infiniband Client to be notified about added and
	6	* removed IB devices of type RDMA.
	7	* Determine device and port characteristics for these IB devices.
	8	*
	9	* Copyright IBM Corp. 2016
	10	*
	11	* Author(s): Ursula Braun <ubraun@linux.vnet.ibm.com>
	12	*/
	13
	14	#include <linux/random.h>
bd4ad577	15	#include <linux/workqueue.h>
a4cf0443 UB	16	#include <rdma/ib_verbs.h>
a4cf0443 UB	17
6812baab	18	#include "smc_pnet.h"
a4cf0443	19	#include "smc_ib.h"
cd6851f3	20	#include "smc_core.h"
f38ba179	21	#include "smc_wr.h"
a4cf0443 UB	22	#include "smc.h"
a4cf0443 UB	23
bd4ad577 UB	24	#define SMC_QP_MIN_RNR_TIMER 5
	25	#define SMC_QP_TIMEOUT 15 /* 4096 * 2 ** timeout usec */
	26	#define SMC_QP_RETRY_CNT 7 /* 7: infinite */
	27	#define SMC_QP_RNR_RETRY 7 /* 7: infinite */
	28
a4cf0443 UB	29	struct smc_ib_devices smc_ib_devices = { /* smc-registered ib devices */
	30	.lock = __SPIN_LOCK_UNLOCKED(smc_ib_devices.lock),
	31	.list = LIST_HEAD_INIT(smc_ib_devices.list),
	32	};
	33
	34	#define SMC_LOCAL_SYSTEMID_RESET "%%%%%%%"
	35
	36	u8 local_systemid[SMC_SYSTEMID_LEN] = SMC_LOCAL_SYSTEMID_RESET; /* unique system
	37	* identifier
	38	*/
	39
bd4ad577 UB	40	static int smc_ib_modify_qp_init(struct smc_link *lnk)
	41	{
	42	struct ib_qp_attr qp_attr;
	43
	44	memset(&qp_attr, 0, sizeof(qp_attr));
	45	qp_attr.qp_state = IB_QPS_INIT;
	46	qp_attr.pkey_index = 0;
	47	qp_attr.port_num = lnk->ibport;
	48	qp_attr.qp_access_flags = IB_ACCESS_LOCAL_WRITE
	49	\| IB_ACCESS_REMOTE_WRITE;
	50	return ib_modify_qp(lnk->roce_qp, &qp_attr,
	51	IB_QP_STATE \| IB_QP_PKEY_INDEX \|
	52	IB_QP_ACCESS_FLAGS \| IB_QP_PORT);
	53	}
	54
	55	static int smc_ib_modify_qp_rtr(struct smc_link *lnk)
	56	{
	57	enum ib_qp_attr_mask qp_attr_mask =
	58	IB_QP_STATE \| IB_QP_AV \| IB_QP_PATH_MTU \| IB_QP_DEST_QPN \|
	59	IB_QP_RQ_PSN \| IB_QP_MAX_DEST_RD_ATOMIC \| IB_QP_MIN_RNR_TIMER;
	60	struct ib_qp_attr qp_attr;
	61
	62	memset(&qp_attr, 0, sizeof(qp_attr));
	63	qp_attr.qp_state = IB_QPS_RTR;
	64	qp_attr.path_mtu = min(lnk->path_mtu, lnk->peer_mtu);
44c58487	65	qp_attr.ah_attr.type = RDMA_AH_ATTR_TYPE_ROCE;
d8966fcd DC	66	rdma_ah_set_port_num(&qp_attr.ah_attr, lnk->ibport);
	67	rdma_ah_set_grh(&qp_attr.ah_attr, NULL, 0, 0, 1, 0);
	68	rdma_ah_set_dgid_raw(&qp_attr.ah_attr, lnk->peer_gid);
44c58487	69	memcpy(&qp_attr.ah_attr.roce.dmac, lnk->peer_mac,
bd4ad577 UB	70	sizeof(lnk->peer_mac));
	71	qp_attr.dest_qp_num = lnk->peer_qpn;
	72	qp_attr.rq_psn = lnk->peer_psn; /* starting receive packet seq # */
	73	qp_attr.max_dest_rd_atomic = 1; /* max # of resources for incoming
	74	* requests
	75	*/
	76	qp_attr.min_rnr_timer = SMC_QP_MIN_RNR_TIMER;
	77
	78	return ib_modify_qp(lnk->roce_qp, &qp_attr, qp_attr_mask);
	79	}
	80
	81	int smc_ib_modify_qp_rts(struct smc_link *lnk)
	82	{
	83	struct ib_qp_attr qp_attr;
	84
	85	memset(&qp_attr, 0, sizeof(qp_attr));
	86	qp_attr.qp_state = IB_QPS_RTS;
	87	qp_attr.timeout = SMC_QP_TIMEOUT; /* local ack timeout */
	88	qp_attr.retry_cnt = SMC_QP_RETRY_CNT; /* retry count */
