xref: /openbmc/linux/drivers/net/ethernet/qlogic/qed/qed_spq.c (revision fb09a1ed)

// SPDX-License-Identifier: (GPL-2.0-only OR BSD-3-Clause)
/* QLogic qed NIC Driver
 * Copyright (c) 2015-2017  QLogic Corporation
 * Copyright (c) 2019-2020 Marvell International Ltd.
 */

#include <linux/types.h>
#include <asm/byteorder.h>
#include <linux/io.h>
#include <linux/delay.h>
#include <linux/dma-mapping.h>
#include <linux/errno.h>
#include <linux/kernel.h>
#include <linux/list.h>
#include <linux/pci.h>
#include <linux/slab.h>
#include <linux/spinlock.h>
#include <linux/string.h>
#include "qed.h"
#include "qed_cxt.h"
#include "qed_dev_api.h"
#include "qed_hsi.h"
#include "qed_hw.h"
#include "qed_int.h"
#include "qed_iscsi.h"
#include "qed_mcp.h"
#include "qed_ooo.h"
#include "qed_reg_addr.h"
#include "qed_sp.h"
#include "qed_sriov.h"
#include "qed_rdma.h"

/***************************************************************************
* Structures & Definitions
***************************************************************************/

#define SPQ_HIGH_PRI_RESERVE_DEFAULT    (1)

#define SPQ_BLOCK_DELAY_MAX_ITER        (10)
#define SPQ_BLOCK_DELAY_US              (10)
#define SPQ_BLOCK_SLEEP_MAX_ITER        (1000)
#define SPQ_BLOCK_SLEEP_MS              (5)

/***************************************************************************
* Blocking Imp. (BLOCK/EBLOCK mode)
***************************************************************************/
static void qed_spq_blocking_cb(struct qed_hwfn *p_hwfn,
				void *cookie,
				union event_ring_data *data, u8 fw_return_code)
{
	struct qed_spq_comp_done *comp_done;

	comp_done = (struct qed_spq_comp_done *)cookie;

	comp_done->fw_return_code = fw_return_code;

	/* Make sure completion done is visible on waiting thread */
	smp_store_release(&comp_done->done, 0x1);
}

static int __qed_spq_block(struct qed_hwfn *p_hwfn,
			   struct qed_spq_entry *p_ent,
			   u8 *p_fw_ret, bool sleep_between_iter)
{
	struct qed_spq_comp_done *comp_done;
	u32 iter_cnt;

	comp_done = (struct qed_spq_comp_done *)p_ent->comp_cb.cookie;
	iter_cnt = sleep_between_iter ? SPQ_BLOCK_SLEEP_MAX_ITER
				      : SPQ_BLOCK_DELAY_MAX_ITER;

	while (iter_cnt--) {
		/* Validate we receive completion update */
		if (smp_load_acquire(&comp_done->done) == 1) { /* ^^^ */
			if (p_fw_ret)
				*p_fw_ret = comp_done->fw_return_code;
			return 0;
		}

		if (sleep_between_iter)
			msleep(SPQ_BLOCK_SLEEP_MS);
		else
			udelay(SPQ_BLOCK_DELAY_US);
	}

	return -EBUSY;
}

static int qed_spq_block(struct qed_hwfn *p_hwfn,
			 struct qed_spq_entry *p_ent,
			 u8 *p_fw_ret, bool skip_quick_poll)
{
	struct qed_spq_comp_done *comp_done;
	struct qed_ptt *p_ptt;
	int rc;

	/* A relatively short polling period w/o sleeping, to allow the FW to
	 * complete the ramrod and thus possibly to avoid the following sleeps.
	 */
	if (!skip_quick_poll) {
		rc = __qed_spq_block(p_hwfn, p_ent, p_fw_ret, false);
		if (!rc)
			return 0;
	}

	/* Move to polling with a sleeping period between iterations */
	rc = __qed_spq_block(p_hwfn, p_ent, p_fw_ret, true);
	if (!rc)
		return 0;

	p_ptt = qed_ptt_acquire(p_hwfn);
	if (!p_ptt) {
		DP_NOTICE(p_hwfn, "ptt, failed to acquire\n");
		return -EAGAIN;
	}

	DP_INFO(p_hwfn, "Ramrod is stuck, requesting MCP drain\n");
	rc = qed_mcp_drain(p_hwfn, p_ptt);
	qed_ptt_release(p_hwfn, p_ptt);
	if (rc) {
		DP_NOTICE(p_hwfn, "MCP drain failed\n");
		goto err;
	}

