[linux-2.6-block.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 <rdma/ib_cache.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);
		if (!smc_ib_port_active(smcibdev, port_idx + 1))
			smc_port_terminate(smcibdev, port_idx + 1);
	}
}

/* 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:
	case IB_EVENT_DEVICE_FATAL:
	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)
{
	if (lnk->roce_pd)
		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)
{
	struct smc_ib_device *smcibdev =
		(struct smc_ib_device *)ibevent->device;
	u8 port_idx;

	switch (ibevent->event) {
	case IB_EVENT_DEVICE_FATAL:
	case IB_EVENT_GID_CHANGE:
	case IB_EVENT_PORT_ERR:
	case IB_EVENT_QP_ACCESS_ERR:
		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_destroy_queue_pair(struct smc_link *lnk)
{
	if (lnk->roce_qp)
		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;
	smcibdev->initialized = 0;
	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	18	#include <rdma/ib_verbs.h>
ddb457c6	19	#include <rdma/ib_cache.h>
a4cf0443	20
6812baab	21	#include "smc_pnet.h"
a4cf0443	22	#include "smc_ib.h"
cd6851f3	23	#include "smc_core.h"
f38ba179	24	#include "smc_wr.h"
a4cf0443 UB	25	#include "smc.h"
a4cf0443 UB	26
c9f4c6cf UB	27	#define SMC_MAX_CQE 32766 /* max. # of completion queue elements */
c9f4c6cf UB	28
bd4ad577 UB	29	#define SMC_QP_MIN_RNR_TIMER 5
	30	#define SMC_QP_TIMEOUT 15 /* 4096 * 2 ** timeout usec */
	31	#define SMC_QP_RETRY_CNT 7 /* 7: infinite */
	32	#define SMC_QP_RNR_RETRY 7 /* 7: infinite */
	33
a4cf0443 UB	34	struct smc_ib_devices smc_ib_devices = { /* smc-registered ib devices */
	35	.lock = __SPIN_LOCK_UNLOCKED(smc_ib_devices.lock),
	36	.list = LIST_HEAD_INIT(smc_ib_devices.list),
	37	};
	38
	39	#define SMC_LOCAL_SYSTEMID_RESET "%%%%%%%"
	40
	41	u8 local_systemid[SMC_SYSTEMID_LEN] = SMC_LOCAL_SYSTEMID_RESET; /* unique system
	42	* identifier
	43	*/
	44
bd4ad577 UB	45	static int smc_ib_modify_qp_init(struct smc_link *lnk)
	46	{
	47	struct ib_qp_attr qp_attr;
	48
	49	memset(&qp_attr, 0, sizeof(qp_attr));
	50	qp_attr.qp_state = IB_QPS_INIT;
	51	qp_attr.pkey_index = 0;
	52	qp_attr.port_num = lnk->ibport;
	53	qp_attr.qp_access_flags = IB_ACCESS_LOCAL_WRITE
	54	\| IB_ACCESS_REMOTE_WRITE;
	55	return ib_modify_qp(lnk->roce_qp, &qp_attr,
	56	IB_QP_STATE \| IB_QP_PKEY_INDEX \|
	57	IB_QP_ACCESS_FLAGS \| IB_QP_PORT);
	58	}
	59
	60	static int smc_ib_modify_qp_rtr(struct smc_link *lnk)
	61	{
	62	enum ib_qp_attr_mask qp_attr_mask =
	63	IB_QP_STATE \| IB_QP_AV \| IB_QP_PATH_MTU \| IB_QP_DEST_QPN \|
	64	IB_QP_RQ_PSN \| IB_QP_MAX_DEST_RD_ATOMIC \| IB_QP_MIN_RNR_TIMER;
	65	struct ib_qp_attr qp_attr;
	66
	67	memset(&qp_attr, 0, sizeof(qp_attr));
	68	qp_attr.qp_state = IB_QPS_RTR;
	69	qp_attr.path_mtu = min(lnk->path_mtu, lnk->peer_mtu);
44c58487	70	qp_attr.ah_attr.type = RDMA_AH_ATTR_TYPE_ROCE;