	89	qp_attr.rnr_retry = SMC_QP_RNR_RETRY; /* RNR retries, 7=infinite */
	90	qp_attr.sq_psn = lnk->psn_initial; /* starting send packet seq # */
	91	qp_attr.max_rd_atomic = 1; /* # of outstanding RDMA reads and
	92	* atomic ops allowed
	93	*/
	94	return ib_modify_qp(lnk->roce_qp, &qp_attr,
	95	IB_QP_STATE \| IB_QP_TIMEOUT \| IB_QP_RETRY_CNT \|
	96	IB_QP_SQ_PSN \| IB_QP_RNR_RETRY \|
	97	IB_QP_MAX_QP_RD_ATOMIC);
	98	}
	99
	100	int smc_ib_modify_qp_reset(struct smc_link *lnk)
	101	{
	102	struct ib_qp_attr qp_attr;
	103
	104	memset(&qp_attr, 0, sizeof(qp_attr));
	105	qp_attr.qp_state = IB_QPS_RESET;
	106	return ib_modify_qp(lnk->roce_qp, &qp_attr, IB_QP_STATE);
	107	}
	108
	109	int smc_ib_ready_link(struct smc_link *lnk)
	110	{
	111	struct smc_link_group *lgr =
	112	container_of(lnk, struct smc_link_group, lnk[0]);
	113	int rc = 0;
	114
	115	rc = smc_ib_modify_qp_init(lnk);
	116	if (rc)
	117	goto out;
	118
	119	rc = smc_ib_modify_qp_rtr(lnk);
	120	if (rc)
	121	goto out;
	122	smc_wr_remember_qp_attr(lnk);
	123	rc = ib_req_notify_cq(lnk->smcibdev->roce_cq_recv,
	124	IB_CQ_SOLICITED_MASK);
	125	if (rc)
	126	goto out;
	127	rc = smc_wr_rx_post_init(lnk);
	128	if (rc)
	129	goto out;
	130	smc_wr_remember_qp_attr(lnk);
	131
	132	if (lgr->role == SMC_SERV) {
	133	rc = smc_ib_modify_qp_rts(lnk);
134	if (rc)
135	goto out;
136	smc_wr_remember_qp_attr(lnk);
137	}
138	out:
139	return rc;
140	}
141
142	/* process context wrapper for might_sleep smc_ib_remember_port_attr */
143	static void smc_ib_port_event_work(struct work_struct *work)
144	{
145	struct smc_ib_device *smcibdev = container_of(
146	work, struct smc_ib_device, port_event_work);
147	u8 port_idx;
148
149	for_each_set_bit(port_idx, &smcibdev->port_event_mask, SMC_MAX_PORTS) {
150	smc_ib_remember_port_attr(smcibdev, port_idx + 1);
151	clear_bit(port_idx, &smcibdev->port_event_mask);
152	}
153	}
154
155	/* can be called in IRQ context */
156	static void smc_ib_global_event_handler(struct ib_event_handler *handler,
157	struct ib_event *ibevent)
158	{
159	struct smc_ib_device *smcibdev;
160	u8 port_idx;
161
162	smcibdev = container_of(handler, struct smc_ib_device, event_handler);
bd4ad577 UB	163
	164	switch (ibevent->event) {
	165	case IB_EVENT_PORT_ERR:
	166	port_idx = ibevent->element.port_num - 1;
	167	set_bit(port_idx, &smcibdev->port_event_mask);
	168	schedule_work(&smcibdev->port_event_work);
	169	/* fall through */
	170	case IB_EVENT_DEVICE_FATAL:
	171	/* tbd in follow-on patch:
	172	* abnormal close of corresponding connections
	173	*/
	174	break;
	175	case IB_EVENT_PORT_ACTIVE:
	176	port_idx = ibevent->element.port_num - 1;
	177	set_bit(port_idx, &smcibdev->port_event_mask);
	178	schedule_work(&smcibdev->port_event_work);
	179	break;
	180	default:
	181	break;
	182	}
	183	}
	184
f38ba179 UB	185	void smc_ib_dealloc_protection_domain(struct smc_link *lnk)
	186	{
	187	ib_dealloc_pd(lnk->roce_pd);
	188	lnk->roce_pd = NULL;
	189	}
	190
	191	int smc_ib_create_protection_domain(struct smc_link *lnk)
	192	{
	193	int rc;
	194
