[mirror_ubuntu-bionic-kernel.git] / net / smc / smc_ib.c

// SPDX-License-Identifier: GPL-2.0
/*
 *  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 <linux/scatterlist.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_MAX_CQE 32766	/* max. # of completion queue elements */

#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;
}

/* synchronize buffer usage for cpu access */
void smc_ib_sync_sg_for_cpu(struct smc_ib_device *smcibdev,
			    struct smc_buf_desc *buf_slot,
			    enum dma_data_direction data_direction)
{
	struct scatterlist *sg;
	unsigned int i;

	/* for now there is just one DMA address */
	for_each_sg(buf_slot->sgt[SMC_SINGLE_LINK].sgl, sg,
		    buf_slot->sgt[SMC_SINGLE_LINK].nents, i) {
		if (!sg_dma_len(sg))
			break;
		ib_dma_sync_single_for_cpu(smcibdev->ibdev,
					   sg_dma_address(sg),
					   sg_dma_len(sg),
					   data_direction);
	}
}

/* synchronize buffer usage for device access */
void smc_ib_sync_sg_for_device(struct smc_ib_device *smcibdev,
			       struct smc_buf_desc *buf_slot,
			       enum dma_data_direction data_direction)
{
	struct scatterlist *sg;
	unsigned int i;

	/* for now there is just one DMA address */
	for_each_sg(buf_slot->sgt[SMC_SINGLE_LINK].sgl, sg,
		    buf_slot->sgt[SMC_SINGLE_LINK].nents, i) {
		if (!sg_dma_len(sg))
			break;
		ib_dma_sync_single_for_device(smcibdev->ibdev,
					      sg_dma_address(sg),
					      sg_dma_len(sg),
					      data_direction);
	}
}

/* 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 ib_gid_attr gattr;
	int rc;

	rc = ib_query_gid(smcibdev->ibdev, ibport, 0,
			  &smcibdev->gid[ibport - 1], &gattr);
	if (rc || !gattr.ndev)
		return -ENODEV;

	memcpy(smcibdev->mac[ibport - 1], gattr.ndev->dev_addr, ETH_ALEN);
	dev_put(gattr.ndev);
	return 0;
}

/* 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;
	/* the SMC protocol requires specification of the RoCE MAC address */
	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_MAX_CQE, .comp_vector = 0 };
	int cqe_size_order, smc_order;
	long rc;

	/* the calculated number of cq entries fits to mlx5 cq allocation */
	cqe_size_order = cache_line_size() == 128 ? 7 : 6;
	smc_order = MAX_ORDER - cqe_size_order - 1;
	if (SMC_MAX_CQE + 2 > (0x00000001 << smc_order) * PAGE_SIZE)
		cqattr.cqe = (0x00000001 << smc_order) * PAGE_SIZE - 2;
	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
b2441318	1	// SPDX-License-Identifier: GPL-2.0
a4cf0443 UB	2	/*
	3	* Shared Memory Communications over RDMA (SMC-R) and RoCE
	4	*
	5	* IB infrastructure:
	6	* Establish SMC-R as an Infiniband Client to be notified about added and
	7	* removed IB devices of type RDMA.
	8	* Determine device and port characteristics for these IB devices.
