/*******************************************************************
 * This file is part of the Emulex Linux Device Driver for         *
 * Fibre Channel Host Bus Adapters.                                *
 * Copyright (C) 2004-2009 Emulex.  All rights reserved.           *
 * EMULEX and SLI are trademarks of Emulex.                        *
 * www.emulex.com                                                  *
 * Portions Copyright (C) 2004-2005 Christoph Hellwig              *
 *                                                                 *
 * This program is free software; you can redistribute it and/or   *
 * modify it under the terms of version 2 of the GNU General       *
 * Public License as published by the Free Software Foundation.    *
 * This program is distributed in the hope that it will be useful. *
 * ALL EXPRESS OR IMPLIED CONDITIONS, REPRESENTATIONS AND          *
 * WARRANTIES, INCLUDING ANY IMPLIED WARRANTY OF MERCHANTABILITY,  *
 * FITNESS FOR A PARTICULAR PURPOSE, OR NON-INFRINGEMENT, ARE      *
 * DISCLAIMED, EXCEPT TO THE EXTENT THAT SUCH DISCLAIMERS ARE HELD *
 * TO BE LEGALLY INVALID.  See the GNU General Public License for  *
 * more details, a copy of which can be found in the file COPYING  *
 * included with this package.                                     *
 *******************************************************************/
#include <linux/pci.h>
#include <linux/slab.h>
#include <linux/interrupt.h>
#include <linux/delay.h>
#include <asm/unaligned.h>

#include <scsi/scsi.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_eh.h>
#include <scsi/scsi_host.h>
#include <scsi/scsi_tcq.h>
#include <scsi/scsi_transport_fc.h>

#include "lpfc_version.h"
#include "lpfc_hw4.h"
#include "lpfc_hw.h"
#include "lpfc_sli.h"
#include "lpfc_sli4.h"
#include "lpfc_nl.h"
#include "lpfc_disc.h"
#include "lpfc_scsi.h"
#include "lpfc.h"
#include "lpfc_logmsg.h"
#include "lpfc_crtn.h"
#include "lpfc_vport.h"

#define LPFC_RESET_WAIT  2
#define LPFC_ABORT_WAIT  2

int _dump_buf_done;

static char *dif_op_str[] = {
	"SCSI_PROT_NORMAL",
	"SCSI_PROT_READ_INSERT",
	"SCSI_PROT_WRITE_STRIP",
	"SCSI_PROT_READ_STRIP",
	"SCSI_PROT_WRITE_INSERT",
	"SCSI_PROT_READ_PASS",
	"SCSI_PROT_WRITE_PASS",
};
static void
lpfc_release_scsi_buf_s4(struct lpfc_hba *phba, struct lpfc_scsi_buf *psb);
static void
lpfc_release_scsi_buf_s3(struct lpfc_hba *phba, struct lpfc_scsi_buf *psb);

static void
lpfc_debug_save_data(struct lpfc_hba *phba, struct scsi_cmnd *cmnd)
{
	void *src, *dst;
	struct scatterlist *sgde = scsi_sglist(cmnd);

	if (!_dump_buf_data) {
		lpfc_printf_log(phba, KERN_ERR, LOG_BG,
			"9050 BLKGRD: ERROR %s _dump_buf_data is NULL\n",
				__func__);
		return;
	}


	if (!sgde) {
		lpfc_printf_log(phba, KERN_ERR, LOG_BG,
			"9051 BLKGRD: ERROR: data scatterlist is null\n");
		return;
	}

	dst = (void *) _dump_buf_data;
	while (sgde) {
		src = sg_virt(sgde);
		memcpy(dst, src, sgde->length);
		dst += sgde->length;
		sgde = sg_next(sgde);
	}
}

static void
lpfc_debug_save_dif(struct lpfc_hba *phba, struct scsi_cmnd *cmnd)
{
	void *src, *dst;
	struct scatterlist *sgde = scsi_prot_sglist(cmnd);

	if (!_dump_buf_dif) {
		lpfc_printf_log(phba, KERN_ERR, LOG_BG,
			"9052 BLKGRD: ERROR %s _dump_buf_data is NULL\n",
				__func__);
		return;
	}

	if (!sgde) {
		lpfc_printf_log(phba, KERN_ERR, LOG_BG,
			"9053 BLKGRD: ERROR: prot scatterlist is null\n");
		return;
	}

	dst = _dump_buf_dif;
	while (sgde) {
		src = sg_virt(sgde);
		memcpy(dst, src, sgde->length);
		dst += sgde->length;
		sgde = sg_next(sgde);
	}
}

/**
 * lpfc_sli4_set_rsp_sgl_last - Set the last bit in the response sge.
 * @phba: Pointer to HBA object.
 * @lpfc_cmd: lpfc scsi command object pointer.
 *
 * This function is called from the lpfc_prep_task_mgmt_cmd function to
 * set the last bit in the response sge entry.
 **/
static void
lpfc_sli4_set_rsp_sgl_last(struct lpfc_hba *phba,
				struct lpfc_scsi_buf *lpfc_cmd)
{
	struct sli4_sge *sgl = (struct sli4_sge *)lpfc_cmd->fcp_bpl;
	if (sgl) {
		sgl += 1;
		sgl->word2 = le32_to_cpu(sgl->word2);
		bf_set(lpfc_sli4_sge_last, sgl, 1);
		sgl->word2 = cpu_to_le32(sgl->word2);
	}
}

/**
 * lpfc_update_stats - Update statistical data for the command completion
 * @phba: Pointer to HBA object.
 * @lpfc_cmd: lpfc scsi command object pointer.
 *
 * This function is called when there is a command completion and this
 * function updates the statistical data for the command completion.
 **/
static void
lpfc_update_stats(struct lpfc_hba *phba, struct  lpfc_scsi_buf *lpfc_cmd)
{
	struct lpfc_rport_data *rdata = lpfc_cmd->rdata;
	struct lpfc_nodelist *pnode = rdata->pnode;
	struct scsi_cmnd *cmd = lpfc_cmd->pCmd;
	unsigned long flags;
	struct Scsi_Host  *shost = cmd->device->host;
	struct lpfc_vport *vport = (struct lpfc_vport *) shost->hostdata;
	unsigned long latency;
	int i;

	if (cmd->result)
		return;

	latency = jiffies_to_msecs((long)jiffies - (long)lpfc_cmd->start_time);

	spin_lock_irqsave(shost->host_lock, flags);
	if (!vport->stat_data_enabled ||
		vport->stat_data_blocked ||
		!pnode->lat_data ||
		(phba->bucket_type == LPFC_NO_BUCKET)) {
		spin_unlock_irqrestore(shost->host_lock, flags);
		return;
	}

	if (phba->bucket_type == LPFC_LINEAR_BUCKET) {
		i = (latency + phba->bucket_step - 1 - phba->bucket_base)/
			phba->bucket_step;
		/* check array subscript bounds */
		if (i < 0)
			i = 0;
		else if (i >= LPFC_MAX_BUCKET_COUNT)
			i = LPFC_MAX_BUCKET_COUNT - 1;
	} else {
		for (i = 0; i < LPFC_MAX_BUCKET_COUNT-1; i++)
			if (latency <= (phba->bucket_base +
				((1<<i)*phba->bucket_step)))
				break;
	}

	pnode->lat_data[i].cmd_count++;
	spin_unlock_irqrestore(shost->host_lock, flags);
}