d8966fcd DC	71	rdma_ah_set_port_num(&qp_attr.ah_attr, lnk->ibport);
	72	rdma_ah_set_grh(&qp_attr.ah_attr, NULL, 0, 0, 1, 0);
	73	rdma_ah_set_dgid_raw(&qp_attr.ah_attr, lnk->peer_gid);
44c58487	74	memcpy(&qp_attr.ah_attr.roce.dmac, lnk->peer_mac,
bd4ad577 UB	75	sizeof(lnk->peer_mac));
	76	qp_attr.dest_qp_num = lnk->peer_qpn;
	77	qp_attr.rq_psn = lnk->peer_psn; /* starting receive packet seq # */
	78	qp_attr.max_dest_rd_atomic = 1; /* max # of resources for incoming
	79	* requests
	80	*/
	81	qp_attr.min_rnr_timer = SMC_QP_MIN_RNR_TIMER;
	82
	83	return ib_modify_qp(lnk->roce_qp, &qp_attr, qp_attr_mask);
	84	}
	85
	86	int smc_ib_modify_qp_rts(struct smc_link *lnk)
	87	{
	88	struct ib_qp_attr qp_attr;
	89
	90	memset(&qp_attr, 0, sizeof(qp_attr));
	91	qp_attr.qp_state = IB_QPS_RTS;
	92	qp_attr.timeout = SMC_QP_TIMEOUT; /* local ack timeout */
	93	qp_attr.retry_cnt = SMC_QP_RETRY_CNT; /* retry count */
	94	qp_attr.rnr_retry = SMC_QP_RNR_RETRY; /* RNR retries, 7=infinite */
	95	qp_attr.sq_psn = lnk->psn_initial; /* starting send packet seq # */
	96	qp_attr.max_rd_atomic = 1; /* # of outstanding RDMA reads and
	97	* atomic ops allowed
	98	*/
	99	return ib_modify_qp(lnk->roce_qp, &qp_attr,
	100	IB_QP_STATE \| IB_QP_TIMEOUT \| IB_QP_RETRY_CNT \|
	101	IB_QP_SQ_PSN \| IB_QP_RNR_RETRY \|
	102	IB_QP_MAX_QP_RD_ATOMIC);
	103	}
	104
	105	int smc_ib_modify_qp_reset(struct smc_link *lnk)
	106	{
	107	struct ib_qp_attr qp_attr;
	108
	109	memset(&qp_attr, 0, sizeof(qp_attr));
	110	qp_attr.qp_state = IB_QPS_RESET;
	111	return ib_modify_qp(lnk->roce_qp, &qp_attr, IB_QP_STATE);
	112	}
	113
	114	int smc_ib_ready_link(struct smc_link *lnk)
	115	{
	116	struct smc_link_group *lgr =
	117	container_of(lnk, struct smc_link_group, lnk[0]);
	118	int rc = 0;
	119
	120	rc = smc_ib_modify_qp_init(lnk);
	121	if (rc)
	122	goto out;
	123
	124	rc = smc_ib_modify_qp_rtr(lnk);
	125	if (rc)
	126	goto out;
	127	smc_wr_remember_qp_attr(lnk);
	128	rc = ib_req_notify_cq(lnk->smcibdev->roce_cq_recv,
	129	IB_CQ_SOLICITED_MASK);
	130	if (rc)
	131	goto out;
	132	rc = smc_wr_rx_post_init(lnk);
	133	if (rc)
	134	goto out;
	135	smc_wr_remember_qp_attr(lnk);
	136
	137	if (lgr->role == SMC_SERV) {
	138	rc = smc_ib_modify_qp_rts(lnk);
139	if (rc)
140	goto out;
141	smc_wr_remember_qp_attr(lnk);
142	}
143	out:
144	return rc;
145	}
146
147	/* process context wrapper for might_sleep smc_ib_remember_port_attr */
148	static void smc_ib_port_event_work(struct work_struct *work)
149	{
150	struct smc_ib_device *smcibdev = container_of(
151	work, struct smc_ib_device, port_event_work);
152	u8 port_idx;
153
154	for_each_set_bit(port_idx, &smcibdev->port_event_mask, SMC_MAX_PORTS) {
155	smc_ib_remember_port_attr(smcibdev, port_idx + 1);
156	clear_bit(port_idx, &smcibdev->port_event_mask);
da05bf29	157	if (!smc_ib_port_active(smcibdev, port_idx + 1))