	/* Retry after drain */
	rc = __qed_spq_block(p_hwfn, p_ent, p_fw_ret, true);
	if (!rc)
		return 0;

	comp_done = (struct qed_spq_comp_done *)p_ent->comp_cb.cookie;
	if (comp_done->done == 1) {
		if (p_fw_ret)
			*p_fw_ret = comp_done->fw_return_code;
		return 0;
	}
err:
	p_ptt = qed_ptt_acquire(p_hwfn);
	if (!p_ptt)
		return -EBUSY;
	qed_hw_err_notify(p_hwfn, p_ptt, QED_HW_ERR_RAMROD_FAIL,
			  "Ramrod is stuck [CID %08x cmd %02x protocol %02x echo %04x]\n",
			  le32_to_cpu(p_ent->elem.hdr.cid),
			  p_ent->elem.hdr.cmd_id,
			  p_ent->elem.hdr.protocol_id,
			  le16_to_cpu(p_ent->elem.hdr.echo));
	qed_ptt_release(p_hwfn, p_ptt);

	return -EBUSY;
}

/***************************************************************************
* SPQ entries inner API
***************************************************************************/
static int qed_spq_fill_entry(struct qed_hwfn *p_hwfn,
			      struct qed_spq_entry *p_ent)
{
	p_ent->flags = 0;

	switch (p_ent->comp_mode) {
	case QED_SPQ_MODE_EBLOCK:
	case QED_SPQ_MODE_BLOCK:
		p_ent->comp_cb.function = qed_spq_blocking_cb;
		break;
	case QED_SPQ_MODE_CB:
		break;
	default:
		DP_NOTICE(p_hwfn, "Unknown SPQE completion mode %d\n",
			  p_ent->comp_mode);
		return -EINVAL;
	}

	DP_VERBOSE(p_hwfn, QED_MSG_SPQ,
		   "Ramrod header: [CID 0x%08x CMD 0x%02x protocol 0x%02x] Data pointer: [%08x:%08x] Completion Mode: %s\n",
		   p_ent->elem.hdr.cid,
		   p_ent->elem.hdr.cmd_id,
		   p_ent->elem.hdr.protocol_id,
		   p_ent->elem.data_ptr.hi,
		   p_ent->elem.data_ptr.lo,
		   D_TRINE(p_ent->comp_mode, QED_SPQ_MODE_EBLOCK,
			   QED_SPQ_MODE_BLOCK, "MODE_EBLOCK", "MODE_BLOCK",
			   "MODE_CB"));

	return 0;
}

/***************************************************************************
* HSI access
***************************************************************************/
static void qed_spq_hw_initialize(struct qed_hwfn *p_hwfn,
				  struct qed_spq *p_spq)
{
	struct core_conn_context *p_cxt;
	struct qed_cxt_info cxt_info;
	u16 physical_q;
	int rc;

	cxt_info.iid = p_spq->cid;

	rc = qed_cxt_get_cid_info(p_hwfn, &cxt_info);

	if (rc < 0) {
		DP_NOTICE(p_hwfn, "Cannot find context info for cid=%d\n",
			  p_spq->cid);
		return;
	}

	p_cxt = cxt_info.p_cxt;

	SET_FIELD(p_cxt->xstorm_ag_context.flags10,
		  XSTORM_CORE_CONN_AG_CTX_DQ_CF_EN, 1);
	SET_FIELD(p_cxt->xstorm_ag_context.flags1,
		  XSTORM_CORE_CONN_AG_CTX_DQ_CF_ACTIVE, 1);
	SET_FIELD(p_cxt->xstorm_ag_context.flags9,
		  XSTORM_CORE_CONN_AG_CTX_CONSOLID_PROD_CF_EN, 1);

	/* QM physical queue */
	physical_q = qed_get_cm_pq_idx(p_hwfn, PQ_FLAGS_LB);
	p_cxt->xstorm_ag_context.physical_q0 = cpu_to_le16(physical_q);

	p_cxt->xstorm_st_context.spq_base_lo =
		DMA_LO_LE(p_spq->chain.p_phys_addr);
	p_cxt->xstorm_st_context.spq_base_hi =
		DMA_HI_LE(p_spq->chain.p_phys_addr);

	DMA_REGPAIR_LE(p_cxt->xstorm_st_context.consolid_base_addr,
		       p_hwfn->p_consq->chain.p_phys_addr);
}

static int qed_spq_hw_post(struct qed_hwfn *p_hwfn,
			   struct qed_spq *p_spq, struct qed_spq_entry *p_ent)
{
	struct qed_chain *p_chain = &p_hwfn->p_spq->chain;
	struct core_db_data *p_db_data = &p_spq->db_data;
	u16 echo = qed_chain_get_prod_idx(p_chain);
	struct slow_path_element	*elem;

	p_ent->elem.hdr.echo	= cpu_to_le16(echo);
	elem = qed_chain_produce(p_chain);
	if (!elem) {
		DP_NOTICE(p_hwfn, "Failed to produce from SPQ chain\n");
		return -EINVAL;
	}

