[linux-2.6-block.git] / net / smc / smc_wr.c

// SPDX-License-Identifier: GPL-2.0
/*
 * Shared Memory Communications over RDMA (SMC-R) and RoCE
 *
 * Work Requests exploiting Infiniband API
 *
 * Work requests (WR) of type ib_post_send or ib_post_recv respectively
 * are submitted to either RC SQ or RC RQ respectively
 * (reliably connected send/receive queue)
 * and become work queue entries (WQEs).
 * While an SQ WR/WQE is pending, we track it until transmission completion.
 * Through a send or receive completion queue (CQ) respectively,
 * we get completion queue entries (CQEs) [aka work completions (WCs)].
 * Since the CQ callback is called from IRQ context, we split work by using
 * bottom halves implemented by tasklets.
 *
 * SMC uses this to exchange LLC (link layer control)
 * and CDC (connection data control) messages.
 *
 * Copyright IBM Corp. 2016
 *
 * Author(s):  Steffen Maier <maier@linux.vnet.ibm.com>
 */

#include <linux/atomic.h>
#include <linux/hashtable.h>
#include <linux/wait.h>
#include <rdma/ib_verbs.h>
#include <asm/div64.h>

#include "smc.h"
#include "smc_wr.h"

#define SMC_WR_MAX_POLL_CQE 10	/* max. # of compl. queue elements in 1 poll */

#define SMC_WR_RX_HASH_BITS 4
static DEFINE_HASHTABLE(smc_wr_rx_hash, SMC_WR_RX_HASH_BITS);
static DEFINE_SPINLOCK(smc_wr_rx_hash_lock);

struct smc_wr_tx_pend {	/* control data for a pending send request */
	u64			wr_id;		/* work request id sent */
	smc_wr_tx_handler	handler;
	enum ib_wc_status	wc_status;	/* CQE status */
	struct smc_link		*link;
	u32			idx;
	struct smc_wr_tx_pend_priv priv;
};

/******************************** send queue *********************************/

/*------------------------------- completion --------------------------------*/

static inline int smc_wr_tx_find_pending_index(struct smc_link *link, u64 wr_id)
{
	u32 i;

	for (i = 0; i < link->wr_tx_cnt; i++) {
		if (link->wr_tx_pends[i].wr_id == wr_id)
			return i;
	}
	return link->wr_tx_cnt;
}

static inline void smc_wr_tx_process_cqe(struct ib_wc *wc)
{
	struct smc_wr_tx_pend pnd_snd;
	struct smc_link *link;
	u32 pnd_snd_idx;
	int i;

	link = wc->qp->qp_context;

	if (wc->opcode == IB_WC_REG_MR) {
		if (wc->status)
			link->wr_reg_state = FAILED;
		else
			link->wr_reg_state = CONFIRMED;
		wake_up(&link->wr_reg_wait);
		return;
	}

	pnd_snd_idx = smc_wr_tx_find_pending_index(link, wc->wr_id);
	if (pnd_snd_idx == link->wr_tx_cnt)
		return;
	link->wr_tx_pends[pnd_snd_idx].wc_status = wc->status;
	memcpy(&pnd_snd, &link->wr_tx_pends[pnd_snd_idx], sizeof(pnd_snd));
	/* clear the full struct smc_wr_tx_pend including .priv */
	memset(&link->wr_tx_pends[pnd_snd_idx], 0,
	       sizeof(link->wr_tx_pends[pnd_snd_idx]));
	memset(&link->wr_tx_bufs[pnd_snd_idx], 0,
	       sizeof(link->wr_tx_bufs[pnd_snd_idx]));
	if (!test_and_clear_bit(pnd_snd_idx, link->wr_tx_mask))
		return;
	if (wc->status) {
		struct smc_link_group *lgr;

		for_each_set_bit(i, link->wr_tx_mask, link->wr_tx_cnt) {
			/* clear full struct smc_wr_tx_pend including .priv */
			memset(&link->wr_tx_pends[i], 0,
			       sizeof(link->wr_tx_pends[i]));
			memset(&link->wr_tx_bufs[i], 0,
			       sizeof(link->wr_tx_bufs[i]));
			clear_bit(i, link->wr_tx_mask);
		}
		/* terminate connections of this link group abnormally */
		lgr = container_of(link, struct smc_link_group,
				   lnk[SMC_SINGLE_LINK]);
		smc_lgr_terminate(lgr);
	}
	if (pnd_snd.handler)
		pnd_snd.handler(&pnd_snd.priv, link, wc->status);
	wake_up(&link->wr_tx_wait);
}

static void smc_wr_tx_tasklet_fn(unsigned long data)
{
	struct smc_ib_device *dev = (struct smc_ib_device *)data;
	struct ib_wc wc[SMC_WR_MAX_POLL_CQE];
	int i = 0, rc;
	int polled = 0;

again:
	polled++;
	do {
		memset(&wc, 0, sizeof(wc));
		rc = ib_poll_cq(dev->roce_cq_send, SMC_WR_MAX_POLL_CQE, wc);
		if (polled == 1) {
			ib_req_notify_cq(dev->roce_cq_send,
					 IB_CQ_NEXT_COMP |
					 IB_CQ_REPORT_MISSED_EVENTS);
		}
		if (!rc)
			break;
		for (i = 0; i < rc; i++)
			smc_wr_tx_process_cqe(&wc[i]);
	} while (rc > 0);
	if (polled == 1)
		goto again;
}

void smc_wr_tx_cq_handler(struct ib_cq *ib_cq, void *cq_context)
{
	struct smc_ib_device *dev = (struct smc_ib_device *)cq_context;

	tasklet_schedule(&dev->send_tasklet);
}

/*---------------------------- request submission ---------------------------*/

static inline int smc_wr_tx_get_free_slot_index(struct smc_link *link, u32 *idx)
{
	*idx = link->wr_tx_cnt;
	for_each_clear_bit(*idx, link->wr_tx_mask, link->wr_tx_cnt) {
		if (!test_and_set_bit(*idx, link->wr_tx_mask))
			return 0;
	}
	*idx = link->wr_tx_cnt;
	return -EBUSY;
}

/**
 * smc_wr_tx_get_free_slot() - returns buffer for message assembly,
 *			and sets info for pending transmit tracking
 * @link:		Pointer to smc_link used to later send the message.
 * @handler:		Send completion handler function pointer.
 * @wr_buf:		Out value returns pointer to message buffer.
 * @wr_pend_priv:	Out value returns pointer serving as handler context.
 *
 * Return: 0 on success, or -errno on error.
 */
int smc_wr_tx_get_free_slot(struct smc_link *link,
			    smc_wr_tx_handler handler,
			    struct smc_wr_buf **wr_buf,
			    struct smc_wr_tx_pend_priv **wr_pend_priv)
{
	struct smc_wr_tx_pend *wr_pend;
	u32 idx = link->wr_tx_cnt;
	struct ib_send_wr *wr_ib;
	u64 wr_id;
	int rc;