897e1c24	195	lnk->roce_pd = ib_alloc_pd(lnk->smcibdev->ibdev, 0);
f38ba179 UB	196	rc = PTR_ERR_OR_ZERO(lnk->roce_pd);
	197	if (IS_ERR(lnk->roce_pd))
	198	lnk->roce_pd = NULL;
	199	return rc;
	200	}
	201
	202	static void smc_ib_qp_event_handler(struct ib_event ibevent, void priv)
	203	{
	204	switch (ibevent->event) {
	205	case IB_EVENT_DEVICE_FATAL:
	206	case IB_EVENT_GID_CHANGE:
	207	case IB_EVENT_PORT_ERR:
	208	case IB_EVENT_QP_ACCESS_ERR:
	209	/* tbd in follow-on patch:
	210	* abnormal close of corresponding connections
	211	*/
	212	break;
	213	default:
	214	break;
	215	}
	216	}
	217
	218	void smc_ib_destroy_queue_pair(struct smc_link *lnk)
	219	{
	220	ib_destroy_qp(lnk->roce_qp);
	221	lnk->roce_qp = NULL;
	222	}
	223
	224	/* create a queue pair within the protection domain for a link */
	225	int smc_ib_create_queue_pair(struct smc_link *lnk)
	226	{
	227	struct ib_qp_init_attr qp_attr = {
	228	.event_handler = smc_ib_qp_event_handler,
	229	.qp_context = lnk,
	230	.send_cq = lnk->smcibdev->roce_cq_send,
	231	.recv_cq = lnk->smcibdev->roce_cq_recv,
	232	.srq = NULL,
	233	.cap = {
f38ba179 UB	234	/* include unsolicited rdma_writes as well,
	235	* there are max. 2 RDMA_WRITE per 1 WR_SEND
	236	*/
652a1e41	237	.max_send_wr = SMC_WR_BUF_CNT * 3,
f38ba179 UB	238	.max_recv_wr = SMC_WR_BUF_CNT * 3,
	239	.max_send_sge = SMC_IB_MAX_SEND_SGE,
	240	.max_recv_sge = 1,
f38ba179 UB	241	},
	242	.sq_sig_type = IB_SIGNAL_REQ_WR,
	243	.qp_type = IB_QPT_RC,
	244	};
	245	int rc;
	246
	247	lnk->roce_qp = ib_create_qp(lnk->roce_pd, &qp_attr);
	248	rc = PTR_ERR_OR_ZERO(lnk->roce_qp);
	249	if (IS_ERR(lnk->roce_qp))
	250	lnk->roce_qp = NULL;
	251	else
	252	smc_wr_remember_qp_attr(lnk);
	253	return rc;
	254	}
	255
897e1c24 UB	256	void smc_ib_put_memory_region(struct ib_mr *mr)
	257	{
	258	ib_dereg_mr(mr);
	259	}
	260
	261	static int smc_ib_map_mr_sg(struct smc_buf_desc *buf_slot)
	262	{
	263	unsigned int offset = 0;
	264	int sg_num;
	265
	266	/* map the largest prefix of a dma mapped SG list */
	267	sg_num = ib_map_mr_sg(buf_slot->mr_rx[SMC_SINGLE_LINK],
	268	buf_slot->sgt[SMC_SINGLE_LINK].sgl,
	269	buf_slot->sgt[SMC_SINGLE_LINK].orig_nents,
	270	&offset, PAGE_SIZE);
	271
	272	return sg_num;
	273	}
	274
	275	/* Allocate a memory region and map the dma mapped SG list of buf_slot */
	276	int smc_ib_get_memory_region(struct ib_pd *pd, int access_flags,
	277	struct smc_buf_desc *buf_slot)
	278	{
	279	if (buf_slot->mr_rx[SMC_SINGLE_LINK])
	280	return 0; /* already done */
	281
	282	buf_slot->mr_rx[SMC_SINGLE_LINK] =
	283	ib_alloc_mr(pd, IB_MR_TYPE_MEM_REG, 1 << buf_slot->order);
	284	if (IS_ERR(buf_slot->mr_rx[SMC_SINGLE_LINK])) {
	285	int rc;
	286
	287	rc = PTR_ERR(buf_slot->mr_rx[SMC_SINGLE_LINK]);
	288	buf_slot->mr_rx[SMC_SINGLE_LINK] = NULL;
	289	return rc;
	290	}
	291
	292	if (smc_ib_map_mr_sg(buf_slot) != 1)
	293	return -EINVAL;
	294
	295	return 0;
	296	}
	297
a3fe3d01 UB	298	/* Map a new TX or RX buffer SG-table to DMA */
	299	int smc_ib_buf_map_sg(struct smc_ib_device *smcibdev,