	*elem = p_ent->elem; /* struct assignment */

	/* send a doorbell on the slow hwfn session */
	p_db_data->spq_prod = cpu_to_le16(qed_chain_get_prod_idx(p_chain));

	/* make sure the SPQE is updated before the doorbell */
	wmb();

	DOORBELL(p_hwfn, p_spq->db_addr_offset, *(u32 *)p_db_data);

	/* make sure doorbell is rang */
	wmb();

	DP_VERBOSE(p_hwfn, QED_MSG_SPQ,
		   "Doorbelled [0x%08x, CID 0x%08x] with Flags: %02x agg_params: %02x, prod: %04x\n",
		   p_spq->db_addr_offset,
		   p_spq->cid,
		   p_db_data->params,
		   p_db_data->agg_flags, qed_chain_get_prod_idx(p_chain));

	return 0;
}

/***************************************************************************
* Asynchronous events
***************************************************************************/
static int
qed_async_event_completion(struct qed_hwfn *p_hwfn,
			   struct event_ring_entry *p_eqe)
{
	qed_spq_async_comp_cb cb;

	if (!p_hwfn->p_spq || (p_eqe->protocol_id >= MAX_PROTOCOL_TYPE))
		return -EINVAL;

	cb = p_hwfn->p_spq->async_comp_cb[p_eqe->protocol_id];
	if (cb) {
		return cb(p_hwfn, p_eqe->opcode, p_eqe->echo,
			  &p_eqe->data, p_eqe->fw_return_code);
	} else {
		DP_NOTICE(p_hwfn,
			  "Unknown Async completion for protocol: %d\n",
			  p_eqe->protocol_id);
		return -EINVAL;
	}
}

int
qed_spq_register_async_cb(struct qed_hwfn *p_hwfn,
			  enum protocol_type protocol_id,
			  qed_spq_async_comp_cb cb)
{
	if (!p_hwfn->p_spq || (protocol_id >= MAX_PROTOCOL_TYPE))
		return -EINVAL;

	p_hwfn->p_spq->async_comp_cb[protocol_id] = cb;
	return 0;
}

void
qed_spq_unregister_async_cb(struct qed_hwfn *p_hwfn,
			    enum protocol_type protocol_id)
{
	if (!p_hwfn->p_spq || (protocol_id >= MAX_PROTOCOL_TYPE))
		return;

	p_hwfn->p_spq->async_comp_cb[protocol_id] = NULL;
}

/***************************************************************************
* EQ API
***************************************************************************/
void qed_eq_prod_update(struct qed_hwfn *p_hwfn, u16 prod)
{
	u32 addr = GTT_BAR0_MAP_REG_USDM_RAM +
		   USTORM_EQE_CONS_OFFSET(p_hwfn->rel_pf_id);

	REG_WR16(p_hwfn, addr, prod);
}

int qed_eq_completion(struct qed_hwfn *p_hwfn, void *cookie)
{
	struct qed_eq *p_eq = cookie;
	struct qed_chain *p_chain = &p_eq->chain;
	int rc = 0;

	/* take a snapshot of the FW consumer */
	u16 fw_cons_idx = le16_to_cpu(*p_eq->p_fw_cons);

	DP_VERBOSE(p_hwfn, QED_MSG_SPQ, "fw_cons_idx %x\n", fw_cons_idx);

	/* Need to guarantee the fw_cons index we use points to a usuable
	 * element (to comply with our chain), so our macros would comply
	 */
	if ((fw_cons_idx & qed_chain_get_usable_per_page(p_chain)) ==
	    qed_chain_get_usable_per_page(p_chain))
		fw_cons_idx += qed_chain_get_unusable_per_page(p_chain);

	/* Complete current segment of eq entries */
	while (fw_cons_idx != qed_chain_get_cons_idx(p_chain)) {
		struct event_ring_entry *p_eqe = qed_chain_consume(p_chain);

		if (!p_eqe) {
			rc = -EINVAL;
			break;
		}

		DP_VERBOSE(p_hwfn, QED_MSG_SPQ,
			   "op %x prot %x res0 %x echo %x fwret %x flags %x\n",
			   p_eqe->opcode,
			   p_eqe->protocol_id,
			   p_eqe->reserved0,
			   le16_to_cpu(p_eqe->echo),
			   p_eqe->fw_return_code,
			   p_eqe->flags);

		if (GET_FIELD(p_eqe->flags, EVENT_RING_ENTRY_ASYNC)) {
			if (qed_async_event_completion(p_hwfn, p_eqe))
				rc = -EINVAL;
		} else if (qed_spq_completion(p_hwfn,
					      p_eqe->echo,
					      p_eqe->fw_return_code,
					      &p_eqe->data)) {
			rc = -EINVAL;
		}

		qed_chain_recycle_consumed(p_chain);
	}

	qed_eq_prod_update(p_hwfn, qed_chain_get_prod_idx(p_chain));