	9	*
	10	* Copyright IBM Corp. 2016
	11	*
	12	* Author(s): Ursula Braun <ubraun@linux.vnet.ibm.com>
	13	*/
	14
	15	#include <linux/random.h>
bd4ad577	16	#include <linux/workqueue.h>
10428dd8	17	#include <linux/scatterlist.h>
a4cf0443 UB	18	#include <rdma/ib_verbs.h>
a4cf0443 UB	19
6812baab	20	#include "smc_pnet.h"
a4cf0443	21	#include "smc_ib.h"
cd6851f3	22	#include "smc_core.h"
f38ba179	23	#include "smc_wr.h"
a4cf0443 UB	24	#include "smc.h"
a4cf0443 UB	25
901e7829 UB	26	#define SMC_MAX_CQE 32766 /* max. # of completion queue elements */
901e7829 UB	27
bd4ad577 UB	28	#define SMC_QP_MIN_RNR_TIMER 5
	29	#define SMC_QP_TIMEOUT 15 /* 4096 * 2 ** timeout usec */
	30	#define SMC_QP_RETRY_CNT 7 /* 7: infinite */
	31	#define SMC_QP_RNR_RETRY 7 /* 7: infinite */
	32
a4cf0443 UB	33	struct smc_ib_devices smc_ib_devices = { /* smc-registered ib devices */
	34	.lock = __SPIN_LOCK_UNLOCKED(smc_ib_devices.lock),
	35	.list = LIST_HEAD_INIT(smc_ib_devices.list),
	36	};
	37
	38	#define SMC_LOCAL_SYSTEMID_RESET "%%%%%%%"
	39
	40	u8 local_systemid[SMC_SYSTEMID_LEN] = SMC_LOCAL_SYSTEMID_RESET; /* unique system
	41	* identifier
	42	*/
	43
bd4ad577 UB	44	static int smc_ib_modify_qp_init(struct smc_link *lnk)
	45	{
	46	struct ib_qp_attr qp_attr;
	47
	48	memset(&qp_attr, 0, sizeof(qp_attr));
	49	qp_attr.qp_state = IB_QPS_INIT;
	50	qp_attr.pkey_index = 0;
	51	qp_attr.port_num = lnk->ibport;
	52	qp_attr.qp_access_flags = IB_ACCESS_LOCAL_WRITE
	53	\| IB_ACCESS_REMOTE_WRITE;
	54	return ib_modify_qp(lnk->roce_qp, &qp_attr,
	55	IB_QP_STATE \| IB_QP_PKEY_INDEX \|
	56	IB_QP_ACCESS_FLAGS \| IB_QP_PORT);
	57	}
	58
	59	static int smc_ib_modify_qp_rtr(struct smc_link *lnk)
	60	{
	61	enum ib_qp_attr_mask qp_attr_mask =
	62	IB_QP_STATE \| IB_QP_AV \| IB_QP_PATH_MTU \| IB_QP_DEST_QPN \|
	63	IB_QP_RQ_PSN \| IB_QP_MAX_DEST_RD_ATOMIC \| IB_QP_MIN_RNR_TIMER;
	64	struct ib_qp_attr qp_attr;
	65
	66	memset(&qp_attr, 0, sizeof(qp_attr));
	67	qp_attr.qp_state = IB_QPS_RTR;
	68	qp_attr.path_mtu = min(lnk->path_mtu, lnk->peer_mtu);
44c58487	69	qp_attr.ah_attr.type = RDMA_AH_ATTR_TYPE_ROCE;
d8966fcd DC	70	rdma_ah_set_port_num(&qp_attr.ah_attr, lnk->ibport);
	71	rdma_ah_set_grh(&qp_attr.ah_attr, NULL, 0, 0, 1, 0);
	72	rdma_ah_set_dgid_raw(&qp_attr.ah_attr, lnk->peer_gid);
44c58487	73	memcpy(&qp_attr.ah_attr.roce.dmac, lnk->peer_mac,
bd4ad577 UB	74	sizeof(lnk->peer_mac));
	75	qp_attr.dest_qp_num = lnk->peer_qpn;
	76	qp_attr.rq_psn = lnk->peer_psn; /* starting receive packet seq # */
	77	qp_attr.max_dest_rd_atomic = 1; /* max # of resources for incoming
	78	* requests
	79	*/
	80	qp_attr.min_rnr_timer = SMC_QP_MIN_RNR_TIMER;
	81
	82	return ib_modify_qp(lnk->roce_qp, &qp_attr, qp_attr_mask);
	83	}
	84
	85	int smc_ib_modify_qp_rts(struct smc_link *lnk)
	86	{
	87	struct ib_qp_attr qp_attr;
	88
	89	memset(&qp_attr, 0, sizeof(qp_attr));
	90	qp_attr.qp_state = IB_QPS_RTS;
	91	qp_attr.timeout = SMC_QP_TIMEOUT; /* local ack timeout */