/**
 * lpfc_send_sdev_queuedepth_change_event - Posts a queuedepth change event
 * @phba: Pointer to HBA context object.
 * @vport: Pointer to vport object.
 * @ndlp: Pointer to FC node associated with the target.
 * @lun: Lun number of the scsi device.
 * @old_val: Old value of the queue depth.
 * @new_val: New value of the queue depth.
 *
 * This function sends an event to the mgmt application indicating
 * there is a change in the scsi device queue depth.
 **/
static void
lpfc_send_sdev_queuedepth_change_event(struct lpfc_hba *phba,
		struct lpfc_vport  *vport,
		struct lpfc_nodelist *ndlp,
		uint32_t lun,
		uint32_t old_val,
		uint32_t new_val)
{
	struct lpfc_fast_path_event *fast_path_evt;
	unsigned long flags;

	fast_path_evt = lpfc_alloc_fast_evt(phba);
	if (!fast_path_evt)
		return;

	fast_path_evt->un.queue_depth_evt.scsi_event.event_type =
		FC_REG_SCSI_EVENT;
	fast_path_evt->un.queue_depth_evt.scsi_event.subcategory =
		LPFC_EVENT_VARQUEDEPTH;

	/* Report all luns with change in queue depth */
	fast_path_evt->un.queue_depth_evt.scsi_event.lun = lun;
	if (ndlp && NLP_CHK_NODE_ACT(ndlp)) {
		memcpy(&fast_path_evt->un.queue_depth_evt.scsi_event.wwpn,
			&ndlp->nlp_portname, sizeof(struct lpfc_name));
		memcpy(&fast_path_evt->un.queue_depth_evt.scsi_event.wwnn,
			&ndlp->nlp_nodename, sizeof(struct lpfc_name));
	}

	fast_path_evt->un.queue_depth_evt.oldval = old_val;
	fast_path_evt->un.queue_depth_evt.newval = new_val;
	fast_path_evt->vport = vport;

	fast_path_evt->work_evt.evt = LPFC_EVT_FASTPATH_MGMT_EVT;
	spin_lock_irqsave(&phba->hbalock, flags);
	list_add_tail(&fast_path_evt->work_evt.evt_listp, &phba->work_list);
	spin_unlock_irqrestore(&phba->hbalock, flags);
	lpfc_worker_wake_up(phba);

	return;
}

/**
 * lpfc_change_queue_depth - Alter scsi device queue depth
 * @sdev: Pointer the scsi device on which to change the queue depth.
 * @qdepth: New queue depth to set the sdev to.
 * @reason: The reason for the queue depth change.
 *
 * This function is called by the midlayer and the LLD to alter the queue
 * depth for a scsi device. This function sets the queue depth to the new
 * value and sends an event out to log the queue depth change.
 **/
int
lpfc_change_queue_depth(struct scsi_device *sdev, int qdepth, int reason)
{
	struct lpfc_vport *vport = (struct lpfc_vport *) sdev->host->hostdata;
	struct lpfc_hba   *phba = vport->phba;
	struct lpfc_rport_data *rdata;
	unsigned long new_queue_depth, old_queue_depth;

	old_queue_depth = sdev->queue_depth;
	scsi_adjust_queue_depth(sdev, scsi_get_tag_type(sdev), qdepth);
	new_queue_depth = sdev->queue_depth;
	rdata = sdev->hostdata;
	if (rdata)
		lpfc_send_sdev_queuedepth_change_event(phba, vport,
						       rdata->pnode, sdev->lun,
						       old_queue_depth,
						       new_queue_depth);
	return sdev->queue_depth;
}

/**
 * lpfc_rampdown_queue_depth - Post RAMP_DOWN_QUEUE event to worker thread
 * @phba: The Hba for which this call is being executed.
 *
 * This routine is called when there is resource error in driver or firmware.
 * This routine posts WORKER_RAMP_DOWN_QUEUE event for @phba. This routine
 * posts at most 1 event each second. This routine wakes up worker thread of
 * @phba to process WORKER_RAM_DOWN_EVENT event.
 *
 * This routine should be called with no lock held.
 **/
void
lpfc_rampdown_queue_depth(struct lpfc_hba *phba)
{
	unsigned long flags;
	uint32_t evt_posted;

	spin_lock_irqsave(&phba->hbalock, flags);
	atomic_inc(&phba->num_rsrc_err);
	phba->last_rsrc_error_time = jiffies;

	if ((phba->last_ramp_down_time + QUEUE_RAMP_DOWN_INTERVAL) > jiffies) {
		spin_unlock_irqrestore(&phba->hbalock, flags);
		return;
	}

	phba->last_ramp_down_time = jiffies;

	spin_unlock_irqrestore(&phba->hbalock, flags);

	spin_lock_irqsave(&phba->pport->work_port_lock, flags);
	evt_posted = phba->pport->work_port_events & WORKER_RAMP_DOWN_QUEUE;
	if (!evt_posted)
		phba->pport->work_port_events |= WORKER_RAMP_DOWN_QUEUE;
	spin_unlock_irqrestore(&phba->pport->work_port_lock, flags);

	if (!evt_posted)
		lpfc_worker_wake_up(phba);
	return;
}

/**
 * lpfc_rampup_queue_depth - Post RAMP_UP_QUEUE event for worker thread
 * @phba: The Hba for which this call is being executed.
 *
 * This routine post WORKER_RAMP_UP_QUEUE event for @phba vport. This routine
 * post at most 1 event every 5 minute after last_ramp_up_time or
 * last_rsrc_error_time.  This routine wakes up worker thread of @phba
 * to process WORKER_RAM_DOWN_EVENT event.
 *
 * This routine should be called with no lock held.
 **/
static inline void
lpfc_rampup_queue_depth(struct lpfc_vport  *vport,
			uint32_t queue_depth)
{
	unsigned long flags;
	struct lpfc_hba *phba = vport->phba;
	uint32_t evt_posted;
	atomic_inc(&phba->num_cmd_success);