	/* Attempt to post pending requests */
	spin_lock_bh(&p_hwfn->p_spq->lock);
	rc = qed_spq_pend_post(p_hwfn);
	spin_unlock_bh(&p_hwfn->p_spq->lock);

	return rc;
}

int qed_eq_alloc(struct qed_hwfn *p_hwfn, u16 num_elem)
{
	struct qed_chain_init_params params = {
		.mode		= QED_CHAIN_MODE_PBL,
		.intended_use	= QED_CHAIN_USE_TO_PRODUCE,
		.cnt_type	= QED_CHAIN_CNT_TYPE_U16,
		.num_elems	= num_elem,
		.elem_size	= sizeof(union event_ring_element),
	};
	struct qed_eq *p_eq;
	int ret;

	/* Allocate EQ struct */
	p_eq = kzalloc(sizeof(*p_eq), GFP_KERNEL);
	if (!p_eq)
		return -ENOMEM;

	ret = qed_chain_alloc(p_hwfn->cdev, &p_eq->chain, &params);
	if (ret) {
		DP_NOTICE(p_hwfn, "Failed to allocate EQ chain\n");
		goto eq_allocate_fail;
	}

	/* register EQ completion on the SP SB */
	qed_int_register_cb(p_hwfn, qed_eq_completion,
			    p_eq, &p_eq->eq_sb_index, &p_eq->p_fw_cons);

	p_hwfn->p_eq = p_eq;
	return 0;

eq_allocate_fail:
	kfree(p_eq);

	return ret;
}

void qed_eq_setup(struct qed_hwfn *p_hwfn)
{
	qed_chain_reset(&p_hwfn->p_eq->chain);
}

void qed_eq_free(struct qed_hwfn *p_hwfn)
{
	if (!p_hwfn->p_eq)
		return;

	qed_chain_free(p_hwfn->cdev, &p_hwfn->p_eq->chain);

	kfree(p_hwfn->p_eq);
	p_hwfn->p_eq = NULL;
}

/***************************************************************************
* CQE API - manipulate EQ functionality
***************************************************************************/
static int qed_cqe_completion(struct qed_hwfn *p_hwfn,
			      struct eth_slow_path_rx_cqe *cqe,
			      enum protocol_type protocol)
{
	if (IS_VF(p_hwfn->cdev))
		return 0;

	/* @@@tmp - it's possible we'll eventually want to handle some
	 * actual commands that can arrive here, but for now this is only
	 * used to complete the ramrod using the echo value on the cqe
	 */
	return qed_spq_completion(p_hwfn, cqe->echo, 0, NULL);
}

int qed_eth_cqe_completion(struct qed_hwfn *p_hwfn,
			   struct eth_slow_path_rx_cqe *cqe)
{
	int rc;

	rc = qed_cqe_completion(p_hwfn, cqe, PROTOCOLID_ETH);
	if (rc)
		DP_NOTICE(p_hwfn,
			  "Failed to handle RXQ CQE [cmd 0x%02x]\n",
			  cqe->ramrod_cmd_id);

	return rc;
}

/***************************************************************************
* Slow hwfn Queue (spq)
***************************************************************************/
void qed_spq_setup(struct qed_hwfn *p_hwfn)
{
	struct qed_spq *p_spq = p_hwfn->p_spq;
	struct qed_spq_entry *p_virt = NULL;
	struct core_db_data *p_db_data;
	void __iomem *db_addr;
	dma_addr_t p_phys = 0;
	u32 i, capacity;
	int rc;

	INIT_LIST_HEAD(&p_spq->pending);
	INIT_LIST_HEAD(&p_spq->completion_pending);
	INIT_LIST_HEAD(&p_spq->free_pool);
	INIT_LIST_HEAD(&p_spq->unlimited_pending);
	spin_lock_init(&p_spq->lock);

	/* SPQ empty pool */
	p_phys	= p_spq->p_phys + offsetof(struct qed_spq_entry, ramrod);
	p_virt	= p_spq->p_virt;

	capacity = qed_chain_get_capacity(&p_spq->chain);
	for (i = 0; i < capacity; i++) {
		DMA_REGPAIR_LE(p_virt->elem.data_ptr, p_phys);

		list_add_tail(&p_virt->list, &p_spq->free_pool);

		p_virt++;
		p_phys += sizeof(struct qed_spq_entry);
	}

	/* Statistics */
	p_spq->normal_count		= 0;
	p_spq->comp_count		= 0;
	p_spq->comp_sent_count		= 0;
	p_spq->unlimited_pending_count	= 0;

	bitmap_zero(p_spq->p_comp_bitmap, SPQ_RING_SIZE);
	p_spq->comp_bitmap_idx = 0;