9fda3510	158	smc_port_terminate(smcibdev, port_idx + 1);
bd4ad577 UB	159	}
	160	}
	161
	162	/* can be called in IRQ context */
	163	static void smc_ib_global_event_handler(struct ib_event_handler *handler,
	164	struct ib_event *ibevent)
	165	{
	166	struct smc_ib_device *smcibdev;
	167	u8 port_idx;
	168
	169	smcibdev = container_of(handler, struct smc_ib_device, event_handler);
bd4ad577 UB	170
	171	switch (ibevent->event) {
	172	case IB_EVENT_PORT_ERR:
bd4ad577	173	case IB_EVENT_DEVICE_FATAL:
bd4ad577 UB	174	case IB_EVENT_PORT_ACTIVE:
	175	port_idx = ibevent->element.port_num - 1;
	176	set_bit(port_idx, &smcibdev->port_event_mask);
	177	schedule_work(&smcibdev->port_event_work);
	178	break;
	179	default:
	180	break;
	181	}
	182	}
	183
f38ba179 UB	184	void smc_ib_dealloc_protection_domain(struct smc_link *lnk)
f38ba179 UB	185	{
da05bf29 UB	186	if (lnk->roce_pd)
da05bf29 UB	187	ib_dealloc_pd(lnk->roce_pd);
f38ba179 UB	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	{
da05bf29 UB	204	struct smc_ib_device *smcibdev =
	205	(struct smc_ib_device *)ibevent->device;
	206	u8 port_idx;
	207
f38ba179 UB	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:
da05bf29 UB	213	port_idx = ibevent->element.port_num - 1;
	214	set_bit(port_idx, &smcibdev->port_event_mask);
	215	schedule_work(&smcibdev->port_event_work);
f38ba179 UB	216	break;
	217	default:
	218	break;
	219	}
	220	}
	221
	222	void smc_ib_destroy_queue_pair(struct smc_link *lnk)
	223	{
da05bf29 UB	224	if (lnk->roce_qp)
da05bf29 UB	225	ib_destroy_qp(lnk->roce_qp);
f38ba179 UB	226	lnk->roce_qp = NULL;
	227	}
	228
	229	/* create a queue pair within the protection domain for a link */
	230	int smc_ib_create_queue_pair(struct smc_link *lnk)
	231	{
	232	struct ib_qp_init_attr qp_attr = {
	233	.event_handler = smc_ib_qp_event_handler,
	234	.qp_context = lnk,
	235	.send_cq = lnk->smcibdev->roce_cq_send,
	236	.recv_cq = lnk->smcibdev->roce_cq_recv,
	237	.srq = NULL,
	238	.cap = {
f38ba179 UB	239	/* include unsolicited rdma_writes as well,
	240	* there are max. 2 RDMA_WRITE per 1 WR_SEND
	241	*/
652a1e41	242	.max_send_wr = SMC_WR_BUF_CNT * 3,
f38ba179 UB	243	.max_recv_wr = SMC_WR_BUF_CNT * 3,
	244	.max_send_sge = SMC_IB_MAX_SEND_SGE,
	245	.max_recv_sge = 1,
f38ba179 UB	246	},
	247	.sq_sig_type = IB_SIGNAL_REQ_WR,
	248	.qp_type = IB_QPT_RC,
	249	};
	250	int rc;
	251
	252	lnk->roce_qp = ib_create_qp(lnk->roce_pd, &qp_attr);
	253	rc = PTR_ERR_OR_ZERO(lnk->roce_qp);
	254	if (IS_ERR(lnk->roce_qp))
	255	lnk->roce_qp = NULL;
	256	else
	257	smc_wr_remember_qp_attr(lnk);
	258	return rc;
	259	}
	260
897e1c24 UB	261	void smc_ib_put_memory_region(struct ib_mr *mr)
	262	{
	263	ib_dereg_mr(mr);
	264	}
	265
	266	static int smc_ib_map_mr_sg(struct smc_buf_desc *buf_slot)
	267	{
	268	unsigned int offset = 0;
	269	int sg_num;
	270