	if (vport->cfg_lun_queue_depth <= queue_depth)
		return;
	spin_lock_irqsave(&phba->hbalock, flags);
	if (time_before(jiffies,
			phba->last_ramp_up_time + QUEUE_RAMP_UP_INTERVAL) ||
	    time_before(jiffies,
			phba->last_rsrc_error_time + QUEUE_RAMP_UP_INTERVAL)) {
		spin_unlock_irqrestore(&phba->hbalock, flags);
		return;
	}
	phba->last_ramp_up_time = jiffies;
	spin_unlock_irqrestore(&phba->hbalock, flags);

	spin_lock_irqsave(&phba->pport->work_port_lock, flags);
	evt_posted = phba->pport->work_port_events & WORKER_RAMP_UP_QUEUE;
	if (!evt_posted)
		phba->pport->work_port_events |= WORKER_RAMP_UP_QUEUE;
	spin_unlock_irqrestore(&phba->pport->work_port_lock, flags);

	if (!evt_posted)
		lpfc_worker_wake_up(phba);
	return;
}

/**
 * lpfc_ramp_down_queue_handler - WORKER_RAMP_DOWN_QUEUE event handler
 * @phba: The Hba for which this call is being executed.
 *
 * This routine is called to  process WORKER_RAMP_DOWN_QUEUE event for worker
 * thread.This routine reduces queue depth for all scsi device on each vport
 * associated with @phba.
 **/
void
lpfc_ramp_down_queue_handler(struct lpfc_hba *phba)
{
	struct lpfc_vport **vports;
	struct Scsi_Host  *shost;
	struct scsi_device *sdev;
	unsigned long new_queue_depth;
	unsigned long num_rsrc_err, num_cmd_success;
	int i;

	num_rsrc_err = atomic_read(&phba->num_rsrc_err);
	num_cmd_success = atomic_read(&phba->num_cmd_success);

	vports = lpfc_create_vport_work_array(phba);
	if (vports != NULL)
		for (i = 0; i <= phba->max_vports && vports[i] != NULL; i++) {
			shost = lpfc_shost_from_vport(vports[i]);
			shost_for_each_device(sdev, shost) {
				new_queue_depth =
					sdev->queue_depth * num_rsrc_err /
					(num_rsrc_err + num_cmd_success);
				if (!new_queue_depth)
					new_queue_depth = sdev->queue_depth - 1;
				else
					new_queue_depth = sdev->queue_depth -
								new_queue_depth;
				lpfc_change_queue_depth(sdev, new_queue_depth,
							SCSI_QDEPTH_DEFAULT);
			}
		}
	lpfc_destroy_vport_work_array(phba, vports);
	atomic_set(&phba->num_rsrc_err, 0);
	atomic_set(&phba->num_cmd_success, 0);
}

/**
 * lpfc_ramp_up_queue_handler - WORKER_RAMP_UP_QUEUE event handler
 * @phba: The Hba for which this call is being executed.
 *
 * This routine is called to  process WORKER_RAMP_UP_QUEUE event for worker
 * thread.This routine increases queue depth for all scsi device on each vport
 * associated with @phba by 1. This routine also sets @phba num_rsrc_err and
 * num_cmd_success to zero.
 **/
void
lpfc_ramp_up_queue_handler(struct lpfc_hba *phba)
{
	struct lpfc_vport **vports;
	struct Scsi_Host  *shost;
	struct scsi_device *sdev;
	int i;

	vports = lpfc_create_vport_work_array(phba);
	if (vports != NULL)
		for (i = 0; i <= phba->max_vports && vports[i] != NULL; i++) {
			shost = lpfc_shost_from_vport(vports[i]);
			shost_for_each_device(sdev, shost) {
				if (vports[i]->cfg_lun_queue_depth <=
				    sdev->queue_depth)
					continue;
				lpfc_change_queue_depth(sdev,
							sdev->queue_depth+1,
							SCSI_QDEPTH_RAMP_UP);
			}
		}
	lpfc_destroy_vport_work_array(phba, vports);
	atomic_set(&phba->num_rsrc_err, 0);
	atomic_set(&phba->num_cmd_success, 0);
}

/**
 * lpfc_scsi_dev_block - set all scsi hosts to block state
 * @phba: Pointer to HBA context object.
 *
 * This function walks vport list and set each SCSI host to block state
 * by invoking fc_remote_port_delete() routine. This function is invoked
 * with EEH when device's PCI slot has been permanently disabled.
 **/
void
lpfc_scsi_dev_block(struct lpfc_hba *phba)
{
	struct lpfc_vport **vports;
	struct Scsi_Host  *shost;
	struct scsi_device *sdev;
	struct fc_rport *rport;
	int i;

	vports = lpfc_create_vport_work_array(phba);
	if (vports != NULL)
		for (i = 0; i <= phba->max_vports && vports[i] != NULL; i++) {
			shost = lpfc_shost_from_vport(vports[i]);
			shost_for_each_device(sdev, shost) {
				rport = starget_to_rport(scsi_target(sdev));
				fc_remote_port_delete(rport);
			}
		}
	lpfc_destroy_vport_work_array(phba, vports);
}

/**
 * lpfc_new_scsi_buf_s3 - Scsi buffer allocator for HBA with SLI3 IF spec
 * @vport: The virtual port for which this call being executed.
 * @num_to_allocate: The requested number of buffers to allocate.
 *
 * This routine allocates a scsi buffer for device with SLI-3 interface spec,
 * the scsi buffer contains all the necessary information needed to initiate
 * a SCSI I/O. The non-DMAable buffer region contains information to build
 * the IOCB. The DMAable region contains memory for the FCP CMND, FCP RSP,
 * and the initial BPL. In addition to allocating memory, the FCP CMND and
 * FCP RSP BDEs are setup in the BPL and the BPL BDE is setup in the IOCB.
 *
 * Return codes:
 *   int - number of scsi buffers that were allocated.
 *   0 = failure, less than num_to_alloc is a partial failure.
 **/
static int
lpfc_new_scsi_buf_s3(struct lpfc_vport *vport, int num_to_alloc)
{
	struct lpfc_hba *phba = vport->phba;
	struct lpfc_scsi_buf *psb;
	struct ulp_bde64 *bpl;
	IOCB_t *iocb;
	dma_addr_t pdma_phys_fcp_cmd;
	dma_addr_t pdma_phys_fcp_rsp;
	dma_addr_t pdma_phys_bpl;
	uint16_t iotag;
	int bcnt;

	for (bcnt = 0; bcnt < num_to_alloc; bcnt++) {
		psb = kzalloc(sizeof(struct lpfc_scsi_buf), GFP_KERNEL);
		if (!psb)
			break;

		/*
		 * Get memory from the pci pool to map the virt space to pci
		 * bus space for an I/O.  The DMA buffer includes space for the
		 * struct fcp_cmnd, struct fcp_rsp and the number of bde's
		 * necessary to support the sg_tablesize.
		 */
		psb->data = pci_pool_alloc(phba->lpfc_scsi_dma_buf_pool,
					GFP_KERNEL, &psb->dma_handle);
		if (!psb->data) {
			kfree(psb);
			break;
		}