	/* SPQ cid, cannot fail */
	qed_cxt_acquire_cid(p_hwfn, PROTOCOLID_CORE, &p_spq->cid);
	qed_spq_hw_initialize(p_hwfn, p_spq);

	/* reset the chain itself */
	qed_chain_reset(&p_spq->chain);

	/* Initialize the address/data of the SPQ doorbell */
	p_spq->db_addr_offset = qed_db_addr(p_spq->cid, DQ_DEMS_LEGACY);
	p_db_data = &p_spq->db_data;
	memset(p_db_data, 0, sizeof(*p_db_data));
	SET_FIELD(p_db_data->params, CORE_DB_DATA_DEST, DB_DEST_XCM);
	SET_FIELD(p_db_data->params, CORE_DB_DATA_AGG_CMD, DB_AGG_CMD_MAX);
	SET_FIELD(p_db_data->params, CORE_DB_DATA_AGG_VAL_SEL,
		  DQ_XCM_CORE_SPQ_PROD_CMD);
	p_db_data->agg_flags = DQ_XCM_CORE_DQ_CF_CMD;

	/* Register the SPQ doorbell with the doorbell recovery mechanism */
	db_addr = (void __iomem *)((u8 __iomem *)p_hwfn->doorbells +
				   p_spq->db_addr_offset);
	rc = qed_db_recovery_add(p_hwfn->cdev, db_addr, &p_spq->db_data,
				 DB_REC_WIDTH_32B, DB_REC_KERNEL);
	if (rc)
		DP_INFO(p_hwfn,
			"Failed to register the SPQ doorbell with the doorbell recovery mechanism\n");
}

int qed_spq_alloc(struct qed_hwfn *p_hwfn)
{
	struct qed_chain_init_params params = {
		.mode		= QED_CHAIN_MODE_SINGLE,
		.intended_use	= QED_CHAIN_USE_TO_PRODUCE,
		.cnt_type	= QED_CHAIN_CNT_TYPE_U16,
		.elem_size	= sizeof(struct slow_path_element),
	};
	struct qed_dev *cdev = p_hwfn->cdev;
	struct qed_spq_entry *p_virt = NULL;
	struct qed_spq *p_spq = NULL;
	dma_addr_t p_phys = 0;
	u32 capacity;
	int ret;

	/* SPQ struct */
	p_spq = kzalloc(sizeof(struct qed_spq), GFP_KERNEL);
	if (!p_spq)
		return -ENOMEM;

	/* SPQ ring */
	ret = qed_chain_alloc(cdev, &p_spq->chain, &params);
	if (ret) {
		DP_NOTICE(p_hwfn, "Failed to allocate SPQ chain\n");
		goto spq_chain_alloc_fail;
	}

	/* allocate and fill the SPQ elements (incl. ramrod data list) */
	capacity = qed_chain_get_capacity(&p_spq->chain);
	ret = -ENOMEM;

	p_virt = dma_alloc_coherent(&cdev->pdev->dev,
				    capacity * sizeof(struct qed_spq_entry),
				    &p_phys, GFP_KERNEL);
	if (!p_virt)
		goto spq_alloc_fail;

	p_spq->p_virt = p_virt;
	p_spq->p_phys = p_phys;
	p_hwfn->p_spq = p_spq;

	return 0;

spq_alloc_fail:
	qed_chain_free(cdev, &p_spq->chain);
spq_chain_alloc_fail:
	kfree(p_spq);

	return ret;
}

void qed_spq_free(struct qed_hwfn *p_hwfn)
{
	struct qed_spq *p_spq = p_hwfn->p_spq;
	void __iomem *db_addr;
	u32 capacity;

	if (!p_spq)
		return;

	/* Delete the SPQ doorbell from the doorbell recovery mechanism */
	db_addr = (void __iomem *)((u8 __iomem *)p_hwfn->doorbells +
				   p_spq->db_addr_offset);
	qed_db_recovery_del(p_hwfn->cdev, db_addr, &p_spq->db_data);

	if (p_spq->p_virt) {
		capacity = qed_chain_get_capacity(&p_spq->chain);
		dma_free_coherent(&p_hwfn->cdev->pdev->dev,
				  capacity *
				  sizeof(struct qed_spq_entry),
				  p_spq->p_virt, p_spq->p_phys);
	}

	qed_chain_free(p_hwfn->cdev, &p_spq->chain);
	kfree(p_spq);
	p_hwfn->p_spq = NULL;
}

int qed_spq_get_entry(struct qed_hwfn *p_hwfn, struct qed_spq_entry **pp_ent)
{
	struct qed_spq *p_spq = p_hwfn->p_spq;
	struct qed_spq_entry *p_ent = NULL;
	int rc = 0;