	271	/* map the largest prefix of a dma mapped SG list */
	272	sg_num = ib_map_mr_sg(buf_slot->mr_rx[SMC_SINGLE_LINK],
	273	buf_slot->sgt[SMC_SINGLE_LINK].sgl,
	274	buf_slot->sgt[SMC_SINGLE_LINK].orig_nents,
	275	&offset, PAGE_SIZE);
	276
	277	return sg_num;
	278	}
	279
	280	/* Allocate a memory region and map the dma mapped SG list of buf_slot */
	281	int smc_ib_get_memory_region(struct ib_pd *pd, int access_flags,
	282	struct smc_buf_desc *buf_slot)
	283	{
	284	if (buf_slot->mr_rx[SMC_SINGLE_LINK])
	285	return 0; /* already done */
	286
	287	buf_slot->mr_rx[SMC_SINGLE_LINK] =
	288	ib_alloc_mr(pd, IB_MR_TYPE_MEM_REG, 1 << buf_slot->order);
	289	if (IS_ERR(buf_slot->mr_rx[SMC_SINGLE_LINK])) {
	290	int rc;
	291
	292	rc = PTR_ERR(buf_slot->mr_rx[SMC_SINGLE_LINK]);
	293	buf_slot->mr_rx[SMC_SINGLE_LINK] = NULL;
	294	return rc;
	295	}
	296
	297	if (smc_ib_map_mr_sg(buf_slot) != 1)
	298	return -EINVAL;
	299
	300	return 0;
	301	}
	302
10428dd8 UB	303	/* synchronize buffer usage for cpu access */
	304	void smc_ib_sync_sg_for_cpu(struct smc_ib_device *smcibdev,
	305	struct smc_buf_desc *buf_slot,
	306	enum dma_data_direction data_direction)
	307	{
	308	struct scatterlist *sg;
	309	unsigned int i;
	310
	311	/* for now there is just one DMA address */
	312	for_each_sg(buf_slot->sgt[SMC_SINGLE_LINK].sgl, sg,
	313	buf_slot->sgt[SMC_SINGLE_LINK].nents, i) {
	314	if (!sg_dma_len(sg))
	315	break;
	316	ib_dma_sync_single_for_cpu(smcibdev->ibdev,
	317	sg_dma_address(sg),
	318	sg_dma_len(sg),
	319	data_direction);
	320	}
	321	}
	322
	323	/* synchronize buffer usage for device access */
	324	void smc_ib_sync_sg_for_device(struct smc_ib_device *smcibdev,
	325	struct smc_buf_desc *buf_slot,
	326	enum dma_data_direction data_direction)
	327	{
	328	struct scatterlist *sg;
	329	unsigned int i;
	330
	331	/* for now there is just one DMA address */
	332	for_each_sg(buf_slot->sgt[SMC_SINGLE_LINK].sgl, sg,
	333	buf_slot->sgt[SMC_SINGLE_LINK].nents, i) {
	334	if (!sg_dma_len(sg))
	335	break;
	336	ib_dma_sync_single_for_device(smcibdev->ibdev,
	337	sg_dma_address(sg),
	338	sg_dma_len(sg),
	339	data_direction);
	340	}
	341	}
	342
a3fe3d01 UB	343	/* Map a new TX or RX buffer SG-table to DMA */
	344	int smc_ib_buf_map_sg(struct smc_ib_device *smcibdev,
	345	struct smc_buf_desc *buf_slot,
	346	enum dma_data_direction data_direction)
	347	{
	348	int mapped_nents;
	349
	350	mapped_nents = ib_dma_map_sg(smcibdev->ibdev,
	351	buf_slot->sgt[SMC_SINGLE_LINK].sgl,
	352	buf_slot->sgt[SMC_SINGLE_LINK].orig_nents,
	353	data_direction);
	354	if (!mapped_nents)
	355	return -ENOMEM;
	356
	357	return mapped_nents;
	358	}
	359
	360	void smc_ib_buf_unmap_sg(struct smc_ib_device *smcibdev,
	361	struct smc_buf_desc *buf_slot,
	362	enum dma_data_direction data_direction)
	363	{
	364	if (!buf_slot->sgt[SMC_SINGLE_LINK].sgl->dma_address)
	365	return; /* already unmapped */