		/* Initialize virtual ptrs to dma_buf region. */
		memset(psb->data, 0, phba->cfg_sg_dma_buf_size);

		/* Allocate iotag for psb->cur_iocbq. */
		iotag = lpfc_sli_next_iotag(phba, &psb->cur_iocbq);
		if (iotag == 0) {
			pci_pool_free(phba->lpfc_scsi_dma_buf_pool,
					psb->data, psb->dma_handle);
			kfree(psb);
			break;
		}
		psb->cur_iocbq.iocb_flag |= LPFC_IO_FCP;

		psb->fcp_cmnd = psb->data;
		psb->fcp_rsp = psb->data + sizeof(struct fcp_cmnd);
		psb->fcp_bpl = psb->data + sizeof(struct fcp_cmnd) +
			sizeof(struct fcp_rsp);

		/* Initialize local short-hand pointers. */
		bpl = psb->fcp_bpl;
		pdma_phys_fcp_cmd = psb->dma_handle;
		pdma_phys_fcp_rsp = psb->dma_handle + sizeof(struct fcp_cmnd);
		pdma_phys_bpl = psb->dma_handle + sizeof(struct fcp_cmnd) +
			sizeof(struct fcp_rsp);

		/*
		 * The first two bdes are the FCP_CMD and FCP_RSP. The balance
		 * are sg list bdes.  Initialize the first two and leave the
		 * rest for queuecommand.
		 */
		bpl[0].addrHigh = le32_to_cpu(putPaddrHigh(pdma_phys_fcp_cmd));
		bpl[0].addrLow = le32_to_cpu(putPaddrLow(pdma_phys_fcp_cmd));
		bpl[0].tus.f.bdeSize = sizeof(struct fcp_cmnd);
		bpl[0].tus.f.bdeFlags = BUFF_TYPE_BDE_64;
		bpl[0].tus.w = le32_to_cpu(bpl[0].tus.w);

		/* Setup the physical region for the FCP RSP */
		bpl[1].addrHigh = le32_to_cpu(putPaddrHigh(pdma_phys_fcp_rsp));
		bpl[1].addrLow = le32_to_cpu(putPaddrLow(pdma_phys_fcp_rsp));
		bpl[1].tus.f.bdeSize = sizeof(struct fcp_rsp);
		bpl[1].tus.f.bdeFlags = BUFF_TYPE_BDE_64;
		bpl[1].tus.w = le32_to_cpu(bpl[1].tus.w);

		/*
		 * Since the IOCB for the FCP I/O is built into this
		 * lpfc_scsi_buf, initialize it with all known data now.
		 */
		iocb = &psb->cur_iocbq.iocb;
		iocb->un.fcpi64.bdl.ulpIoTag32 = 0;
		if ((phba->sli_rev == 3) &&
				!(phba->sli3_options & LPFC_SLI3_BG_ENABLED)) {
			/* fill in immediate fcp command BDE */
			iocb->un.fcpi64.bdl.bdeFlags = BUFF_TYPE_BDE_IMMED;
			iocb->un.fcpi64.bdl.bdeSize = sizeof(struct fcp_cmnd);
			iocb->un.fcpi64.bdl.addrLow = offsetof(IOCB_t,
					unsli3.fcp_ext.icd);
			iocb->un.fcpi64.bdl.addrHigh = 0;
			iocb->ulpBdeCount = 0;
			iocb->ulpLe = 0;
			/* fill in responce BDE */
			iocb->unsli3.fcp_ext.rbde.tus.f.bdeFlags =
							BUFF_TYPE_BDE_64;
			iocb->unsli3.fcp_ext.rbde.tus.f.bdeSize =
				sizeof(struct fcp_rsp);
			iocb->unsli3.fcp_ext.rbde.addrLow =
				putPaddrLow(pdma_phys_fcp_rsp);
			iocb->unsli3.fcp_ext.rbde.addrHigh =
				putPaddrHigh(pdma_phys_fcp_rsp);
		} else {
			iocb->un.fcpi64.bdl.bdeFlags = BUFF_TYPE_BLP_64;
			iocb->un.fcpi64.bdl.bdeSize =
					(2 * sizeof(struct ulp_bde64));
			iocb->un.fcpi64.bdl.addrLow =
					putPaddrLow(pdma_phys_bpl);
			iocb->un.fcpi64.bdl.addrHigh =
					putPaddrHigh(pdma_phys_bpl);
			iocb->ulpBdeCount = 1;
			iocb->ulpLe = 1;
		}
		iocb->ulpClass = CLASS3;
		psb->status = IOSTAT_SUCCESS;
		/* Put it back into the SCSI buffer list */
		lpfc_release_scsi_buf_s3(phba, psb);

	}

	return bcnt;
}

/**
 * lpfc_sli4_fcp_xri_aborted - Fast-path process of fcp xri abort
 * @phba: pointer to lpfc hba data structure.
 * @axri: pointer to the fcp xri abort wcqe structure.
 *
 * This routine is invoked by the worker thread to process a SLI4 fast-path
 * FCP aborted xri.
 **/
void
lpfc_sli4_fcp_xri_aborted(struct lpfc_hba *phba,
			  struct sli4_wcqe_xri_aborted *axri)
{
	uint16_t xri = bf_get(lpfc_wcqe_xa_xri, axri);
	struct lpfc_scsi_buf *psb, *next_psb;
	unsigned long iflag = 0;
	struct lpfc_iocbq *iocbq;
	int i;

	spin_lock_irqsave(&phba->hbalock, iflag);
	spin_lock(&phba->sli4_hba.abts_scsi_buf_list_lock);
	list_for_each_entry_safe(psb, next_psb,
		&phba->sli4_hba.lpfc_abts_scsi_buf_list, list) {
		if (psb->cur_iocbq.sli4_xritag == xri) {
			list_del(&psb->list);
			psb->exch_busy = 0;
			psb->status = IOSTAT_SUCCESS;
			spin_unlock(
				&phba->sli4_hba.abts_scsi_buf_list_lock);
			spin_unlock_irqrestore(&phba->hbalock, iflag);
			lpfc_release_scsi_buf_s4(phba, psb);
			return;
		}
	}
	spin_unlock(&phba->sli4_hba.abts_scsi_buf_list_lock);
	for (i = 1; i <= phba->sli.last_iotag; i++) {
		iocbq = phba->sli.iocbq_lookup[i];

		if (!(iocbq->iocb_flag &  LPFC_IO_FCP) ||
			(iocbq->iocb_flag & LPFC_IO_LIBDFC))
			continue;
		if (iocbq->sli4_xritag != xri)
			continue;
		psb = container_of(iocbq, struct lpfc_scsi_buf, cur_iocbq);
		psb->exch_busy = 0;
		spin_unlock_irqrestore(&phba->hbalock, iflag);
		return;