	*wr_buf = NULL;
	*wr_pend_priv = NULL;
	if (in_softirq()) {
		rc = smc_wr_tx_get_free_slot_index(link, &idx);
		if (rc)
			return rc;
	} else {
		struct smc_link_group *lgr;

		lgr = container_of(link, struct smc_link_group,
				   lnk[SMC_SINGLE_LINK]);
		rc = wait_event_timeout(
			link->wr_tx_wait,
			list_empty(&lgr->list) || /* lgr terminated */
			(smc_wr_tx_get_free_slot_index(link, &idx) != -EBUSY),
			SMC_WR_TX_WAIT_FREE_SLOT_TIME);
		if (!rc) {
			/* timeout - terminate connections */
			smc_lgr_terminate(lgr);
			return -EPIPE;
		}
		if (idx == link->wr_tx_cnt)
			return -EPIPE;
	}
	wr_id = smc_wr_tx_get_next_wr_id(link);
	wr_pend = &link->wr_tx_pends[idx];
	wr_pend->wr_id = wr_id;
	wr_pend->handler = handler;
	wr_pend->link = link;
	wr_pend->idx = idx;
	wr_ib = &link->wr_tx_ibs[idx];
	wr_ib->wr_id = wr_id;
	*wr_buf = &link->wr_tx_bufs[idx];
	*wr_pend_priv = &wr_pend->priv;
	return 0;
}

int smc_wr_tx_put_slot(struct smc_link *link,
		       struct smc_wr_tx_pend_priv *wr_pend_priv)
{
	struct smc_wr_tx_pend *pend;

	pend = container_of(wr_pend_priv, struct smc_wr_tx_pend, priv);
	if (pend->idx < link->wr_tx_cnt) {
		/* clear the full struct smc_wr_tx_pend including .priv */
		memset(&link->wr_tx_pends[pend->idx], 0,
		       sizeof(link->wr_tx_pends[pend->idx]));
		memset(&link->wr_tx_bufs[pend->idx], 0,
		       sizeof(link->wr_tx_bufs[pend->idx]));
		test_and_clear_bit(pend->idx, link->wr_tx_mask);
		return 1;
	}

	return 0;
}

/* Send prepared WR slot via ib_post_send.
 * @priv: pointer to smc_wr_tx_pend_priv identifying prepared message buffer
 */
int smc_wr_tx_send(struct smc_link *link, struct smc_wr_tx_pend_priv *priv)
{
	struct ib_send_wr *failed_wr = NULL;
	struct smc_wr_tx_pend *pend;
	int rc;

	ib_req_notify_cq(link->smcibdev->roce_cq_send,
			 IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS);
	pend = container_of(priv, struct smc_wr_tx_pend, priv);
	rc = ib_post_send(link->roce_qp, &link->wr_tx_ibs[pend->idx],
			  &failed_wr);
	if (rc) {
		struct smc_link_group *lgr =
			container_of(link, struct smc_link_group,
				     lnk[SMC_SINGLE_LINK]);

		smc_wr_tx_put_slot(link, priv);
		smc_lgr_terminate(lgr);
	}
	return rc;
}

/* Register a memory region and wait for result. */
int smc_wr_reg_send(struct smc_link *link, struct ib_mr *mr)
{
	struct ib_send_wr *failed_wr = NULL;
	int rc;

	ib_req_notify_cq(link->smcibdev->roce_cq_send,
			 IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS);
	link->wr_reg_state = POSTED;
	link->wr_reg.wr.wr_id = (u64)(uintptr_t)mr;
	link->wr_reg.mr = mr;
	link->wr_reg.key = mr->rkey;
	failed_wr = &link->wr_reg.wr;
	rc = ib_post_send(link->roce_qp, &link->wr_reg.wr, &failed_wr);
	WARN_ON(failed_wr != &link->wr_reg.wr);
	if (rc)
		return rc;

	rc = wait_event_interruptible_timeout(link->wr_reg_wait,
					      (link->wr_reg_state != POSTED),
					      SMC_WR_REG_MR_WAIT_TIME);
	if (!rc) {
		/* timeout - terminate connections */
		struct smc_link_group *lgr;

		lgr = container_of(link, struct smc_link_group,
				   lnk[SMC_SINGLE_LINK]);
		smc_lgr_terminate(lgr);
		return -EPIPE;
	}
	if (rc == -ERESTARTSYS)
		return -EINTR;
	switch (link->wr_reg_state) {
	case CONFIRMED:
		rc = 0;
		break;
	case FAILED:
		rc = -EIO;
		break;
	case POSTED:
		rc = -EPIPE;
		break;
	}
	return rc;
}

void smc_wr_tx_dismiss_slots(struct smc_link *link, u8 wr_tx_hdr_type,
			     smc_wr_tx_filter filter,
			     smc_wr_tx_dismisser dismisser,
			     unsigned long data)
{
	struct smc_wr_tx_pend_priv *tx_pend;
	struct smc_wr_rx_hdr *wr_tx;
	int i;

	for_each_set_bit(i, link->wr_tx_mask, link->wr_tx_cnt) {
		wr_tx = (struct smc_wr_rx_hdr *)&link->wr_tx_bufs[i];
		if (wr_tx->type != wr_tx_hdr_type)
			continue;
		tx_pend = &link->wr_tx_pends[i].priv;
		if (filter(tx_pend, data))
			dismisser(tx_pend);
	}
}

/****************************** receive queue ********************************/

int smc_wr_rx_register_handler(struct smc_wr_rx_handler *handler)
{
	struct smc_wr_rx_handler *h_iter;
	int rc = 0;

	spin_lock(&smc_wr_rx_hash_lock);
	hash_for_each_possible(smc_wr_rx_hash, h_iter, list, handler->type) {
		if (h_iter->type == handler->type) {
			rc = -EEXIST;
			goto out_unlock;
		}
	}
	hash_add(smc_wr_rx_hash, &handler->list, handler->type);
out_unlock:
	spin_unlock(&smc_wr_rx_hash_lock);
	return rc;
}

/* Demultiplex a received work request based on the message type to its handler.
 * Relies on smc_wr_rx_hash having been completely filled before any IB WRs,
 * and not being modified any more afterwards so we don't need to lock it.
 */
static inline void smc_wr_rx_demultiplex(struct ib_wc *wc)
{
	struct smc_link *link = (struct smc_link *)wc->qp->qp_context;
	struct smc_wr_rx_handler *handler;
	struct smc_wr_rx_hdr *wr_rx;
	u64 temp_wr_id;
	u32 index;

	if (wc->byte_len < sizeof(*wr_rx))
		return; /* short message */
	temp_wr_id = wc->wr_id;
	index = do_div(temp_wr_id, link->wr_rx_cnt);
	wr_rx = (struct smc_wr_rx_hdr *)&link->wr_rx_bufs[index];
	hash_for_each_possible(smc_wr_rx_hash, handler, list, wr_rx->type) {
		if (handler->type == wr_rx->type)
			handler->handler(wc, wr_rx);
	}
}

static inline void smc_wr_rx_process_cqes(struct ib_wc wc[], int num)
{
	struct smc_link *link;
	int i;

	for (i = 0; i < num; i++) {
		link = wc[i].qp->qp_context;
		if (wc[i].status == IB_WC_SUCCESS) {
			smc_wr_rx_demultiplex(&wc[i]);
			smc_wr_rx_post(link); /* refill WR RX */
		} else {
			struct smc_link_group *lgr;