	92	qp_attr.retry_cnt = SMC_QP_RETRY_CNT; /* retry count */
	93	qp_attr.rnr_retry = SMC_QP_RNR_RETRY; /* RNR retries, 7=infinite */
	94	qp_attr.sq_psn = lnk->psn_initial; /* starting send packet seq # */
	95	qp_attr.max_rd_atomic = 1; /* # of outstanding RDMA reads and
	96	* atomic ops allowed
	97	*/
	98	return ib_modify_qp(lnk->roce_qp, &qp_attr,
	99	IB_QP_STATE \| IB_QP_TIMEOUT \| IB_QP_RETRY_CNT \|
	100	IB_QP_SQ_PSN \| IB_QP_RNR_RETRY \|
	101	IB_QP_MAX_QP_RD_ATOMIC);
	102	}
	103
	104	int smc_ib_modify_qp_reset(struct smc_link *lnk)
	105	{
	106	struct ib_qp_attr qp_attr;
	107
	108	memset(&qp_attr, 0, sizeof(qp_attr));
	109	qp_attr.qp_state = IB_QPS_RESET;
	110	return ib_modify_qp(lnk->roce_qp, &qp_attr, IB_QP_STATE);
	111	}
	112
	113	int smc_ib_ready_link(struct smc_link *lnk)
	114	{
	115	struct smc_link_group *lgr =
	116	container_of(lnk, struct smc_link_group, lnk[0]);
	117	int rc = 0;
	118
	119	rc = smc_ib_modify_qp_init(lnk);
	120	if (rc)
	121	goto out;
	122
	123	rc = smc_ib_modify_qp_rtr(lnk);
	124	if (rc)
	125	goto out;
	126	smc_wr_remember_qp_attr(lnk);
	127	rc = ib_req_notify_cq(lnk->smcibdev->roce_cq_recv,
	128	IB_CQ_SOLICITED_MASK);
	129	if (rc)
	130	goto out;
	131	rc = smc_wr_rx_post_init(lnk);
	132	if (rc)
	133	goto out;
	134	smc_wr_remember_qp_attr(lnk);
	135
	136	if (lgr->role == SMC_SERV) {
	137	rc = smc_ib_modify_qp_rts(lnk);
138	if (rc)
139	goto out;
140	smc_wr_remember_qp_attr(lnk);
141	}
142	out:
143	return rc;
144	}
145
146	/* process context wrapper for might_sleep smc_ib_remember_port_attr */
147	static void smc_ib_port_event_work(struct work_struct *work)
148	{
149	struct smc_ib_device *smcibdev = container_of(
150	work, struct smc_ib_device, port_event_work);
151	u8 port_idx;
152
153	for_each_set_bit(port_idx, &smcibdev->port_event_mask, SMC_MAX_PORTS) {
154	smc_ib_remember_port_attr(smcibdev, port_idx + 1);
155	clear_bit(port_idx, &smcibdev->port_event_mask);
156	}
157	}
158
159	/* can be called in IRQ context */
160	static void smc_ib_global_event_handler(struct ib_event_handler *handler,
161	struct ib_event *ibevent)
162	{
163	struct smc_ib_device *smcibdev;
164	u8 port_idx;
165
166	smcibdev = container_of(handler, struct smc_ib_device, event_handler);
bd4ad577 UB	167
	168	switch (ibevent->event) {
	169	case IB_EVENT_PORT_ERR:
	170	port_idx = ibevent->element.port_num - 1;
	171	set_bit(port_idx, &smcibdev->port_event_mask);
	172	schedule_work(&smcibdev->port_event_work);
	173	/* fall through */
	174	case IB_EVENT_DEVICE_FATAL:
	175	/* tbd in follow-on patch:
	176	* abnormal close of corresponding connections
	177	*/
	178	break;
	179	case IB_EVENT_PORT_ACTIVE:
	180	port_idx = ibevent->element.port_num - 1;
	181	set_bit(port_idx, &smcibdev->port_event_mask);
	182	schedule_work(&smcibdev->port_event_work);
	183	break;
	184	default:
	185	break;
	186	}
	187	}
	188