	spin_lock_bh(&p_spq->lock);

	if (list_empty(&p_spq->free_pool)) {
		p_ent = kzalloc(sizeof(*p_ent), GFP_ATOMIC);
		if (!p_ent) {
			DP_NOTICE(p_hwfn,
				  "Failed to allocate an SPQ entry for a pending ramrod\n");
			rc = -ENOMEM;
			goto out_unlock;
		}
		p_ent->queue = &p_spq->unlimited_pending;
	} else {
		p_ent = list_first_entry(&p_spq->free_pool,
					 struct qed_spq_entry, list);
		list_del(&p_ent->list);
		p_ent->queue = &p_spq->pending;
	}

	*pp_ent = p_ent;

out_unlock:
	spin_unlock_bh(&p_spq->lock);
	return rc;
}

/* Locked variant; Should be called while the SPQ lock is taken */
static void __qed_spq_return_entry(struct qed_hwfn *p_hwfn,
				   struct qed_spq_entry *p_ent)
{
	list_add_tail(&p_ent->list, &p_hwfn->p_spq->free_pool);
}

void qed_spq_return_entry(struct qed_hwfn *p_hwfn, struct qed_spq_entry *p_ent)
{
	spin_lock_bh(&p_hwfn->p_spq->lock);
	__qed_spq_return_entry(p_hwfn, p_ent);
	spin_unlock_bh(&p_hwfn->p_spq->lock);
}

/**
 * qed_spq_add_entry() - Add a new entry to the pending list.
 *                       Should be used while lock is being held.
 *
 * @p_hwfn: HW device data.
 * @p_ent: An entry to add.
 * @priority: Desired priority.
 *
 * Adds an entry to the pending list is there is room (an empty
 * element is available in the free_pool), or else places the
 * entry in the unlimited_pending pool.
 *
 * Return: zero on success, -EINVAL on invalid @priority.
 */
static int qed_spq_add_entry(struct qed_hwfn *p_hwfn,
			     struct qed_spq_entry *p_ent,
			     enum spq_priority priority)
{
	struct qed_spq *p_spq = p_hwfn->p_spq;

	if (p_ent->queue == &p_spq->unlimited_pending) {

		if (list_empty(&p_spq->free_pool)) {
			list_add_tail(&p_ent->list, &p_spq->unlimited_pending);
			p_spq->unlimited_pending_count++;

			return 0;
		} else {
			struct qed_spq_entry *p_en2;

			p_en2 = list_first_entry(&p_spq->free_pool,
						 struct qed_spq_entry, list);
			list_del(&p_en2->list);

			/* Copy the ring element physical pointer to the new
			 * entry, since we are about to override the entire ring
			 * entry and don't want to lose the pointer.
			 */
			p_ent->elem.data_ptr = p_en2->elem.data_ptr;

			*p_en2 = *p_ent;

			/* EBLOCK responsible to free the allocated p_ent */
			if (p_ent->comp_mode != QED_SPQ_MODE_EBLOCK)
				kfree(p_ent);
			else
				p_ent->post_ent = p_en2;

			p_ent = p_en2;
		}
	}

	/* entry is to be placed in 'pending' queue */
	switch (priority) {
	case QED_SPQ_PRIORITY_NORMAL:
		list_add_tail(&p_ent->list, &p_spq->pending);
		p_spq->normal_count++;
		break;
	case QED_SPQ_PRIORITY_HIGH:
		list_add(&p_ent->list, &p_spq->pending);
		p_spq->high_count++;
		break;
	default:
		return -EINVAL;
	}

	return 0;
}

/***************************************************************************
* Accessor
***************************************************************************/
u32 qed_spq_get_cid(struct qed_hwfn *p_hwfn)
{
	if (!p_hwfn->p_spq)
		return 0xffffffff;      /* illegal */
	return p_hwfn->p_spq->cid;
}

/***************************************************************************
* Posting new Ramrods
***************************************************************************/
static int qed_spq_post_list(struct qed_hwfn *p_hwfn,
			     struct list_head *head, u32 keep_reserve)
{
	struct qed_spq *p_spq = p_hwfn->p_spq;
	int rc;

	while (qed_chain_get_elem_left(&p_spq->chain) > keep_reserve &&
	       !list_empty(head)) {
		struct qed_spq_entry *p_ent =
			list_first_entry(head, struct qed_spq_entry, list);
		list_move_tail(&p_ent->list, &p_spq->completion_pending);
		p_spq->comp_sent_count++;

		rc = qed_spq_hw_post(p_hwfn, p_spq, p_ent);
		if (rc) {
			list_del(&p_ent->list);
			__qed_spq_return_entry(p_hwfn, p_ent);
			return rc;
		}
	}