	}
	spin_unlock_irqrestore(&phba->hbalock, iflag);
}

/**
 * lpfc_sli4_repost_scsi_sgl_list - Repsot the Scsi buffers sgl pages as block
 * @phba: pointer to lpfc hba data structure.
 *
 * This routine walks the list of scsi buffers that have been allocated and
 * repost them to the HBA by using SGL block post. This is needed after a
 * pci_function_reset/warm_start or start. The lpfc_hba_down_post_s4 routine
 * is responsible for moving all scsi buffers on the lpfc_abts_scsi_sgl_list
 * to the lpfc_scsi_buf_list. If the repost fails, reject all scsi buffers.
 *
 * Returns: 0 = success, non-zero failure.
 **/
int
lpfc_sli4_repost_scsi_sgl_list(struct lpfc_hba *phba)
{
	struct lpfc_scsi_buf *psb;
	int index, status, bcnt = 0, rcnt = 0, rc = 0;
	LIST_HEAD(sblist);

	for (index = 0; index < phba->sli4_hba.scsi_xri_cnt; index++) {
		psb = phba->sli4_hba.lpfc_scsi_psb_array[index];
		if (psb) {
			/* Remove from SCSI buffer list */
			list_del(&psb->list);
			/* Add it to a local SCSI buffer list */
			list_add_tail(&psb->list, &sblist);
			if (++rcnt == LPFC_NEMBED_MBOX_SGL_CNT) {
				bcnt = rcnt;
				rcnt = 0;
			}
		} else
			/* A hole present in the XRI array, need to skip */
			bcnt = rcnt;

		if (index == phba->sli4_hba.scsi_xri_cnt - 1)
			/* End of XRI array for SCSI buffer, complete */
			bcnt = rcnt;

		/* Continue until collect up to a nembed page worth of sgls */
		if (bcnt == 0)
			continue;
		/* Now, post the SCSI buffer list sgls as a block */
		status = lpfc_sli4_post_scsi_sgl_block(phba, &sblist, bcnt);
		/* Reset SCSI buffer count for next round of posting */
		bcnt = 0;
		while (!list_empty(&sblist)) {
			list_remove_head(&sblist, psb, struct lpfc_scsi_buf,
					 list);
			if (status) {
				/* Put this back on the abort scsi list */
				psb->exch_busy = 1;
				rc++;
			} else {
				psb->exch_busy = 0;
				psb->status = IOSTAT_SUCCESS;
			}
			/* Put it back into the SCSI buffer list */
			lpfc_release_scsi_buf_s4(phba, psb);
		}
	}
	return rc;
}

/**
 * lpfc_new_scsi_buf_s4 - Scsi buffer allocator for HBA with SLI4 IF spec
 * @vport: The virtual port for which this call being executed.
 * @num_to_allocate: The requested number of buffers to allocate.
 *
 * This routine allocates a scsi buffer for device with SLI-4 interface spec,
 * the scsi buffer contains all the necessary information needed to initiate
 * a SCSI I/O.
 *
 * Return codes:
 *   int - number of scsi buffers that were allocated.
 *   0 = failure, less than num_to_alloc is a partial failure.
 **/
static int
lpfc_new_scsi_buf_s4(struct lpfc_vport *vport, int num_to_alloc)
{
	struct lpfc_hba *phba = vport->phba;
	struct lpfc_scsi_buf *psb;
	struct sli4_sge *sgl;
	IOCB_t *iocb;
	dma_addr_t pdma_phys_fcp_cmd;
	dma_addr_t pdma_phys_fcp_rsp;
	dma_addr_t pdma_phys_bpl, pdma_phys_bpl1;
	uint16_t iotag, last_xritag = NO_XRI;
	int status = 0, index;
	int bcnt;
	int non_sequential_xri = 0;
	int rc = 0;
	LIST_HEAD(sblist);

	for (bcnt = 0; bcnt < num_to_alloc; bcnt++) {
		psb = kzalloc(sizeof(struct lpfc_scsi_buf), GFP_KERNEL);
		if (!psb)
			break;

		/*
		 * Get memory from the pci pool to map the virt space to pci bus
		 * space for an I/O.  The DMA buffer includes space for the
		 * struct fcp_cmnd, struct fcp_rsp and the number of bde's
		 * necessary to support the sg_tablesize.
		 */
		psb->data = pci_pool_alloc(phba->lpfc_scsi_dma_buf_pool,
						GFP_KERNEL, &psb->dma_handle);
		if (!psb->data) {
			kfree(psb);
			break;
		}

		/* Initialize virtual ptrs to dma_buf region. */
		memset(psb->data, 0, phba->cfg_sg_dma_buf_size);

		/* Allocate iotag for psb->cur_iocbq. */
		iotag = lpfc_sli_next_iotag(phba, &psb->cur_iocbq);
		if (iotag == 0) {
			kfree(psb);
			break;
		}

		psb->cur_iocbq.sli4_xritag = lpfc_sli4_next_xritag(phba);
		if (psb->cur_iocbq.sli4_xritag == NO_XRI) {
			pci_pool_free(phba->lpfc_scsi_dma_buf_pool,
			      psb->data, psb->dma_handle);
			kfree(psb);
			break;
		}
		if (last_xritag != NO_XRI
			&& psb->cur_iocbq.sli4_xritag != (last_xritag+1)) {
			non_sequential_xri = 1;
		} else
			list_add_tail(&psb->list, &sblist);
		last_xritag = psb->cur_iocbq.sli4_xritag;

		index = phba->sli4_hba.scsi_xri_cnt++;
		psb->cur_iocbq.iocb_flag |= LPFC_IO_FCP;

		psb->fcp_bpl = psb->data;
		psb->fcp_cmnd = (psb->data + phba->cfg_sg_dma_buf_size)
			- (sizeof(struct fcp_cmnd) + sizeof(struct fcp_rsp));
		psb->fcp_rsp = (struct fcp_rsp *)((uint8_t *)psb->fcp_cmnd +
					sizeof(struct fcp_cmnd));

		/* Initialize local short-hand pointers. */
		sgl = (struct sli4_sge *)psb->fcp_bpl;
		pdma_phys_bpl = psb->dma_handle;
		pdma_phys_fcp_cmd =
			(psb->dma_handle + phba->cfg_sg_dma_buf_size)
			 - (sizeof(struct fcp_cmnd) + sizeof(struct fcp_rsp));
		pdma_phys_fcp_rsp = pdma_phys_fcp_cmd + sizeof(struct fcp_cmnd);