f38ba179 UB	189	void smc_ib_dealloc_protection_domain(struct smc_link *lnk)
	190	{
	191	ib_dealloc_pd(lnk->roce_pd);
	192	lnk->roce_pd = NULL;
	193	}
	194
	195	int smc_ib_create_protection_domain(struct smc_link *lnk)
	196	{
	197	int rc;
	198
897e1c24	199	lnk->roce_pd = ib_alloc_pd(lnk->smcibdev->ibdev, 0);
f38ba179 UB	200	rc = PTR_ERR_OR_ZERO(lnk->roce_pd);
	201	if (IS_ERR(lnk->roce_pd))
	202	lnk->roce_pd = NULL;
	203	return rc;
	204	}
	205
	206	static void smc_ib_qp_event_handler(struct ib_event ibevent, void priv)
	207	{
	208	switch (ibevent->event) {
	209	case IB_EVENT_DEVICE_FATAL:
	210	case IB_EVENT_GID_CHANGE:
	211	case IB_EVENT_PORT_ERR:
	212	case IB_EVENT_QP_ACCESS_ERR:
	213	/* tbd in follow-on patch:
	214	* abnormal close of corresponding connections
	215	*/
	216	break;
	217	default:
	218	break;
	219	}
	220	}
	221
	222	void smc_ib_destroy_queue_pair(struct smc_link *lnk)
	223	{
	224	ib_destroy_qp(lnk->roce_qp);
	225	lnk->roce_qp = NULL;
	226	}
	227
	228	/* create a queue pair within the protection domain for a link */
	229	int smc_ib_create_queue_pair(struct smc_link *lnk)
	230	{
	231	struct ib_qp_init_attr qp_attr = {
	232	.event_handler = smc_ib_qp_event_handler,
	233	.qp_context = lnk,
	234	.send_cq = lnk->smcibdev->roce_cq_send,
	235	.recv_cq = lnk->smcibdev->roce_cq_recv,
	236	.srq = NULL,
	237	.cap = {
f38ba179 UB	238	/* include unsolicited rdma_writes as well,
	239	* there are max. 2 RDMA_WRITE per 1 WR_SEND
	240	*/
652a1e41	241	.max_send_wr = SMC_WR_BUF_CNT * 3,
f38ba179 UB	242	.max_recv_wr = SMC_WR_BUF_CNT * 3,
	243	.max_send_sge = SMC_IB_MAX_SEND_SGE,
	244	.max_recv_sge = 1,
f38ba179 UB	245	},
	246	.sq_sig_type = IB_SIGNAL_REQ_WR,
	247	.qp_type = IB_QPT_RC,
	248	};
	249	int rc;
	250
	251	lnk->roce_qp = ib_create_qp(lnk->roce_pd, &qp_attr);
	252	rc = PTR_ERR_OR_ZERO(lnk->roce_qp);
	253	if (IS_ERR(lnk->roce_qp))
	254	lnk->roce_qp = NULL;
	255	else
	256	smc_wr_remember_qp_attr(lnk);
	257	return rc;
	258	}
	259
897e1c24 UB	260	void smc_ib_put_memory_region(struct ib_mr *mr)
	261	{
	262	ib_dereg_mr(mr);
	263	}
	264
	265	static int smc_ib_map_mr_sg(struct smc_buf_desc *buf_slot)
	266	{
	267	unsigned int offset = 0;
	268	int sg_num;
	269
	270	/* map the largest prefix of a dma mapped SG list */
	271	sg_num = ib_map_mr_sg(buf_slot->mr_rx[SMC_SINGLE_LINK],
	272	buf_slot->sgt[SMC_SINGLE_LINK].sgl,
	273	buf_slot->sgt[SMC_SINGLE_LINK].orig_nents,
	274	&offset, PAGE_SIZE);
	275
	276	return sg_num;
	277	}
	278
	279	/* Allocate a memory region and map the dma mapped SG list of buf_slot */
	280	int smc_ib_get_memory_region(struct ib_pd *pd, int access_flags,
	281	struct smc_buf_desc *buf_slot)
	282	{
	283	if (buf_slot->mr_rx[SMC_SINGLE_LINK])
	284	return 0; /* already done */
	285
	286	buf_slot->mr_rx[SMC_SINGLE_LINK] =
	287	ib_alloc_mr(pd, IB_MR_TYPE_MEM_REG, 1 << buf_slot->order);
	288	if (IS_ERR(buf_slot->mr_rx[SMC_SINGLE_LINK])) {
	289	int rc;
	290