			/* handle status errors */
			switch (wc[i].status) {
			case IB_WC_RETRY_EXC_ERR:
			case IB_WC_RNR_RETRY_EXC_ERR:
			case IB_WC_WR_FLUSH_ERR:
				/* terminate connections of this link group
				 * abnormally
				 */
				lgr = container_of(link, struct smc_link_group,
						   lnk[SMC_SINGLE_LINK]);
				smc_lgr_terminate(lgr);
				break;
			default:
				smc_wr_rx_post(link); /* refill WR RX */
				break;
			}
		}
	}
}

static void smc_wr_rx_tasklet_fn(unsigned long data)
{
	struct smc_ib_device *dev = (struct smc_ib_device *)data;
	struct ib_wc wc[SMC_WR_MAX_POLL_CQE];
	int polled = 0;
	int rc;

again:
	polled++;
	do {
		memset(&wc, 0, sizeof(wc));
		rc = ib_poll_cq(dev->roce_cq_recv, SMC_WR_MAX_POLL_CQE, wc);
		if (polled == 1) {
			ib_req_notify_cq(dev->roce_cq_recv,
					 IB_CQ_SOLICITED_MASK
					 | IB_CQ_REPORT_MISSED_EVENTS);
		}
		if (!rc)
			break;
		smc_wr_rx_process_cqes(&wc[0], rc);
	} while (rc > 0);
	if (polled == 1)
		goto again;
}

void smc_wr_rx_cq_handler(struct ib_cq *ib_cq, void *cq_context)
{
	struct smc_ib_device *dev = (struct smc_ib_device *)cq_context;

	tasklet_schedule(&dev->recv_tasklet);
}

int smc_wr_rx_post_init(struct smc_link *link)
{
	u32 i;
	int rc = 0;

	for (i = 0; i < link->wr_rx_cnt; i++)
		rc = smc_wr_rx_post(link);
	return rc;
}

/***************************** init, exit, misc ******************************/

void smc_wr_remember_qp_attr(struct smc_link *lnk)
{
	struct ib_qp_attr *attr = &lnk->qp_attr;
	struct ib_qp_init_attr init_attr;

	memset(attr, 0, sizeof(*attr));
	memset(&init_attr, 0, sizeof(init_attr));
	ib_query_qp(lnk->roce_qp, attr,
		    IB_QP_STATE |
		    IB_QP_CUR_STATE |
		    IB_QP_PKEY_INDEX |
		    IB_QP_PORT |
		    IB_QP_QKEY |
		    IB_QP_AV |
		    IB_QP_PATH_MTU |
		    IB_QP_TIMEOUT |
		    IB_QP_RETRY_CNT |
		    IB_QP_RNR_RETRY |
		    IB_QP_RQ_PSN |
		    IB_QP_ALT_PATH |
		    IB_QP_MIN_RNR_TIMER |
		    IB_QP_SQ_PSN |
		    IB_QP_PATH_MIG_STATE |
		    IB_QP_CAP |
		    IB_QP_DEST_QPN,
		    &init_attr);

	lnk->wr_tx_cnt = min_t(size_t, SMC_WR_BUF_CNT,
			       lnk->qp_attr.cap.max_send_wr);
	lnk->wr_rx_cnt = min_t(size_t, SMC_WR_BUF_CNT * 3,
			       lnk->qp_attr.cap.max_recv_wr);
}

static void smc_wr_init_sge(struct smc_link *lnk)
{
	u32 i;

	for (i = 0; i < lnk->wr_tx_cnt; i++) {
		lnk->wr_tx_sges[i].addr =
			lnk->wr_tx_dma_addr + i * SMC_WR_BUF_SIZE;
		lnk->wr_tx_sges[i].length = SMC_WR_TX_SIZE;
		lnk->wr_tx_sges[i].lkey = lnk->roce_pd->local_dma_lkey;
		lnk->wr_tx_ibs[i].next = NULL;
		lnk->wr_tx_ibs[i].sg_list = &lnk->wr_tx_sges[i];
		lnk->wr_tx_ibs[i].num_sge = 1;
		lnk->wr_tx_ibs[i].opcode = IB_WR_SEND;
		lnk->wr_tx_ibs[i].send_flags =
			IB_SEND_SIGNALED | IB_SEND_SOLICITED;
	}
	for (i = 0; i < lnk->wr_rx_cnt; i++) {
		lnk->wr_rx_sges[i].addr =
			lnk->wr_rx_dma_addr + i * SMC_WR_BUF_SIZE;
		lnk->wr_rx_sges[i].length = SMC_WR_BUF_SIZE;
		lnk->wr_rx_sges[i].lkey = lnk->roce_pd->local_dma_lkey;
		lnk->wr_rx_ibs[i].next = NULL;
		lnk->wr_rx_ibs[i].sg_list = &lnk->wr_rx_sges[i];
		lnk->wr_rx_ibs[i].num_sge = 1;
	}
	lnk->wr_reg.wr.next = NULL;
	lnk->wr_reg.wr.num_sge = 0;
	lnk->wr_reg.wr.send_flags = IB_SEND_SIGNALED;
	lnk->wr_reg.wr.opcode = IB_WR_REG_MR;
	lnk->wr_reg.access = IB_ACCESS_LOCAL_WRITE | IB_ACCESS_REMOTE_WRITE;
}

void smc_wr_free_link(struct smc_link *lnk)
{
	struct ib_device *ibdev;

	memset(lnk->wr_tx_mask, 0,
	       BITS_TO_LONGS(SMC_WR_BUF_CNT) * sizeof(*lnk->wr_tx_mask));

	if (!lnk->smcibdev)
		return;
	ibdev = lnk->smcibdev->ibdev;

	if (lnk->wr_rx_dma_addr) {
		ib_dma_unmap_single(ibdev, lnk->wr_rx_dma_addr,
				    SMC_WR_BUF_SIZE * lnk->wr_rx_cnt,
				    DMA_FROM_DEVICE);
		lnk->wr_rx_dma_addr = 0;
	}
	if (lnk->wr_tx_dma_addr) {
		ib_dma_unmap_single(ibdev, lnk->wr_tx_dma_addr,
				    SMC_WR_BUF_SIZE * lnk->wr_tx_cnt,
				    DMA_TO_DEVICE);
		lnk->wr_tx_dma_addr = 0;
	}
}

void smc_wr_free_link_mem(struct smc_link *lnk)
{
	kfree(lnk->wr_tx_pends);
	lnk->wr_tx_pends = NULL;
	kfree(lnk->wr_tx_mask);
	lnk->wr_tx_mask = NULL;
	kfree(lnk->wr_tx_sges);
	lnk->wr_tx_sges = NULL;
	kfree(lnk->wr_rx_sges);
	lnk->wr_rx_sges = NULL;
	kfree(lnk->wr_rx_ibs);
	lnk->wr_rx_ibs = NULL;
	kfree(lnk->wr_tx_ibs);
	lnk->wr_tx_ibs = NULL;
	kfree(lnk->wr_tx_bufs);
	lnk->wr_tx_bufs = NULL;
	kfree(lnk->wr_rx_bufs);
	lnk->wr_rx_bufs = NULL;
}