	366
	367	ib_dma_unmap_sg(smcibdev->ibdev,
	368	buf_slot->sgt[SMC_SINGLE_LINK].sgl,
	369	buf_slot->sgt[SMC_SINGLE_LINK].orig_nents,
	370	data_direction);
	371	buf_slot->sgt[SMC_SINGLE_LINK].sgl->dma_address = 0;
	372	}
	373
a4cf0443 UB	374	static int smc_ib_fill_gid_and_mac(struct smc_ib_device *smcibdev, u8 ibport)
a4cf0443 UB	375	{
92f4e77c JG	376	struct ib_gid_attr gattr;
92f4e77c JG	377	int rc;
d921c420	378
92f4e77c JG	379	rc = ib_query_gid(smcibdev->ibdev, ibport, 0,
	380	&smcibdev->gid[ibport - 1], &gattr);
	381	if (rc \|\| !gattr.ndev)
	382	return -ENODEV;
	383
	384	memcpy(smcibdev->mac[ibport - 1], gattr.ndev->dev_addr, ETH_ALEN);
	385	dev_put(gattr.ndev);
	386	return 0;
a4cf0443 UB	387	}
	388
	389	/* Create an identifier unique for this instance of SMC-R.
	390	* The MAC-address of the first active registered IB device
	391	* plus a random 2-byte number is used to create this identifier.
	392	* This name is delivered to the peer during connection initialization.
	393	*/
	394	static inline void smc_ib_define_local_systemid(struct smc_ib_device *smcibdev,
	395	u8 ibport)
	396	{
	397	memcpy(&local_systemid[2], &smcibdev->mac[ibport - 1],
	398	sizeof(smcibdev->mac[ibport - 1]));
	399	get_random_bytes(&local_systemid[0], 2);
	400	}
	401
	402	bool smc_ib_port_active(struct smc_ib_device *smcibdev, u8 ibport)
	403	{
	404	return smcibdev->pattr[ibport - 1].state == IB_PORT_ACTIVE;
	405	}
	406
	407	int smc_ib_remember_port_attr(struct smc_ib_device *smcibdev, u8 ibport)
	408	{
	409	int rc;
	410
	411	memset(&smcibdev->pattr[ibport - 1], 0,
	412	sizeof(smcibdev->pattr[ibport - 1]));
	413	rc = ib_query_port(smcibdev->ibdev, ibport,
	414	&smcibdev->pattr[ibport - 1]);
	415	if (rc)
	416	goto out;
d921c420	417	/* the SMC protocol requires specification of the RoCE MAC address */
a4cf0443 UB	418	rc = smc_ib_fill_gid_and_mac(smcibdev, ibport);
	419	if (rc)
	420	goto out;
	421	if (!strncmp(local_systemid, SMC_LOCAL_SYSTEMID_RESET,
	422	sizeof(local_systemid)) &&
	423	smc_ib_port_active(smcibdev, ibport))
	424	/* create unique system identifier */
	425	smc_ib_define_local_systemid(smcibdev, ibport);
	426	out:
	427	return rc;
	428	}
	429
bd4ad577 UB	430	long smc_ib_setup_per_ibdev(struct smc_ib_device *smcibdev)
	431	{
	432	struct ib_cq_init_attr cqattr = {
c9f4c6cf UB	433	.cqe = SMC_MAX_CQE, .comp_vector = 0 };
c9f4c6cf UB	434	int cqe_size_order, smc_order;
bd4ad577 UB	435	long rc;
bd4ad577 UB	436
c9f4c6cf UB	437	/* the calculated number of cq entries fits to mlx5 cq allocation */
	438	cqe_size_order = cache_line_size() == 128 ? 7 : 6;
	439	smc_order = MAX_ORDER - cqe_size_order - 1;
	440	if (SMC_MAX_CQE + 2 > (0x00000001 << smc_order) * PAGE_SIZE)
	441	cqattr.cqe = (0x00000001 << smc_order) * PAGE_SIZE - 2;
bd4ad577 UB	442	smcibdev->roce_cq_send = ib_create_cq(smcibdev->ibdev,