	291	rc = PTR_ERR(buf_slot->mr_rx[SMC_SINGLE_LINK]);
	292	buf_slot->mr_rx[SMC_SINGLE_LINK] = NULL;
	293	return rc;
	294	}
	295
	296	if (smc_ib_map_mr_sg(buf_slot) != 1)
	297	return -EINVAL;
	298
	299	return 0;
	300	}
	301
10428dd8 UB	302	/* synchronize buffer usage for cpu access */
	303	void smc_ib_sync_sg_for_cpu(struct smc_ib_device *smcibdev,
	304	struct smc_buf_desc *buf_slot,
	305	enum dma_data_direction data_direction)
	306	{
	307	struct scatterlist *sg;
	308	unsigned int i;
	309
	310	/* for now there is just one DMA address */
	311	for_each_sg(buf_slot->sgt[SMC_SINGLE_LINK].sgl, sg,
	312	buf_slot->sgt[SMC_SINGLE_LINK].nents, i) {
	313	if (!sg_dma_len(sg))
	314	break;
	315	ib_dma_sync_single_for_cpu(smcibdev->ibdev,
	316	sg_dma_address(sg),
	317	sg_dma_len(sg),
	318	data_direction);
	319	}
	320	}
	321
	322	/* synchronize buffer usage for device access */
	323	void smc_ib_sync_sg_for_device(struct smc_ib_device *smcibdev,
	324	struct smc_buf_desc *buf_slot,
	325	enum dma_data_direction data_direction)
	326	{
	327	struct scatterlist *sg;
	328	unsigned int i;
	329
	330	/* for now there is just one DMA address */
	331	for_each_sg(buf_slot->sgt[SMC_SINGLE_LINK].sgl, sg,
	332	buf_slot->sgt[SMC_SINGLE_LINK].nents, i) {
	333	if (!sg_dma_len(sg))
	334	break;
	335	ib_dma_sync_single_for_device(smcibdev->ibdev,
	336	sg_dma_address(sg),
	337	sg_dma_len(sg),
	338	data_direction);
	339	}
	340	}
	341
a3fe3d01 UB	342	/* Map a new TX or RX buffer SG-table to DMA */
	343	int smc_ib_buf_map_sg(struct smc_ib_device *smcibdev,
	344	struct smc_buf_desc *buf_slot,
	345	enum dma_data_direction data_direction)
	346	{
	347	int mapped_nents;
	348
	349	mapped_nents = ib_dma_map_sg(smcibdev->ibdev,
	350	buf_slot->sgt[SMC_SINGLE_LINK].sgl,
	351	buf_slot->sgt[SMC_SINGLE_LINK].orig_nents,
	352	data_direction);
	353	if (!mapped_nents)
	354	return -ENOMEM;
	355
	356	return mapped_nents;
	357	}
	358
	359	void smc_ib_buf_unmap_sg(struct smc_ib_device *smcibdev,
	360	struct smc_buf_desc *buf_slot,
	361	enum dma_data_direction data_direction)
	362	{
	363	if (!buf_slot->sgt[SMC_SINGLE_LINK].sgl->dma_address)
	364	return; /* already unmapped */
	365
	366	ib_dma_unmap_sg(smcibdev->ibdev,
	367	buf_slot->sgt[SMC_SINGLE_LINK].sgl,
	368	buf_slot->sgt[SMC_SINGLE_LINK].orig_nents,
	369	data_direction);
	370	buf_slot->sgt[SMC_SINGLE_LINK].sgl->dma_address = 0;
	371	}
	372
a4cf0443 UB	373	static int smc_ib_fill_gid_and_mac(struct smc_ib_device *smcibdev, u8 ibport)
a4cf0443 UB	374	{
d921c420	375	struct ib_gid_attr gattr;
a4cf0443 UB	376	int rc;
	377
	378	rc = ib_query_gid(smcibdev->ibdev, ibport, 0,
d921c420 UB	379	&smcibdev->gid[ibport - 1], &gattr);
	380	if (rc \|\| !gattr.ndev)
	381	return -ENODEV;
	382
	383	memcpy(smcibdev->mac[ibport - 1], gattr.ndev->dev_addr, ETH_ALEN);
	384	dev_put(gattr.ndev);
	385	return 0;
a4cf0443 UB	386	}
	387
	388	/* Create an identifier unique for this instance of SMC-R.