int smc_wr_alloc_link_mem(struct smc_link *link)
{
	/* allocate link related memory */
	link->wr_tx_bufs = kcalloc(SMC_WR_BUF_CNT, SMC_WR_BUF_SIZE, GFP_KERNEL);
	if (!link->wr_tx_bufs)
		goto no_mem;
	link->wr_rx_bufs = kcalloc(SMC_WR_BUF_CNT * 3, SMC_WR_BUF_SIZE,
				   GFP_KERNEL);
	if (!link->wr_rx_bufs)
		goto no_mem_wr_tx_bufs;
	link->wr_tx_ibs = kcalloc(SMC_WR_BUF_CNT, sizeof(link->wr_tx_ibs[0]),
				  GFP_KERNEL);
	if (!link->wr_tx_ibs)
		goto no_mem_wr_rx_bufs;
	link->wr_rx_ibs = kcalloc(SMC_WR_BUF_CNT * 3,
				  sizeof(link->wr_rx_ibs[0]),
				  GFP_KERNEL);
	if (!link->wr_rx_ibs)
		goto no_mem_wr_tx_ibs;
	link->wr_tx_sges = kcalloc(SMC_WR_BUF_CNT, sizeof(link->wr_tx_sges[0]),
				   GFP_KERNEL);
	if (!link->wr_tx_sges)
		goto no_mem_wr_rx_ibs;
	link->wr_rx_sges = kcalloc(SMC_WR_BUF_CNT * 3,
				   sizeof(link->wr_rx_sges[0]),
				   GFP_KERNEL);
	if (!link->wr_rx_sges)
		goto no_mem_wr_tx_sges;
	link->wr_tx_mask = kzalloc(
		BITS_TO_LONGS(SMC_WR_BUF_CNT) * sizeof(*link->wr_tx_mask),
		GFP_KERNEL);
	if (!link->wr_tx_mask)
		goto no_mem_wr_rx_sges;
	link->wr_tx_pends = kcalloc(SMC_WR_BUF_CNT,
				    sizeof(link->wr_tx_pends[0]),
				    GFP_KERNEL);
	if (!link->wr_tx_pends)
		goto no_mem_wr_tx_mask;
	return 0;

no_mem_wr_tx_mask:
	kfree(link->wr_tx_mask);
no_mem_wr_rx_sges:
	kfree(link->wr_rx_sges);
no_mem_wr_tx_sges:
	kfree(link->wr_tx_sges);
no_mem_wr_rx_ibs:
	kfree(link->wr_rx_ibs);
no_mem_wr_tx_ibs:
	kfree(link->wr_tx_ibs);
no_mem_wr_rx_bufs:
	kfree(link->wr_rx_bufs);
no_mem_wr_tx_bufs:
	kfree(link->wr_tx_bufs);
no_mem:
	return -ENOMEM;
}

void smc_wr_remove_dev(struct smc_ib_device *smcibdev)
{
	tasklet_kill(&smcibdev->recv_tasklet);
	tasklet_kill(&smcibdev->send_tasklet);
}

void smc_wr_add_dev(struct smc_ib_device *smcibdev)
{
	tasklet_init(&smcibdev->recv_tasklet, smc_wr_rx_tasklet_fn,
		     (unsigned long)smcibdev);
	tasklet_init(&smcibdev->send_tasklet, smc_wr_tx_tasklet_fn,
		     (unsigned long)smcibdev);
}

int smc_wr_create_link(struct smc_link *lnk)
{
	struct ib_device *ibdev = lnk->smcibdev->ibdev;
	int rc = 0;

	smc_wr_tx_set_wr_id(&lnk->wr_tx_id, 0);
	lnk->wr_rx_id = 0;
	lnk->wr_rx_dma_addr = ib_dma_map_single(
		ibdev, lnk->wr_rx_bufs,	SMC_WR_BUF_SIZE * lnk->wr_rx_cnt,
		DMA_FROM_DEVICE);
	if (ib_dma_mapping_error(ibdev, lnk->wr_rx_dma_addr)) {
		lnk->wr_rx_dma_addr = 0;
		rc = -EIO;
		goto out;
	}
	lnk->wr_tx_dma_addr = ib_dma_map_single(
		ibdev, lnk->wr_tx_bufs,	SMC_WR_BUF_SIZE * lnk->wr_tx_cnt,
		DMA_TO_DEVICE);
	if (ib_dma_mapping_error(ibdev, lnk->wr_tx_dma_addr)) {
		rc = -EIO;
		goto dma_unmap;
	}
	smc_wr_init_sge(lnk);
	memset(lnk->wr_tx_mask, 0,
	       BITS_TO_LONGS(SMC_WR_BUF_CNT) * sizeof(*lnk->wr_tx_mask));
	init_waitqueue_head(&lnk->wr_tx_wait);
	init_waitqueue_head(&lnk->wr_reg_wait);
	return rc;