		/*
		 * The first two bdes are the FCP_CMD and FCP_RSP.  The balance
		 * are sg list bdes.  Initialize the first two and leave the
		 * rest for queuecommand.
		 */
		sgl->addr_hi = cpu_to_le32(putPaddrHigh(pdma_phys_fcp_cmd));
		sgl->addr_lo = cpu_to_le32(putPaddrLow(pdma_phys_fcp_cmd));
		bf_set(lpfc_sli4_sge_last, sgl, 0);
		sgl->word2 = cpu_to_le32(sgl->word2);
		sgl->sge_len = cpu_to_le32(sizeof(struct fcp_cmnd));
		sgl++;

		/* Setup the physical region for the FCP RSP */
		sgl->addr_hi = cpu_to_le32(putPaddrHigh(pdma_phys_fcp_rsp));
		sgl->addr_lo = cpu_to_le32(putPaddrLow(pdma_phys_fcp_rsp));
		bf_set(lpfc_sli4_sge_last, sgl, 1);
		sgl->word2 = cpu_to_le32(sgl->word2);
		sgl->sge_len = cpu_to_le32(sizeof(struct fcp_rsp));

		/*
		 * Since the IOCB for the FCP I/O is built into this
		 * lpfc_scsi_buf, initialize it with all known data now.
		 */
		iocb = &psb->cur_iocbq.iocb;
		iocb->un.fcpi64.bdl.ulpIoTag32 = 0;
		iocb->un.fcpi64.bdl.bdeFlags = BUFF_TYPE_BDE_64;
		/* setting the BLP size to 2 * sizeof BDE may not be correct.
		 * We are setting the bpl to point to out sgl. An sgl's
		 * entries are 16 bytes, a bpl entries are 12 bytes.
		 */
		iocb->un.fcpi64.bdl.bdeSize = sizeof(struct fcp_cmnd);
		iocb->un.fcpi64.bdl.addrLow = putPaddrLow(pdma_phys_fcp_cmd);
		iocb->un.fcpi64.bdl.addrHigh = putPaddrHigh(pdma_phys_fcp_cmd);
		iocb->ulpBdeCount = 1;
		iocb->ulpLe = 1;
		iocb->ulpClass = CLASS3;
		if (phba->cfg_sg_dma_buf_size > SGL_PAGE_SIZE)
			pdma_phys_bpl1 = pdma_phys_bpl + SGL_PAGE_SIZE;
		else
			pdma_phys_bpl1 = 0;
		psb->dma_phys_bpl = pdma_phys_bpl;
		phba->sli4_hba.lpfc_scsi_psb_array[index] = psb;
		if (non_sequential_xri) {
			status = lpfc_sli4_post_sgl(phba, pdma_phys_bpl,
						pdma_phys_bpl1,
						psb->cur_iocbq.sli4_xritag);
			if (status) {
				/* Put this back on the abort scsi list */
				psb->exch_busy = 1;
				rc++;
			} else {
				psb->exch_busy = 0;
				psb->status = IOSTAT_SUCCESS;
			}
			/* Put it back into the SCSI buffer list */
			lpfc_release_scsi_buf_s4(phba, psb);
			break;
		}
	}
	if (bcnt) {
		status = lpfc_sli4_post_scsi_sgl_block(phba, &sblist, bcnt);
		/* Reset SCSI buffer count for next round of posting */
		while (!list_empty(&sblist)) {
			list_remove_head(&sblist, psb, struct lpfc_scsi_buf,
				 list);
			if (status) {
				/* Put this back on the abort scsi list */
				psb->exch_busy = 1;
				rc++;
			} else {
				psb->exch_busy = 0;
				psb->status = IOSTAT_SUCCESS;
			}
			/* Put it back into the SCSI buffer list */
			lpfc_release_scsi_buf_s4(phba, psb);
		}
	}

	return bcnt + non_sequential_xri - rc;
}

/**
 * lpfc_new_scsi_buf - Wrapper funciton for scsi buffer allocator
 * @vport: The virtual port for which this call being executed.
 * @num_to_allocate: The requested number of buffers to allocate.
 *
 * This routine wraps the actual SCSI buffer allocator function pointer from
 * the lpfc_hba struct.
 *
 * Return codes:
 *   int - number of scsi buffers that were allocated.
 *   0 = failure, less than num_to_alloc is a partial failure.
 **/
static inline int
lpfc_new_scsi_buf(struct lpfc_vport *vport, int num_to_alloc)
{
	return vport->phba->lpfc_new_scsi_buf(vport, num_to_alloc);
}

/**
 * lpfc_get_scsi_buf - Get a scsi buffer from lpfc_scsi_buf_list of the HBA
 * @phba: The HBA for which this call is being executed.
 *
 * This routine removes a scsi buffer from head of @phba lpfc_scsi_buf_list list
 * and returns to caller.
 *
 * Return codes:
 *   NULL - Error
 *   Pointer to lpfc_scsi_buf - Success
 **/
static struct lpfc_scsi_buf*
lpfc_get_scsi_buf(struct lpfc_hba * phba)
{
	struct  lpfc_scsi_buf * lpfc_cmd = NULL;
	struct list_head *scsi_buf_list = &phba->lpfc_scsi_buf_list;
	unsigned long iflag = 0;

	spin_lock_irqsave(&phba->scsi_buf_list_lock, iflag);
	list_remove_head(scsi_buf_list, lpfc_cmd, struct lpfc_scsi_buf, list);
	if (lpfc_cmd) {
		lpfc_cmd->seg_cnt = 0;
		lpfc_cmd->nonsg_phys = 0;
		lpfc_cmd->prot_seg_cnt = 0;
	}
	spin_unlock_irqrestore(&phba->scsi_buf_list_lock, iflag);
	return  lpfc_cmd;
}

/**
 * lpfc_release_scsi_buf - Return a scsi buffer back to hba scsi buf list
 * @phba: The Hba for which this call is being executed.
 * @psb: The scsi buffer which is being released.
 *
 * This routine releases @psb scsi buffer by adding it to tail of @phba
 * lpfc_scsi_buf_list list.
 **/
static void
lpfc_release_scsi_buf_s3(struct lpfc_hba *phba, struct lpfc_scsi_buf *psb)
{
	unsigned long iflag = 0;

	spin_lock_irqsave(&phba->scsi_buf_list_lock, iflag);
	psb->pCmd = NULL;
	list_add_tail(&psb->list, &phba->lpfc_scsi_buf_list);
	spin_unlock_irqrestore(&phba->scsi_buf_list_lock, iflag);
}