	443	smc_wr_tx_cq_handler, NULL,
	444	smcibdev, &cqattr);
	445	rc = PTR_ERR_OR_ZERO(smcibdev->roce_cq_send);
	446	if (IS_ERR(smcibdev->roce_cq_send)) {
	447	smcibdev->roce_cq_send = NULL;
	448	return rc;
	449	}
	450	smcibdev->roce_cq_recv = ib_create_cq(smcibdev->ibdev,
	451	smc_wr_rx_cq_handler, NULL,
	452	smcibdev, &cqattr);
	453	rc = PTR_ERR_OR_ZERO(smcibdev->roce_cq_recv);
	454	if (IS_ERR(smcibdev->roce_cq_recv)) {
	455	smcibdev->roce_cq_recv = NULL;
	456	goto err;
	457	}
	458	INIT_IB_EVENT_HANDLER(&smcibdev->event_handler, smcibdev->ibdev,
	459	smc_ib_global_event_handler);
	460	ib_register_event_handler(&smcibdev->event_handler);
	461	smc_wr_add_dev(smcibdev);
	462	smcibdev->initialized = 1;
	463	return rc;
	464
	465	err:
	466	ib_destroy_cq(smcibdev->roce_cq_send);
	467	return rc;
	468	}
	469
	470	static void smc_ib_cleanup_per_ibdev(struct smc_ib_device *smcibdev)
	471	{
	472	if (!smcibdev->initialized)
	473	return;
da05bf29	474	smcibdev->initialized = 0;
bd4ad577 UB	475	smc_wr_remove_dev(smcibdev);
	476	ib_unregister_event_handler(&smcibdev->event_handler);
	477	ib_destroy_cq(smcibdev->roce_cq_recv);
	478	ib_destroy_cq(smcibdev->roce_cq_send);
	479	}
	480
a4cf0443 UB	481	static struct ib_client smc_ib_client;
	482
	483	/* callback function for ib_register_client() */
	484	static void smc_ib_add_dev(struct ib_device *ibdev)
	485	{
	486	struct smc_ib_device *smcibdev;
	487
	488	if (ibdev->node_type != RDMA_NODE_IB_CA)
	489	return;
	490
	491	smcibdev = kzalloc(sizeof(*smcibdev), GFP_KERNEL);
	492	if (!smcibdev)
	493	return;
	494
	495	smcibdev->ibdev = ibdev;
bd4ad577	496	INIT_WORK(&smcibdev->port_event_work, smc_ib_port_event_work);
a4cf0443 UB	497
	498	spin_lock(&smc_ib_devices.lock);
	499	list_add_tail(&smcibdev->list, &smc_ib_devices.list);
	500	spin_unlock(&smc_ib_devices.lock);
	501	ib_set_client_data(ibdev, &smc_ib_client, smcibdev);
	502	}
	503
	504	/* callback function for ib_register_client() */
	505	static void smc_ib_remove_dev(struct ib_device ibdev, void client_data)
	506	{
	507	struct smc_ib_device *smcibdev;
	508
	509	smcibdev = ib_get_client_data(ibdev, &smc_ib_client);
	510	ib_set_client_data(ibdev, &smc_ib_client, NULL);
	511	spin_lock(&smc_ib_devices.lock);
	512	list_del_init(&smcibdev->list); /* remove from smc_ib_devices */
	513	spin_unlock(&smc_ib_devices.lock);
6812baab	514	smc_pnet_remove_by_ibdev(smcibdev);
bd4ad577	515	smc_ib_cleanup_per_ibdev(smcibdev);
a4cf0443 UB	516	kfree(smcibdev);
	517	}
	518
	519	static struct ib_client smc_ib_client = {
	520	.name = "smc_ib",
	521	.add = smc_ib_add_dev,
	522	.remove = smc_ib_remove_dev,
	523	};
	524
	525	int __init smc_ib_register_client(void)
	526	{
	527	return ib_register_client(&smc_ib_client);
	528	}
	529
	530	void smc_ib_unregister_client(void)
	531	{
	532	ib_unregister_client(&smc_ib_client);
	533	}