dma_unmap:
	ib_dma_unmap_single(ibdev, lnk->wr_rx_dma_addr,
			    SMC_WR_BUF_SIZE * lnk->wr_rx_cnt,
			    DMA_FROM_DEVICE);
	lnk->wr_rx_dma_addr = 0;
out:
	return rc;
}
Commit	Line	Data
b2441318	1	// SPDX-License-Identifier: GPL-2.0
f38ba179 UB	2	/*
	3	* Shared Memory Communications over RDMA (SMC-R) and RoCE
	4	*
	5	* Work Requests exploiting Infiniband API
	6	*
	7	* Work requests (WR) of type ib_post_send or ib_post_recv respectively
	8	* are submitted to either RC SQ or RC RQ respectively
	9	* (reliably connected send/receive queue)
	10	* and become work queue entries (WQEs).
	11	* While an SQ WR/WQE is pending, we track it until transmission completion.
	12	* Through a send or receive completion queue (CQ) respectively,
	13	* we get completion queue entries (CQEs) [aka work completions (WCs)].
	14	* Since the CQ callback is called from IRQ context, we split work by using
	15	* bottom halves implemented by tasklets.
	16	*
	17	* SMC uses this to exchange LLC (link layer control)
	18	* and CDC (connection data control) messages.
	19	*
	20	* Copyright IBM Corp. 2016
	21	*
	22	* Author(s): Steffen Maier <maier@linux.vnet.ibm.com>
	23	*/
	24
	25	#include <linux/atomic.h>
	26	#include <linux/hashtable.h>
	27	#include <linux/wait.h>
	28	#include <rdma/ib_verbs.h>
	29	#include <asm/div64.h>
	30
	31	#include "smc.h"
	32	#include "smc_wr.h"
	33
	34	#define SMC_WR_MAX_POLL_CQE 10 /* max. # of compl. queue elements in 1 poll */
	35
	36	#define SMC_WR_RX_HASH_BITS 4
	37	static DEFINE_HASHTABLE(smc_wr_rx_hash, SMC_WR_RX_HASH_BITS);
	38	static DEFINE_SPINLOCK(smc_wr_rx_hash_lock);
	39
	40	struct smc_wr_tx_pend { /* control data for a pending send request */
	41	u64 wr_id; /* work request id sent */
	42	smc_wr_tx_handler handler;
	43	enum ib_wc_status wc_status; /* CQE status */
	44	struct smc_link *link;
	45	u32 idx;
	46	struct smc_wr_tx_pend_priv priv;
	47	};
	48
	49	/****************************** send queue *******************************/
	50
	51	/------------------------------- completion --------------------------------/
	52
	53	static inline int smc_wr_tx_find_pending_index(struct smc_link *link, u64 wr_id)
	54	{
	55	u32 i;
	56
	57	for (i = 0; i < link->wr_tx_cnt; i++) {
	58	if (link->wr_tx_pends[i].wr_id == wr_id)
	59	return i;
	60	}
	61	return link->wr_tx_cnt;
	62	}
	63
	64	static inline void smc_wr_tx_process_cqe(struct ib_wc *wc)
	65	{
66	struct smc_wr_tx_pend pnd_snd;
67	struct smc_link *link;
68	u32 pnd_snd_idx;
69	int i;
70
71	link = wc->qp->qp_context;
652a1e41 UB	72
	73	if (wc->opcode == IB_WC_REG_MR) {
	74	if (wc->status)
	75	link->wr_reg_state = FAILED;
	76	else
	77	link->wr_reg_state = CONFIRMED;
	78	wake_up(&link->wr_reg_wait);
	79	return;
	80	}
	81
f38ba179 UB	82	pnd_snd_idx = smc_wr_tx_find_pending_index(link, wc->wr_id);
	83	if (pnd_snd_idx == link->wr_tx_cnt)
	84	return;
	85	link->wr_tx_pends[pnd_snd_idx].wc_status = wc->status;
	86	memcpy(&pnd_snd, &link->wr_tx_pends[pnd_snd_idx], sizeof(pnd_snd));
	87	/* clear the full struct smc_wr_tx_pend including .priv */
	88	memset(&link->wr_tx_pends[pnd_snd_idx], 0,
	89	sizeof(link->wr_tx_pends[pnd_snd_idx]));
	90	memset(&link->wr_tx_bufs[pnd_snd_idx], 0,
	91	sizeof(link->wr_tx_bufs[pnd_snd_idx]));
	92	if (!test_and_clear_bit(pnd_snd_idx, link->wr_tx_mask))
	93	return;
	94	if (wc->status) {
b38d7324 UB	95	struct smc_link_group *lgr;
b38d7324 UB	96
f38ba179 UB	97	for_each_set_bit(i, link->wr_tx_mask, link->wr_tx_cnt) {
	98	/* clear full struct smc_wr_tx_pend including .priv */
	99	memset(&link->wr_tx_pends[i], 0,
	100	sizeof(link->wr_tx_pends[i]));
	101	memset(&link->wr_tx_bufs[i], 0,
	102	sizeof(link->wr_tx_bufs[i]));
	103	clear_bit(i, link->wr_tx_mask);
	104	}
b38d7324 UB	105	/* terminate connections of this link group abnormally */
	106	lgr = container_of(link, struct smc_link_group,
	107	lnk[SMC_SINGLE_LINK]);
	108	smc_lgr_terminate(lgr);
f38ba179 UB	109	}
	110	if (pnd_snd.handler)
	111	pnd_snd.handler(&pnd_snd.priv, link, wc->status);
	112	wake_up(&link->wr_tx_wait);
	113	}
	114
	115	static void smc_wr_tx_tasklet_fn(unsigned long data)
	116	{
	117	struct smc_ib_device dev = (struct smc_ib_device )data;
	118	struct ib_wc wc[SMC_WR_MAX_POLL_CQE];
	119	int i = 0, rc;
	120	int polled = 0;
	121
	122	again:
	123	polled++;
	124	do {
86e780d3	125	memset(&wc, 0, sizeof(wc));
f38ba179 UB	126	rc = ib_poll_cq(dev->roce_cq_send, SMC_WR_MAX_POLL_CQE, wc);
	127	if (polled == 1) {
	128	ib_req_notify_cq(dev->roce_cq_send,
	129	IB_CQ_NEXT_COMP \|
	130	IB_CQ_REPORT_MISSED_EVENTS);
	131	}
	132	if (!rc)
	133	break;
	134	for (i = 0; i < rc; i++)
	135	smc_wr_tx_process_cqe(&wc[i]);
	136	} while (rc > 0);
	137	if (polled == 1)
	138	goto again;
	139	}
	140
	141	void smc_wr_tx_cq_handler(struct ib_cq ib_cq, void cq_context)
	142	{
	143	struct smc_ib_device dev = (struct smc_ib_device )cq_context;
	144
	145	tasklet_schedule(&dev->send_tasklet);
	146	}
	147
	148	/---------------------------- request submission ---------------------------/
	149
	150	static inline int smc_wr_tx_get_free_slot_index(struct smc_link link, u32 idx)
	151	{
	152	*idx = link->wr_tx_cnt;
	153	for_each_clear_bit(*idx, link->wr_tx_mask, link->wr_tx_cnt) {