	300	struct smc_buf_desc *buf_slot,
	301	enum dma_data_direction data_direction)
	302	{
	303	int mapped_nents;
	304
	305	mapped_nents = ib_dma_map_sg(smcibdev->ibdev,
	306	buf_slot->sgt[SMC_SINGLE_LINK].sgl,
	307	buf_slot->sgt[SMC_SINGLE_LINK].orig_nents,
	308	data_direction);
	309	if (!mapped_nents)
	310	return -ENOMEM;
	311
	312	return mapped_nents;
	313	}
	314
	315	void smc_ib_buf_unmap_sg(struct smc_ib_device *smcibdev,
	316	struct smc_buf_desc *buf_slot,
	317	enum dma_data_direction data_direction)
	318	{
	319	if (!buf_slot->sgt[SMC_SINGLE_LINK].sgl->dma_address)
	320	return; /* already unmapped */
	321
	322	ib_dma_unmap_sg(smcibdev->ibdev,
	323	buf_slot->sgt[SMC_SINGLE_LINK].sgl,
	324	buf_slot->sgt[SMC_SINGLE_LINK].orig_nents,
	325	data_direction);
	326	buf_slot->sgt[SMC_SINGLE_LINK].sgl->dma_address = 0;
	327	}
	328
a4cf0443 UB	329	static int smc_ib_fill_gid_and_mac(struct smc_ib_device *smcibdev, u8 ibport)
	330	{
	331	struct net_device *ndev;
	332	int rc;
	333
	334	rc = ib_query_gid(smcibdev->ibdev, ibport, 0,
	335	&smcibdev->gid[ibport - 1], NULL);
	336	/* the SMC protocol requires specification of the roce MAC address;
	337	* if net_device cannot be determined, it can be derived from gid 0
	338	*/
	339	ndev = smcibdev->ibdev->get_netdev(smcibdev->ibdev, ibport);
	340	if (ndev) {
	341	memcpy(&smcibdev->mac, ndev->dev_addr, ETH_ALEN);
	342	} else if (!rc) {
	343	memcpy(&smcibdev->mac[ibport - 1][0],
	344	&smcibdev->gid[ibport - 1].raw[8], 3);
	345	memcpy(&smcibdev->mac[ibport - 1][3],
	346	&smcibdev->gid[ibport - 1].raw[13], 3);
	347	smcibdev->mac[ibport - 1][0] &= ~0x02;
	348	}
	349	return rc;
	350	}
	351
	352	/* Create an identifier unique for this instance of SMC-R.
	353	* The MAC-address of the first active registered IB device
	354	* plus a random 2-byte number is used to create this identifier.
	355	* This name is delivered to the peer during connection initialization.
	356	*/
	357	static inline void smc_ib_define_local_systemid(struct smc_ib_device *smcibdev,
	358	u8 ibport)
	359	{
	360	memcpy(&local_systemid[2], &smcibdev->mac[ibport - 1],
	361	sizeof(smcibdev->mac[ibport - 1]));
	362	get_random_bytes(&local_systemid[0], 2);
	363	}
	364
	365	bool smc_ib_port_active(struct smc_ib_device *smcibdev, u8 ibport)
	366	{
	367	return smcibdev->pattr[ibport - 1].state == IB_PORT_ACTIVE;
	368	}
	369
	370	int smc_ib_remember_port_attr(struct smc_ib_device *smcibdev, u8 ibport)
	371	{
	372	int rc;
	373
	374	memset(&smcibdev->pattr[ibport - 1], 0,
	375	sizeof(smcibdev->pattr[ibport - 1]));
	376	rc = ib_query_port(smcibdev->ibdev, ibport,
	377	&smcibdev->pattr[ibport - 1]);
	378	if (rc)
	379	goto out;
	380	rc = smc_ib_fill_gid_and_mac(smcibdev, ibport);
	381	if (rc)
	382	goto out;
	383	if (!strncmp(local_systemid, SMC_LOCAL_SYSTEMID_RESET,
	384	sizeof(local_systemid)) &&
	385	smc_ib_port_active(smcibdev, ibport))
	386	/* create unique system identifier */
	387	smc_ib_define_local_systemid(smcibdev, ibport);
	388	out:
	389	return rc;
	390	}
	391