	return 0;
}

int qed_spq_pend_post(struct qed_hwfn *p_hwfn)
{
	struct qed_spq *p_spq = p_hwfn->p_spq;
	struct qed_spq_entry *p_ent = NULL;

	while (!list_empty(&p_spq->free_pool)) {
		if (list_empty(&p_spq->unlimited_pending))
			break;

		p_ent = list_first_entry(&p_spq->unlimited_pending,
					 struct qed_spq_entry, list);
		if (!p_ent)
			return -EINVAL;

		list_del(&p_ent->list);

		qed_spq_add_entry(p_hwfn, p_ent, p_ent->priority);
	}

	return qed_spq_post_list(p_hwfn, &p_spq->pending,
				 SPQ_HIGH_PRI_RESERVE_DEFAULT);
}

static void qed_spq_recov_set_ret_code(struct qed_spq_entry *p_ent,
				       u8 *fw_return_code)
{
	if (!fw_return_code)
		return;

	if (p_ent->elem.hdr.protocol_id == PROTOCOLID_ROCE ||
	    p_ent->elem.hdr.protocol_id == PROTOCOLID_IWARP)
		*fw_return_code = RDMA_RETURN_OK;
}

/* Avoid overriding of SPQ entries when getting out-of-order completions, by
 * marking the completions in a bitmap and increasing the chain consumer only
 * for the first successive completed entries.
 */
static void qed_spq_comp_bmap_update(struct qed_hwfn *p_hwfn, __le16 echo)
{
	u16 pos = le16_to_cpu(echo) % SPQ_RING_SIZE;
	struct qed_spq *p_spq = p_hwfn->p_spq;

	__set_bit(pos, p_spq->p_comp_bitmap);
	while (test_bit(p_spq->comp_bitmap_idx,
			p_spq->p_comp_bitmap)) {
		__clear_bit(p_spq->comp_bitmap_idx,
			    p_spq->p_comp_bitmap);
		p_spq->comp_bitmap_idx++;
		qed_chain_return_produced(&p_spq->chain);
	}
}

int qed_spq_post(struct qed_hwfn *p_hwfn,
		 struct qed_spq_entry *p_ent, u8 *fw_return_code)
{
	int rc = 0;
	struct qed_spq *p_spq = p_hwfn ? p_hwfn->p_spq : NULL;
	bool b_ret_ent = true;
	bool eblock;

	if (!p_hwfn)
		return -EINVAL;

	if (!p_ent) {
		DP_NOTICE(p_hwfn, "Got a NULL pointer\n");
		return -EINVAL;
	}

	if (p_hwfn->cdev->recov_in_prog) {
		DP_VERBOSE(p_hwfn,
			   QED_MSG_SPQ,
			   "Recovery is in progress. Skip spq post [cmd %02x protocol %02x]\n",
			   p_ent->elem.hdr.cmd_id, p_ent->elem.hdr.protocol_id);

		/* Let the flow complete w/o any error handling */
		qed_spq_recov_set_ret_code(p_ent, fw_return_code);
		return 0;
	}

	/* Complete the entry */
	rc = qed_spq_fill_entry(p_hwfn, p_ent);

	spin_lock_bh(&p_spq->lock);

	/* Check return value after LOCK is taken for cleaner error flow */
	if (rc)
		goto spq_post_fail;

	/* Check if entry is in block mode before qed_spq_add_entry,
	 * which might kfree p_ent.
	 */
	eblock = (p_ent->comp_mode == QED_SPQ_MODE_EBLOCK);

	/* Add the request to the pending queue */
	rc = qed_spq_add_entry(p_hwfn, p_ent, p_ent->priority);
	if (rc)
		goto spq_post_fail;

	rc = qed_spq_pend_post(p_hwfn);
	if (rc) {
		/* Since it's possible that pending failed for a different
		 * entry [although unlikely], the failed entry was already
		 * dealt with; No need to return it here.
		 */
		b_ret_ent = false;
		goto spq_post_fail;
	}

	spin_unlock_bh(&p_spq->lock);

	if (eblock) {
		/* For entries in QED BLOCK mode, the completion code cannot
		 * perform the necessary cleanup - if it did, we couldn't
		 * access p_ent here to see whether it's successful or not.
		 * Thus, after gaining the answer perform the cleanup here.
		 */
		rc = qed_spq_block(p_hwfn, p_ent, fw_return_code,
				   p_ent->queue == &p_spq->unlimited_pending);

		if (p_ent->queue == &p_spq->unlimited_pending) {
			struct qed_spq_entry *p_post_ent = p_ent->post_ent;

			kfree(p_ent);

			/* Return the entry which was actually posted */
			p_ent = p_post_ent;
		}

		if (rc)
			goto spq_post_fail2;