	154	if (!test_and_set_bit(*idx, link->wr_tx_mask))
	155	return 0;
	156	}
	157	*idx = link->wr_tx_cnt;
	158	return -EBUSY;
	159	}
	160
	161	/**
	162	* smc_wr_tx_get_free_slot() - returns buffer for message assembly,
	163	* and sets info for pending transmit tracking
	164	* @link: Pointer to smc_link used to later send the message.
	165	* @handler: Send completion handler function pointer.
	166	* @wr_buf: Out value returns pointer to message buffer.
	167	* @wr_pend_priv: Out value returns pointer serving as handler context.
	168	*
	169	* Return: 0 on success, or -errno on error.
	170	*/
	171	int smc_wr_tx_get_free_slot(struct smc_link *link,
	172	smc_wr_tx_handler handler,
	173	struct smc_wr_buf **wr_buf,
	174	struct smc_wr_tx_pend_priv **wr_pend_priv)
	175	{
	176	struct smc_wr_tx_pend *wr_pend;
1a0a04c7	177	u32 idx = link->wr_tx_cnt;
f38ba179 UB	178	struct ib_send_wr *wr_ib;
f38ba179 UB	179	u64 wr_id;
f38ba179 UB	180	int rc;
	181
	182	*wr_buf = NULL;
	183	*wr_pend_priv = NULL;
	184	if (in_softirq()) {
	185	rc = smc_wr_tx_get_free_slot_index(link, &idx);
	186	if (rc)
	187	return rc;
	188	} else {
1a0a04c7 UB	189	struct smc_link_group *lgr;
	190
	191	lgr = container_of(link, struct smc_link_group,
	192	lnk[SMC_SINGLE_LINK]);
86e780d3	193	rc = wait_event_timeout(
f38ba179	194	link->wr_tx_wait,
1a0a04c7	195	list_empty(&lgr->list) \|\| /* lgr terminated */
f38ba179 UB	196	(smc_wr_tx_get_free_slot_index(link, &idx) != -EBUSY),
	197	SMC_WR_TX_WAIT_FREE_SLOT_TIME);
	198	if (!rc) {
b38d7324	199	/* timeout - terminate connections */
b38d7324	200	smc_lgr_terminate(lgr);
f38ba179 UB	201	return -EPIPE;
f38ba179 UB	202	}
f38ba179 UB	203	if (idx == link->wr_tx_cnt)
	204	return -EPIPE;
	205	}
	206	wr_id = smc_wr_tx_get_next_wr_id(link);
	207	wr_pend = &link->wr_tx_pends[idx];
	208	wr_pend->wr_id = wr_id;
	209	wr_pend->handler = handler;
	210	wr_pend->link = link;
	211	wr_pend->idx = idx;
	212	wr_ib = &link->wr_tx_ibs[idx];
	213	wr_ib->wr_id = wr_id;
	214	*wr_buf = &link->wr_tx_bufs[idx];
	215	*wr_pend_priv = &wr_pend->priv;
	216	return 0;
	217	}
	218
	219	int smc_wr_tx_put_slot(struct smc_link *link,
	220	struct smc_wr_tx_pend_priv *wr_pend_priv)
	221	{
	222	struct smc_wr_tx_pend *pend;
	223
	224	pend = container_of(wr_pend_priv, struct smc_wr_tx_pend, priv);
	225	if (pend->idx < link->wr_tx_cnt) {
	226	/* clear the full struct smc_wr_tx_pend including .priv */
	227	memset(&link->wr_tx_pends[pend->idx], 0,
	228	sizeof(link->wr_tx_pends[pend->idx]));
	229	memset(&link->wr_tx_bufs[pend->idx], 0,
	230	sizeof(link->wr_tx_bufs[pend->idx]));
	231	test_and_clear_bit(pend->idx, link->wr_tx_mask);
	232	return 1;
	233	}
	234
	235	return 0;
	236	}
	237
	238	/* Send prepared WR slot via ib_post_send.
	239	* @priv: pointer to smc_wr_tx_pend_priv identifying prepared message buffer
	240	*/
	241	int smc_wr_tx_send(struct smc_link link, struct smc_wr_tx_pend_priv priv)
	242	{
	243	struct ib_send_wr *failed_wr = NULL;
	244	struct smc_wr_tx_pend *pend;
	245	int rc;
	246
	247	ib_req_notify_cq(link->smcibdev->roce_cq_send,
8301fa44	248	IB_CQ_NEXT_COMP \| IB_CQ_REPORT_MISSED_EVENTS);
f38ba179 UB	249	pend = container_of(priv, struct smc_wr_tx_pend, priv);
	250	rc = ib_post_send(link->roce_qp, &link->wr_tx_ibs[pend->idx],
	251	&failed_wr);
b4772b3a UB	252	if (rc) {
	253	struct smc_link_group *lgr =
	254	container_of(link, struct smc_link_group,
	255	lnk[SMC_SINGLE_LINK]);
	256
f38ba179	257	smc_wr_tx_put_slot(link, priv);
b4772b3a UB	258	smc_lgr_terminate(lgr);
b4772b3a UB	259	}
f38ba179 UB	260	return rc;
	261	}
	262
652a1e41 UB	263	/* Register a memory region and wait for result. */
	264	int smc_wr_reg_send(struct smc_link link, struct ib_mr mr)
	265	{
	266	struct ib_send_wr *failed_wr = NULL;
	267	int rc;
	268
	269	ib_req_notify_cq(link->smcibdev->roce_cq_send,
	270	IB_CQ_NEXT_COMP \| IB_CQ_REPORT_MISSED_EVENTS);
	271	link->wr_reg_state = POSTED;
	272	link->wr_reg.wr.wr_id = (u64)(uintptr_t)mr;
	273	link->wr_reg.mr = mr;
	274	link->wr_reg.key = mr->rkey;
	275	failed_wr = &link->wr_reg.wr;
	276	rc = ib_post_send(link->roce_qp, &link->wr_reg.wr, &failed_wr);
	277	WARN_ON(failed_wr != &link->wr_reg.wr);
	278	if (rc)
	279	return rc;
	280
	281	rc = wait_event_interruptible_timeout(link->wr_reg_wait,
	282	(link->wr_reg_state != POSTED),
	283	SMC_WR_REG_MR_WAIT_TIME);
	284	if (!rc) {
	285	/* timeout - terminate connections */
	286	struct smc_link_group *lgr;
	287
	288	lgr = container_of(link, struct smc_link_group,
	289	lnk[SMC_SINGLE_LINK]);
	290	smc_lgr_terminate(lgr);
	291	return -EPIPE;
	292	}
	293	if (rc == -ERESTARTSYS)
	294	return -EINTR;
	295	switch (link->wr_reg_state) {
	296	case CONFIRMED:
	297	rc = 0;
	298	break;
	299	case FAILED:
	300	rc = -EIO;
	301	break;
	302	case POSTED:
	303	rc = -EPIPE;
	304	break;
	305	}
	306	return rc;
	307	}
	308
86e780d3	309	void smc_wr_tx_dismiss_slots(struct smc_link *link, u8 wr_tx_hdr_type,
5f08318f UB	310	smc_wr_tx_filter filter,
	311	smc_wr_tx_dismisser dismisser,
	312	unsigned long data)
	313	{
	314	struct smc_wr_tx_pend_priv *tx_pend;
86e780d3	315	struct smc_wr_rx_hdr *wr_tx;
5f08318f UB	316	int i;
	317
	318	for_each_set_bit(i, link->wr_tx_mask, link->wr_tx_cnt) {