	389	* The MAC-address of the first active registered IB device
	390	* plus a random 2-byte number is used to create this identifier.
	391	* This name is delivered to the peer during connection initialization.
	392	*/
	393	static inline void smc_ib_define_local_systemid(struct smc_ib_device *smcibdev,
	394	u8 ibport)
	395	{
	396	memcpy(&local_systemid[2], &smcibdev->mac[ibport - 1],
	397	sizeof(smcibdev->mac[ibport - 1]));
	398	get_random_bytes(&local_systemid[0], 2);
	399	}
	400
	401	bool smc_ib_port_active(struct smc_ib_device *smcibdev, u8 ibport)
	402	{
	403	return smcibdev->pattr[ibport - 1].state == IB_PORT_ACTIVE;
	404	}
	405
	406	int smc_ib_remember_port_attr(struct smc_ib_device *smcibdev, u8 ibport)
	407	{
	408	int rc;
	409
	410	memset(&smcibdev->pattr[ibport - 1], 0,
	411	sizeof(smcibdev->pattr[ibport - 1]));
	412	rc = ib_query_port(smcibdev->ibdev, ibport,
	413	&smcibdev->pattr[ibport - 1]);
	414	if (rc)
	415	goto out;
d921c420	416	/* the SMC protocol requires specification of the RoCE MAC address */
a4cf0443 UB	417	rc = smc_ib_fill_gid_and_mac(smcibdev, ibport);
	418	if (rc)
	419	goto out;
	420	if (!strncmp(local_systemid, SMC_LOCAL_SYSTEMID_RESET,
	421	sizeof(local_systemid)) &&
	422	smc_ib_port_active(smcibdev, ibport))
	423	/* create unique system identifier */
	424	smc_ib_define_local_systemid(smcibdev, ibport);
	425	out:
	426	return rc;
	427	}
	428
bd4ad577 UB	429	long smc_ib_setup_per_ibdev(struct smc_ib_device *smcibdev)
	430	{
	431	struct ib_cq_init_attr cqattr = {
901e7829 UB	432	.cqe = SMC_MAX_CQE, .comp_vector = 0 };
901e7829 UB	433	int cqe_size_order, smc_order;
bd4ad577 UB	434	long rc;
bd4ad577 UB	435
901e7829 UB	436	/* the calculated number of cq entries fits to mlx5 cq allocation */
	437	cqe_size_order = cache_line_size() == 128 ? 7 : 6;
	438	smc_order = MAX_ORDER - cqe_size_order - 1;
	439	if (SMC_MAX_CQE + 2 > (0x00000001 << smc_order) * PAGE_SIZE)
	440	cqattr.cqe = (0x00000001 << smc_order) * PAGE_SIZE - 2;
bd4ad577 UB	441	smcibdev->roce_cq_send = ib_create_cq(smcibdev->ibdev,
	442	smc_wr_tx_cq_handler, NULL,
	443	smcibdev, &cqattr);
	444	rc = PTR_ERR_OR_ZERO(smcibdev->roce_cq_send);
	445	if (IS_ERR(smcibdev->roce_cq_send)) {
	446	smcibdev->roce_cq_send = NULL;
	447	return rc;
	448	}
	449	smcibdev->roce_cq_recv = ib_create_cq(smcibdev->ibdev,
	450	smc_wr_rx_cq_handler, NULL,
	451	smcibdev, &cqattr);
	452	rc = PTR_ERR_OR_ZERO(smcibdev->roce_cq_recv);