/**
 * lpfc_release_scsi_buf_s4: Return a scsi buffer back to hba scsi buf list.
 * @phba: The Hba for which this call is being executed.
 * @psb: The scsi buffer which is being released.
 *
 * This routine releases @psb scsi buffer by adding it to tail of @phba
 * lpfc_scsi_buf_list list. For SLI4 XRI's are tied to the scsi buffer
 * and cannot be reused for at least RA_TOV amount of time if it was
 * aborted.
 **/
static void
lpfc_release_scsi_buf_s4(struct lpfc_hba *phba, struct lpfc_scsi_buf *psb)
{
	unsigned long iflag = 0;

	if (psb->exch_busy) {
		spin_lock_irqsave(&phba->sli4_hba.abts_scsi_buf_list_lock,
					iflag);
		psb->pCmd = NULL;
		list_add_tail(&psb->list,
			&phba->sli4_hba.lpfc_abts_scsi_buf_list);
		spin_unlock_irqrestore(&phba->sli4_hba.abts_scsi_buf_list_lock,
					iflag);
	} else {

		spin_lock_irqsave(&phba->scsi_buf_list_lock, iflag);
		psb->pCmd = NULL;
		list_add_tail(&psb->list, &phba->lpfc_scsi_buf_list);
		spin_unlock_irqrestore(&phba->scsi_buf_list_lock, iflag);
	}
}

/**
 * lpfc_release_scsi_buf: Return a scsi buffer back to hba scsi buf list.
 * @phba: The Hba for which this call is being executed.
 * @psb: The scsi buffer which is being released.
 *
 * This routine releases @psb scsi buffer by adding it to tail of @phba
 * lpfc_scsi_buf_list list.
 **/
static void
lpfc_release_scsi_buf(struct lpfc_hba *phba, struct lpfc_scsi_buf *psb)
{

	phba->lpfc_release_scsi_buf(phba, psb);
}

/**
 * lpfc_scsi_prep_dma_buf_s3 - DMA mapping for scsi buffer to SLI3 IF spec
 * @phba: The Hba for which this call is being executed.
 * @lpfc_cmd: The scsi buffer which is going to be mapped.
 *
 * This routine does the pci dma mapping for scatter-gather list of scsi cmnd
 * field of @lpfc_cmd for device with SLI-3 interface spec. This routine scans
 * through sg elements and format the bdea. This routine also initializes all
 * IOCB fields which are dependent on scsi command request buffer.
 *
 * Return codes:
 *   1 - Error
 *   0 - Success
 **/
static int
lpfc_scsi_prep_dma_buf_s3(struct lpfc_hba *phba, struct lpfc_scsi_buf *lpfc_cmd)
{
	struct scsi_cmnd *scsi_cmnd = lpfc_cmd->pCmd;
	struct scatterlist *sgel = NULL;
	struct fcp_cmnd *fcp_cmnd = lpfc_cmd->fcp_cmnd;
	struct ulp_bde64 *bpl = lpfc_cmd->fcp_bpl;
	struct lpfc_iocbq *iocbq = &lpfc_cmd->cur_iocbq;
	IOCB_t *iocb_cmd = &lpfc_cmd->cur_iocbq.iocb;
	struct ulp_bde64 *data_bde = iocb_cmd->unsli3.fcp_ext.dbde;
	dma_addr_t physaddr;
	uint32_t num_bde = 0;
	int nseg, datadir = scsi_cmnd->sc_data_direction;

	/*
	 * There are three possibilities here - use scatter-gather segment, use
	 * the single mapping, or neither.  Start the lpfc command prep by
	 * bumping the bpl beyond the fcp_cmnd and fcp_rsp regions to the first
	 * data bde entry.
	 */
	bpl += 2;
	if (scsi_sg_count(scsi_cmnd)) {
		/*
		 * The driver stores the segment count returned from pci_map_sg
		 * because this a count of dma-mappings used to map the use_sg
		 * pages.  They are not guaranteed to be the same for those
		 * architectures that implement an IOMMU.
		 */

		nseg = dma_map_sg(&phba->pcidev->dev, scsi_sglist(scsi_cmnd),
				  scsi_sg_count(scsi_cmnd), datadir);
		if (unlikely(!nseg))
			return 1;

		lpfc_cmd->seg_cnt = nseg;
		if (lpfc_cmd->seg_cnt > phba->cfg_sg_seg_cnt) {
			lpfc_printf_log(phba, KERN_ERR, LOG_BG,
				"9064 BLKGRD: %s: Too many sg segments from "
			       "dma_map_sg.  Config %d, seg_cnt %d\n",
			       __func__, phba->cfg_sg_seg_cnt,
			       lpfc_cmd->seg_cnt);
			scsi_dma_unmap(scsi_cmnd);
			return 1;
		}

		/*
		 * The driver established a maximum scatter-gather segment count
		 * during probe that limits the number of sg elements in any
		 * single scsi command.  Just run through the seg_cnt and format
		 * the bde's.
		 * When using SLI-3 the driver will try to fit all the BDEs into
		 * the IOCB. If it can't then the BDEs get added to a BPL as it
		 * does for SLI-2 mode.
		 */
		scsi_for_each_sg(scsi_cmnd, sgel, nseg, num_bde) {
			physaddr = sg_dma_address(sgel);
			if (phba->sli_rev == 3 &&
			    !(phba->sli3_options & LPFC_SLI3_BG_ENABLED) &&
			    !(iocbq->iocb_flag & DSS_SECURITY_OP) &&
			    nseg <= LPFC_EXT_DATA_BDE_COUNT) {
				data_bde->tus.f.bdeFlags = BUFF_TYPE_BDE_64;
				data_bde->tus.f.bdeSize = sg_dma_len(sgel);
				data_bde->addrLow = putPaddrLow(physaddr);
				data_bde->addrHigh = putPaddrHigh(physaddr);
				data_bde++;
			} else {
				bpl->tus.f.bdeFlags = BUFF_TYPE_BDE_64;
				bpl->tus.f.bdeSize = sg_dma_len(sgel);
				bpl->tus.w = le32_to_cpu(bpl->tus.w);
				bpl->addrLow =
					le32_to_cpu(putPaddrLow(physaddr));
				bpl->addrHigh =
					le32_to_cpu(putPaddrHigh(physaddr));
				bpl++;
			}
		}
	}