86e780d3 UB	319	wr_tx = (struct smc_wr_rx_hdr *)&link->wr_tx_bufs[i];
86e780d3 UB	320	if (wr_tx->type != wr_tx_hdr_type)
5f08318f UB	321	continue;
	322	tx_pend = &link->wr_tx_pends[i].priv;
	323	if (filter(tx_pend, data))
	324	dismisser(tx_pend);
	325	}
	326	}
	327
f38ba179 UB	328	/**************************** receive queue ******************************/
	329
	330	int smc_wr_rx_register_handler(struct smc_wr_rx_handler *handler)
	331	{
	332	struct smc_wr_rx_handler *h_iter;
	333	int rc = 0;
	334
	335	spin_lock(&smc_wr_rx_hash_lock);
	336	hash_for_each_possible(smc_wr_rx_hash, h_iter, list, handler->type) {
	337	if (h_iter->type == handler->type) {
	338	rc = -EEXIST;
	339	goto out_unlock;
	340	}
	341	}
	342	hash_add(smc_wr_rx_hash, &handler->list, handler->type);
	343	out_unlock:
	344	spin_unlock(&smc_wr_rx_hash_lock);
	345	return rc;
	346	}
	347
	348	/* Demultiplex a received work request based on the message type to its handler.
	349	* Relies on smc_wr_rx_hash having been completely filled before any IB WRs,
	350	* and not being modified any more afterwards so we don't need to lock it.
	351	*/
	352	static inline void smc_wr_rx_demultiplex(struct ib_wc *wc)
	353	{
	354	struct smc_link link = (struct smc_link )wc->qp->qp_context;
	355	struct smc_wr_rx_handler *handler;
	356	struct smc_wr_rx_hdr *wr_rx;
	357	u64 temp_wr_id;
	358	u32 index;
	359
	360	if (wc->byte_len < sizeof(*wr_rx))
	361	return; /* short message */
	362	temp_wr_id = wc->wr_id;
	363	index = do_div(temp_wr_id, link->wr_rx_cnt);
	364	wr_rx = (struct smc_wr_rx_hdr *)&link->wr_rx_bufs[index];
	365	hash_for_each_possible(smc_wr_rx_hash, handler, list, wr_rx->type) {
	366	if (handler->type == wr_rx->type)
	367	handler->handler(wc, wr_rx);
	368	}
	369	}
	370
	371	static inline void smc_wr_rx_process_cqes(struct ib_wc wc[], int num)
	372	{
	373	struct smc_link *link;
	374	int i;
	375
	376	for (i = 0; i < num; i++) {
	377	link = wc[i].qp->qp_context;
	378	if (wc[i].status == IB_WC_SUCCESS) {
	379	smc_wr_rx_demultiplex(&wc[i]);
	380	smc_wr_rx_post(link); /* refill WR RX */
	381	} else {
b38d7324 UB	382	struct smc_link_group *lgr;
b38d7324 UB	383
f38ba179 UB	384	/* handle status errors */
	385	switch (wc[i].status) {
	386	case IB_WC_RETRY_EXC_ERR:
	387	case IB_WC_RNR_RETRY_EXC_ERR:
	388	case IB_WC_WR_FLUSH_ERR:
b38d7324 UB	389	/* terminate connections of this link group
	390	* abnormally
	391	*/
	392	lgr = container_of(link, struct smc_link_group,
	393	lnk[SMC_SINGLE_LINK]);
	394	smc_lgr_terminate(lgr);
f38ba179 UB	395	break;
	396	default:
	397	smc_wr_rx_post(link); /* refill WR RX */
	398	break;
	399	}
	400	}
	401	}
	402	}
	403
	404	static void smc_wr_rx_tasklet_fn(unsigned long data)
	405	{
	406	struct smc_ib_device dev = (struct smc_ib_device )data;
	407	struct ib_wc wc[SMC_WR_MAX_POLL_CQE];
	408	int polled = 0;
	409	int rc;
	410
	411	again:
	412	polled++;
	413	do {
	414	memset(&wc, 0, sizeof(wc));
	415	rc = ib_poll_cq(dev->roce_cq_recv, SMC_WR_MAX_POLL_CQE, wc);
	416	if (polled == 1) {
	417	ib_req_notify_cq(dev->roce_cq_recv,
	418	IB_CQ_SOLICITED_MASK
	419	\| IB_CQ_REPORT_MISSED_EVENTS);
	420	}
	421	if (!rc)
	422	break;
	423	smc_wr_rx_process_cqes(&wc[0], rc);
	424	} while (rc > 0);
	425	if (polled == 1)
	426	goto again;
	427	}
	428
	429	void smc_wr_rx_cq_handler(struct ib_cq ib_cq, void cq_context)
	430	{
	431	struct smc_ib_device dev = (struct smc_ib_device )cq_context;
	432
	433	tasklet_schedule(&dev->recv_tasklet);
	434	}
	435
	436	int smc_wr_rx_post_init(struct smc_link *link)
	437	{
	438	u32 i;
	439	int rc = 0;
	440
	441	for (i = 0; i < link->wr_rx_cnt; i++)
	442	rc = smc_wr_rx_post(link);
	443	return rc;
	444	}
	445
	446	/*************************** init, exit, misc ****************************/
	447
	448	void smc_wr_remember_qp_attr(struct smc_link *lnk)
	449	{
	450	struct ib_qp_attr *attr = &lnk->qp_attr;
	451	struct ib_qp_init_attr init_attr;
	452
	453	memset(attr, 0, sizeof(*attr));
	454	memset(&init_attr, 0, sizeof(init_attr));
	455	ib_query_qp(lnk->roce_qp, attr,
	456	IB_QP_STATE \|
	457	IB_QP_CUR_STATE \|
	458	IB_QP_PKEY_INDEX \|
459	IB_QP_PORT \|
460	IB_QP_QKEY \|
461	IB_QP_AV \|
462	IB_QP_PATH_MTU \|
463	IB_QP_TIMEOUT \|
464	IB_QP_RETRY_CNT \|
465	IB_QP_RNR_RETRY \|
466	IB_QP_RQ_PSN \|
467	IB_QP_ALT_PATH \|
468	IB_QP_MIN_RNR_TIMER \|
469	IB_QP_SQ_PSN \|
470	IB_QP_PATH_MIG_STATE \|
471	IB_QP_CAP \|
472	IB_QP_DEST_QPN,
473	&init_attr);
474
475	lnk->wr_tx_cnt = min_t(size_t, SMC_WR_BUF_CNT,
476	lnk->qp_attr.cap.max_send_wr);
477	lnk->wr_rx_cnt = min_t(size_t, SMC_WR_BUF_CNT * 3,
478	lnk->qp_attr.cap.max_recv_wr);
479	}
480
481	static void smc_wr_init_sge(struct smc_link *lnk)
482	{
483	u32 i;
484
485	for (i = 0; i < lnk->wr_tx_cnt; i++) {
486	lnk->wr_tx_sges[i].addr =
487	lnk->wr_tx_dma_addr + i * SMC_WR_BUF_SIZE;
488	lnk->wr_tx_sges[i].length = SMC_WR_TX_SIZE;
bd4ad577	489	lnk->wr_tx_sges[i].lkey = lnk->roce_pd->local_dma_lkey;
f38ba179 UB	490	lnk->wr_tx_ibs[i].next = NULL;
	491	lnk->wr_tx_ibs[i].sg_list = &lnk->wr_tx_sges[i];
	492	lnk->wr_tx_ibs[i].num_sge = 1;
	493	lnk->wr_tx_ibs[i].opcode = IB_WR_SEND;
	494	lnk->wr_tx_ibs[i].send_flags =