	453	if (IS_ERR(smcibdev->roce_cq_recv)) {
	454	smcibdev->roce_cq_recv = NULL;
	455	goto err;
	456	}
	457	INIT_IB_EVENT_HANDLER(&smcibdev->event_handler, smcibdev->ibdev,
	458	smc_ib_global_event_handler);
	459	ib_register_event_handler(&smcibdev->event_handler);
	460	smc_wr_add_dev(smcibdev);
	461	smcibdev->initialized = 1;
	462	return rc;
	463
	464	err:
	465	ib_destroy_cq(smcibdev->roce_cq_send);
	466	return rc;
	467	}
	468
	469	static void smc_ib_cleanup_per_ibdev(struct smc_ib_device *smcibdev)
	470	{
	471	if (!smcibdev->initialized)
	472	return;
	473	smc_wr_remove_dev(smcibdev);
	474	ib_unregister_event_handler(&smcibdev->event_handler);
	475	ib_destroy_cq(smcibdev->roce_cq_recv);
	476	ib_destroy_cq(smcibdev->roce_cq_send);
	477	}
	478
a4cf0443 UB	479	static struct ib_client smc_ib_client;
	480
	481	/* callback function for ib_register_client() */
	482	static void smc_ib_add_dev(struct ib_device *ibdev)
	483	{
	484	struct smc_ib_device *smcibdev;
	485
	486	if (ibdev->node_type != RDMA_NODE_IB_CA)
	487	return;
	488
	489	smcibdev = kzalloc(sizeof(*smcibdev), GFP_KERNEL);
	490	if (!smcibdev)
	491	return;
	492
	493	smcibdev->ibdev = ibdev;
bd4ad577	494	INIT_WORK(&smcibdev->port_event_work, smc_ib_port_event_work);
a4cf0443 UB	495
	496	spin_lock(&smc_ib_devices.lock);
	497	list_add_tail(&smcibdev->list, &smc_ib_devices.list);
	498	spin_unlock(&smc_ib_devices.lock);
	499	ib_set_client_data(ibdev, &smc_ib_client, smcibdev);
	500	}
	501
	502	/* callback function for ib_register_client() */
	503	static void smc_ib_remove_dev(struct ib_device ibdev, void client_data)
	504	{
	505	struct smc_ib_device *smcibdev;
	506
	507	smcibdev = ib_get_client_data(ibdev, &smc_ib_client);
	508	ib_set_client_data(ibdev, &smc_ib_client, NULL);
	509	spin_lock(&smc_ib_devices.lock);
	510	list_del_init(&smcibdev->list); /* remove from smc_ib_devices */
	511	spin_unlock(&smc_ib_devices.lock);
6812baab	512	smc_pnet_remove_by_ibdev(smcibdev);
bd4ad577	513	smc_ib_cleanup_per_ibdev(smcibdev);
a4cf0443 UB	514	kfree(smcibdev);
	515	}
	516
	517	static struct ib_client smc_ib_client = {
	518	.name = "smc_ib",
	519	.add = smc_ib_add_dev,
	520	.remove = smc_ib_remove_dev,
	521	};
	522
	523	int __init smc_ib_register_client(void)
	524	{
	525	return ib_register_client(&smc_ib_client);
	526	}
	527
	528	void smc_ib_unregister_client(void)
	529	{
	530	ib_unregister_client(&smc_ib_client);
	531	}