bd4ad577 UB	392	long smc_ib_setup_per_ibdev(struct smc_ib_device *smcibdev)
	393	{
	394	struct ib_cq_init_attr cqattr = {
	395	.cqe = SMC_WR_MAX_CQE, .comp_vector = 0 };
	396	long rc;
	397
	398	smcibdev->roce_cq_send = ib_create_cq(smcibdev->ibdev,
	399	smc_wr_tx_cq_handler, NULL,
	400	smcibdev, &cqattr);
	401	rc = PTR_ERR_OR_ZERO(smcibdev->roce_cq_send);
	402	if (IS_ERR(smcibdev->roce_cq_send)) {
	403	smcibdev->roce_cq_send = NULL;
	404	return rc;
	405	}
	406	smcibdev->roce_cq_recv = ib_create_cq(smcibdev->ibdev,
	407	smc_wr_rx_cq_handler, NULL,
	408	smcibdev, &cqattr);
	409	rc = PTR_ERR_OR_ZERO(smcibdev->roce_cq_recv);
	410	if (IS_ERR(smcibdev->roce_cq_recv)) {
	411	smcibdev->roce_cq_recv = NULL;
	412	goto err;
	413	}
	414	INIT_IB_EVENT_HANDLER(&smcibdev->event_handler, smcibdev->ibdev,
	415	smc_ib_global_event_handler);
	416	ib_register_event_handler(&smcibdev->event_handler);
	417	smc_wr_add_dev(smcibdev);
	418	smcibdev->initialized = 1;
	419	return rc;
	420
	421	err:
	422	ib_destroy_cq(smcibdev->roce_cq_send);
	423	return rc;
	424	}
	425
	426	static void smc_ib_cleanup_per_ibdev(struct smc_ib_device *smcibdev)
	427	{
	428	if (!smcibdev->initialized)
	429	return;
	430	smc_wr_remove_dev(smcibdev);
	431	ib_unregister_event_handler(&smcibdev->event_handler);
	432	ib_destroy_cq(smcibdev->roce_cq_recv);
	433	ib_destroy_cq(smcibdev->roce_cq_send);
	434	}
	435
a4cf0443 UB	436	static struct ib_client smc_ib_client;
	437
	438	/* callback function for ib_register_client() */
	439	static void smc_ib_add_dev(struct ib_device *ibdev)
	440	{
	441	struct smc_ib_device *smcibdev;
	442
	443	if (ibdev->node_type != RDMA_NODE_IB_CA)
	444	return;
	445
	446	smcibdev = kzalloc(sizeof(*smcibdev), GFP_KERNEL);
	447	if (!smcibdev)
	448	return;
	449
	450	smcibdev->ibdev = ibdev;
bd4ad577	451	INIT_WORK(&smcibdev->port_event_work, smc_ib_port_event_work);
a4cf0443 UB	452
	453	spin_lock(&smc_ib_devices.lock);
	454	list_add_tail(&smcibdev->list, &smc_ib_devices.list);
	455	spin_unlock(&smc_ib_devices.lock);
	456	ib_set_client_data(ibdev, &smc_ib_client, smcibdev);
	457	}
	458
	459	/* callback function for ib_register_client() */
	460	static void smc_ib_remove_dev(struct ib_device ibdev, void client_data)
	461	{
	462	struct smc_ib_device *smcibdev;
	463
	464	smcibdev = ib_get_client_data(ibdev, &smc_ib_client);
	465	ib_set_client_data(ibdev, &smc_ib_client, NULL);
	466	spin_lock(&smc_ib_devices.lock);
	467	list_del_init(&smcibdev->list); /* remove from smc_ib_devices */
	468	spin_unlock(&smc_ib_devices.lock);
6812baab	469	smc_pnet_remove_by_ibdev(smcibdev);
bd4ad577	470	smc_ib_cleanup_per_ibdev(smcibdev);
a4cf0443 UB	471	kfree(smcibdev);
	472	}
	473
	474	static struct ib_client smc_ib_client = {
	475	.name = "smc_ib",
	476	.add = smc_ib_add_dev,
	477	.remove = smc_ib_remove_dev,
	478	};
	479
	480	int __init smc_ib_register_client(void)
	481	{
	482	return ib_register_client(&smc_ib_client);
	483	}
	484
	485	void smc_ib_unregister_client(void)
	486	{
	487	ib_unregister_client(&smc_ib_client);
	488	}