2c9c1682	495	IB_SEND_SIGNALED \| IB_SEND_SOLICITED;
f38ba179 UB	496	}
	497	for (i = 0; i < lnk->wr_rx_cnt; i++) {
	498	lnk->wr_rx_sges[i].addr =
	499	lnk->wr_rx_dma_addr + i * SMC_WR_BUF_SIZE;
	500	lnk->wr_rx_sges[i].length = SMC_WR_BUF_SIZE;
bd4ad577	501	lnk->wr_rx_sges[i].lkey = lnk->roce_pd->local_dma_lkey;
f38ba179 UB	502	lnk->wr_rx_ibs[i].next = NULL;
	503	lnk->wr_rx_ibs[i].sg_list = &lnk->wr_rx_sges[i];
	504	lnk->wr_rx_ibs[i].num_sge = 1;
	505	}
652a1e41 UB	506	lnk->wr_reg.wr.next = NULL;
	507	lnk->wr_reg.wr.num_sge = 0;
	508	lnk->wr_reg.wr.send_flags = IB_SEND_SIGNALED;
	509	lnk->wr_reg.wr.opcode = IB_WR_REG_MR;
	510	lnk->wr_reg.access = IB_ACCESS_LOCAL_WRITE \| IB_ACCESS_REMOTE_WRITE;
f38ba179 UB	511	}
	512
	513	void smc_wr_free_link(struct smc_link *lnk)
	514	{
	515	struct ib_device *ibdev;
	516
	517	memset(lnk->wr_tx_mask, 0,
	518	BITS_TO_LONGS(SMC_WR_BUF_CNT) * sizeof(*lnk->wr_tx_mask));
	519
	520	if (!lnk->smcibdev)
	521	return;
	522	ibdev = lnk->smcibdev->ibdev;
	523
	524	if (lnk->wr_rx_dma_addr) {
	525	ib_dma_unmap_single(ibdev, lnk->wr_rx_dma_addr,
	526	SMC_WR_BUF_SIZE * lnk->wr_rx_cnt,
	527	DMA_FROM_DEVICE);
	528	lnk->wr_rx_dma_addr = 0;
	529	}
	530	if (lnk->wr_tx_dma_addr) {
	531	ib_dma_unmap_single(ibdev, lnk->wr_tx_dma_addr,
	532	SMC_WR_BUF_SIZE * lnk->wr_tx_cnt,
	533	DMA_TO_DEVICE);
	534	lnk->wr_tx_dma_addr = 0;
	535	}
	536	}
	537
	538	void smc_wr_free_link_mem(struct smc_link *lnk)
	539	{
	540	kfree(lnk->wr_tx_pends);
	541	lnk->wr_tx_pends = NULL;
	542	kfree(lnk->wr_tx_mask);
	543	lnk->wr_tx_mask = NULL;
	544	kfree(lnk->wr_tx_sges);
	545	lnk->wr_tx_sges = NULL;
	546	kfree(lnk->wr_rx_sges);
	547	lnk->wr_rx_sges = NULL;
	548	kfree(lnk->wr_rx_ibs);
	549	lnk->wr_rx_ibs = NULL;
	550	kfree(lnk->wr_tx_ibs);
	551	lnk->wr_tx_ibs = NULL;
	552	kfree(lnk->wr_tx_bufs);
	553	lnk->wr_tx_bufs = NULL;
	554	kfree(lnk->wr_rx_bufs);
	555	lnk->wr_rx_bufs = NULL;
	556	}
	557
	558	int smc_wr_alloc_link_mem(struct smc_link *link)
	559	{
	560	/* allocate link related memory */
	561	link->wr_tx_bufs = kcalloc(SMC_WR_BUF_CNT, SMC_WR_BUF_SIZE, GFP_KERNEL);
	562	if (!link->wr_tx_bufs)
	563	goto no_mem;
	564	link->wr_rx_bufs = kcalloc(SMC_WR_BUF_CNT * 3, SMC_WR_BUF_SIZE,
	565	GFP_KERNEL);
	566	if (!link->wr_rx_bufs)
	567	goto no_mem_wr_tx_bufs;
	568	link->wr_tx_ibs = kcalloc(SMC_WR_BUF_CNT, sizeof(link->wr_tx_ibs[0]),
	569	GFP_KERNEL);
	570	if (!link->wr_tx_ibs)
	571	goto no_mem_wr_rx_bufs;
	572	link->wr_rx_ibs = kcalloc(SMC_WR_BUF_CNT * 3,
	573	sizeof(link->wr_rx_ibs[0]),
	574	GFP_KERNEL);
575	if (!link->wr_rx_ibs)
576	goto no_mem_wr_tx_ibs;
577	link->wr_tx_sges = kcalloc(SMC_WR_BUF_CNT, sizeof(link->wr_tx_sges[0]),
578	GFP_KERNEL);
579	if (!link->wr_tx_sges)
580	goto no_mem_wr_rx_ibs;
581	link->wr_rx_sges = kcalloc(SMC_WR_BUF_CNT * 3,
582	sizeof(link->wr_rx_sges[0]),
583	GFP_KERNEL);
584	if (!link->wr_rx_sges)
585	goto no_mem_wr_tx_sges;
586	link->wr_tx_mask = kzalloc(
587	BITS_TO_LONGS(SMC_WR_BUF_CNT) * sizeof(*link->wr_tx_mask),
588	GFP_KERNEL);
589	if (!link->wr_tx_mask)
590	goto no_mem_wr_rx_sges;
591	link->wr_tx_pends = kcalloc(SMC_WR_BUF_CNT,
592	sizeof(link->wr_tx_pends[0]),
593	GFP_KERNEL);
594	if (!link->wr_tx_pends)
595	goto no_mem_wr_tx_mask;
596	return 0;
597
598	no_mem_wr_tx_mask:
599	kfree(link->wr_tx_mask);
600	no_mem_wr_rx_sges:
601	kfree(link->wr_rx_sges);
602	no_mem_wr_tx_sges:
603	kfree(link->wr_tx_sges);
604	no_mem_wr_rx_ibs:
605	kfree(link->wr_rx_ibs);
606	no_mem_wr_tx_ibs:
607	kfree(link->wr_tx_ibs);
608	no_mem_wr_rx_bufs:
609	kfree(link->wr_rx_bufs);
610	no_mem_wr_tx_bufs:
611	kfree(link->wr_tx_bufs);
612	no_mem:
613	return -ENOMEM;
614	}
615
616	void smc_wr_remove_dev(struct smc_ib_device *smcibdev)
617	{
618	tasklet_kill(&smcibdev->recv_tasklet);
619	tasklet_kill(&smcibdev->send_tasklet);
620	}
621
622	void smc_wr_add_dev(struct smc_ib_device *smcibdev)
623	{
624	tasklet_init(&smcibdev->recv_tasklet, smc_wr_rx_tasklet_fn,
625	(unsigned long)smcibdev);
626	tasklet_init(&smcibdev->send_tasklet, smc_wr_tx_tasklet_fn,
627	(unsigned long)smcibdev);
628	}
629
630	int smc_wr_create_link(struct smc_link *lnk)
631	{
632	struct ib_device *ibdev = lnk->smcibdev->ibdev;
633	int rc = 0;
634
635	smc_wr_tx_set_wr_id(&lnk->wr_tx_id, 0);
636	lnk->wr_rx_id = 0;
637	lnk->wr_rx_dma_addr = ib_dma_map_single(
638	ibdev, lnk->wr_rx_bufs, SMC_WR_BUF_SIZE * lnk->wr_rx_cnt,
639	DMA_FROM_DEVICE);
640	if (ib_dma_mapping_error(ibdev, lnk->wr_rx_dma_addr)) {
641	lnk->wr_rx_dma_addr = 0;
642	rc = -EIO;
643	goto out;
644	}
645	lnk->wr_tx_dma_addr = ib_dma_map_single(
646	ibdev, lnk->wr_tx_bufs, SMC_WR_BUF_SIZE * lnk->wr_tx_cnt,
647	DMA_TO_DEVICE);
648	if (ib_dma_mapping_error(ibdev, lnk->wr_tx_dma_addr)) {
649	rc = -EIO;
650	goto dma_unmap;
651	}
652	smc_wr_init_sge(lnk);
653	memset(lnk->wr_tx_mask, 0,
654	BITS_TO_LONGS(SMC_WR_BUF_CNT) * sizeof(*lnk->wr_tx_mask));
652a1e41 UB	655	init_waitqueue_head(&lnk->wr_tx_wait);
652a1e41 UB	656	init_waitqueue_head(&lnk->wr_reg_wait);
f38ba179 UB	657	return rc;
	658
	659	dma_unmap:
	660	ib_dma_unmap_single(ibdev, lnk->wr_rx_dma_addr,
	661	SMC_WR_BUF_SIZE * lnk->wr_rx_cnt,
	662	DMA_FROM_DEVICE);
	663	lnk->wr_rx_dma_addr = 0;
	664	out:
	665	return rc;
	666	}