		/* return to pool */
		qed_spq_return_entry(p_hwfn, p_ent);
	}
	return rc;

spq_post_fail2:
	spin_lock_bh(&p_spq->lock);
	list_del(&p_ent->list);
	qed_spq_comp_bmap_update(p_hwfn, p_ent->elem.hdr.echo);

spq_post_fail:
	/* return to the free pool */
	if (b_ret_ent)
		__qed_spq_return_entry(p_hwfn, p_ent);
	spin_unlock_bh(&p_spq->lock);

	return rc;
}

int qed_spq_completion(struct qed_hwfn *p_hwfn,
		       __le16 echo,
		       u8 fw_return_code,
		       union event_ring_data *p_data)
{
	struct qed_spq		*p_spq;
	struct qed_spq_entry	*p_ent = NULL;
	struct qed_spq_entry	*tmp;
	struct qed_spq_entry	*found = NULL;

	if (!p_hwfn)
		return -EINVAL;

	p_spq = p_hwfn->p_spq;
	if (!p_spq)
		return -EINVAL;

	spin_lock_bh(&p_spq->lock);
	list_for_each_entry_safe(p_ent, tmp, &p_spq->completion_pending, list) {
		if (p_ent->elem.hdr.echo == echo) {
			list_del(&p_ent->list);
			qed_spq_comp_bmap_update(p_hwfn, echo);
			p_spq->comp_count++;
			found = p_ent;
			break;
		}

		/* This is relatively uncommon - depends on scenarios
		 * which have mutliple per-PF sent ramrods.
		 */
		DP_VERBOSE(p_hwfn, QED_MSG_SPQ,
			   "Got completion for echo %04x - doesn't match echo %04x in completion pending list\n",
			   le16_to_cpu(echo),
			   le16_to_cpu(p_ent->elem.hdr.echo));
	}

	/* Release lock before callback, as callback may post
	 * an additional ramrod.
	 */
	spin_unlock_bh(&p_spq->lock);

	if (!found) {
		DP_NOTICE(p_hwfn,
			  "Failed to find an entry this EQE [echo %04x] completes\n",
			  le16_to_cpu(echo));
		return -EEXIST;
	}

	DP_VERBOSE(p_hwfn, QED_MSG_SPQ,
		   "Complete EQE [echo %04x]: func %p cookie %p)\n",
		   le16_to_cpu(echo),
		   p_ent->comp_cb.function, p_ent->comp_cb.cookie);
	if (found->comp_cb.function)
		found->comp_cb.function(p_hwfn, found->comp_cb.cookie, p_data,
					fw_return_code);
	else
		DP_VERBOSE(p_hwfn,
			   QED_MSG_SPQ,
			   "Got a completion without a callback function\n");

	if (found->comp_mode != QED_SPQ_MODE_EBLOCK)
		/* EBLOCK  is responsible for returning its own entry into the
		 * free list.
		 */
		qed_spq_return_entry(p_hwfn, found);

	return 0;
}

#define QED_SPQ_CONSQ_ELEM_SIZE		0x80

int qed_consq_alloc(struct qed_hwfn *p_hwfn)
{
	struct qed_chain_init_params params = {
		.mode		= QED_CHAIN_MODE_PBL,
		.intended_use	= QED_CHAIN_USE_TO_PRODUCE,
		.cnt_type	= QED_CHAIN_CNT_TYPE_U16,
		.num_elems	= QED_CHAIN_PAGE_SIZE / QED_SPQ_CONSQ_ELEM_SIZE,
		.elem_size	= QED_SPQ_CONSQ_ELEM_SIZE,
	};
	struct qed_consq *p_consq;
	int ret;

	/* Allocate ConsQ struct */
	p_consq = kzalloc(sizeof(*p_consq), GFP_KERNEL);
	if (!p_consq)
		return -ENOMEM;

	/* Allocate and initialize ConsQ chain */
	ret = qed_chain_alloc(p_hwfn->cdev, &p_consq->chain, &params);
	if (ret) {
		DP_NOTICE(p_hwfn, "Failed to allocate ConsQ chain");
		goto consq_alloc_fail;
	}

	p_hwfn->p_consq = p_consq;

	return 0;

consq_alloc_fail:
	kfree(p_consq);

	return ret;
}

void qed_consq_setup(struct qed_hwfn *p_hwfn)
{
	qed_chain_reset(&p_hwfn->p_consq->chain);
}

void qed_consq_free(struct qed_hwfn *p_hwfn)
{
	if (!p_hwfn->p_consq)
		return;

	qed_chain_free(p_hwfn->cdev, &p_hwfn->p_consq->chain);

	kfree(p_hwfn->p_consq);
	p_hwfn->p_consq = NULL;
}