	/*
	 * Finish initializing those IOCB fields that are dependent on the
	 * scsi_cmnd request_buffer.  Note that for SLI-2 the bdeSize is
	 * explicitly reinitialized and for SLI-3 the extended bde count is
	 * explicitly reinitialized since all iocb memory resources are reused.
	 */
	if (phba->sli_rev == 3 &&
	    !(phba->sli3_options & LPFC_SLI3_BG_ENABLED) &&
	    !(iocbq->iocb_flag & DSS_SECURITY_OP)) {
		if (num_bde > LPFC_EXT_DATA_BDE_COUNT) {
			/*
			 * The extended IOCB format can only fit 3 BDE or a BPL.
			 * This I/O has more than 3 BDE so the 1st data bde will
			 * be a BPL that is filled in here.
			 */
			physaddr = lpfc_cmd->dma_handle;
			data_bde->tus.f.bdeFlags = BUFF_TYPE_BLP_64;
			data_bde->tus.f.bdeSize = (num_bde *
						   sizeof(struct ulp_bde64));
			physaddr += (sizeof(struct fcp_cmnd) +
				     sizeof(struct fcp_rsp) +
				     (2 * sizeof(struct ulp_bde64)));
			data_bde->addrHigh = putPaddrHigh(physaddr);
			data_bde->addrLow = putPaddrLow(physaddr);
			/* ebde count includes the responce bde and data bpl */
			iocb_cmd->unsli3.fcp_ext.ebde_count = 2;
		} else {
			/* ebde count includes the responce bde and data bdes */
			iocb_cmd->unsli3.fcp_ext.ebde_count = (num_bde + 1);
		}
	} else {
		iocb_cmd->un.fcpi64.bdl.bdeSize =
			((num_bde + 2) * sizeof(struct ulp_bde64));
		iocb_cmd->unsli3.fcp_ext.ebde_count = (num_bde + 1);
	}
	fcp_cmnd->fcpDl = cpu_to_be32(scsi_bufflen(scsi_cmnd));

	/*
	 * Due to difference in data length between DIF/non-DIF paths,
	 * we need to set word 4 of IOCB here
	 */
	iocb_cmd->un.fcpi.fcpi_parm = scsi_bufflen(scsi_cmnd);
	return 0;
}

/*
 * Given a scsi cmnd, determine the BlockGuard profile to be used
 * with the cmd
 */
static int
lpfc_sc_to_sli_prof(struct lpfc_hba *phba, struct scsi_cmnd *sc)
{
	uint8_t guard_type = scsi_host_get_guard(sc->device->host);
	uint8_t ret_prof = LPFC_PROF_INVALID;

	if (guard_type == SHOST_DIX_GUARD_IP) {
		switch (scsi_get_prot_op(sc)) {
		case SCSI_PROT_READ_INSERT:
		case SCSI_PROT_WRITE_STRIP:
			ret_prof = LPFC_PROF_AST2;
			break;

		case SCSI_PROT_READ_STRIP:
		case SCSI_PROT_WRITE_INSERT:
			ret_prof = LPFC_PROF_A1;
			break;

		case SCSI_PROT_READ_PASS:
		case SCSI_PROT_WRITE_PASS:
			ret_prof = LPFC_PROF_AST1;
			break;

		case SCSI_PROT_NORMAL:
		default:
			lpfc_printf_log(phba, KERN_ERR, LOG_BG,
				"9063 BLKGRD:Bad op/guard:%d/%d combination\n",
					scsi_get_prot_op(sc), guard_type);
			break;

		}
	} else if (guard_type == SHOST_DIX_GUARD_CRC) {
		switch (scsi_get_prot_op(sc)) {
		case SCSI_PROT_READ_STRIP:
		case SCSI_PROT_WRITE_INSERT:
			ret_prof = LPFC_PROF_A1;
			break;

		case SCSI_PROT_READ_PASS:
		case SCSI_PROT_WRITE_PASS:
			ret_prof = LPFC_PROF_C1;
			break;

		case SCSI_PROT_READ_INSERT:
		case SCSI_PROT_WRITE_STRIP:
		case SCSI_PROT_NORMAL:
		default:
			lpfc_printf_log(phba, KERN_ERR, LOG_BG,
				"9075 BLKGRD: Bad op/guard:%d/%d combination\n",
					scsi_get_prot_op(sc), guard_type);
			break;
		}
	} else {
		/* unsupported format */
		BUG();
	}

	return ret_prof;
}

struct scsi_dif_tuple {
	__be16 guard_tag;       /* Checksum */
	__be16 app_tag;         /* Opaque storage */
	__be32 ref_tag;         /* Target LBA or indirect LBA */
};

static inline unsigned
lpfc_cmd_blksize(struct scsi_cmnd *sc)
{
	return sc->device->sector_size;
}

/**
 * lpfc_get_cmd_dif_parms - Extract DIF parameters from SCSI command
 * @sc:             in: SCSI command
 * @apptagmask:     out: app tag mask
 * @apptagval:      out: app tag value
 * @reftag:         out: ref tag (reference tag)
 *
 * Description:
 *   Extract DIF parameters from the command if possible.  Otherwise,
 *   use default parameters.
 *
 **/
static inline void
lpfc_get_cmd_dif_parms(struct scsi_cmnd *sc, uint16_t *apptagmask,
		uint16_t *apptagval, uint32_t *reftag)
{
	struct  scsi_dif_tuple *spt;
	unsigned char op = scsi_get_prot_op(sc);
	unsigned int protcnt = scsi_prot_sg_count(sc);
	static int cnt;

	if (protcnt && (op == SCSI_PROT_WRITE_STRIP ||
				op == SCSI_PROT_WRITE_PASS)) {

		cnt++;
		spt = page_address(sg_page(scsi_prot_sglist(sc))) +
			scsi_prot_sglist(sc)[0].offset;
		*apptagmask = 0;
		*apptagval = 0;
		*reftag = cpu_to_be32(spt->ref_tag);

	} else {
		/* SBC defines ref tag to be lower 32bits of LBA */
		*reftag = (uint32_t) (0xffffffff & scsi_get_lba(sc));
		*apptagmask = 0;
		*apptagval = 0;
	}
}

/*
 * This function sets up buffer list for protection groups of
 * type LPFC_PG_TYPE_NO_DIF
 *
 * This is usually used when the HBA is instructed to generate
 * DIFs and insert them into data stream (or strip DIF from
 * incoming data stream)
 *
 * The buffer list consists of just one protection group described
 * below:
 *                                +-------------------------+
 *   start of prot group  -->     |          PDE_1          |
 *                                +-------------------------+
 *                                |         Data BDE        |
 *                                +-------------------------+
 *                                |more Data BDE's ... (opt)|
 *                                +-------------------------+
 *
 * @sc: pointer to scsi command we're working on
 * @bpl: pointer to buffer list for protection groups
 * @datacnt: number of segments of data that have been dma mapped
 *
 * Note: Data s/g buffers have been dma mapped
 */
static int
lpfc_bg_setup_bpl(struct lpfc_hba *phba, struct scsi_cmnd *sc,
		struct ulp_bde64 *bpl, int datasegcnt)
{
	struct scatterlist *sgde = NULL; /* s/g data entry */
	struct lpfc_pde *pde1 = NULL;
	dma_addr_t physaddr;
	int i = 0, num_bde = 0;
	int datadir = sc->sc_data_direction;
	int prof = LPFC_PROF_INVALID;
	unsigned blksize;
	uint32_t reftag;
	uint16_t apptagmask, apptagval;

	pde1 = (struct lpfc_pde *) bpl;
	prof = lpfc_sc_to_sli_prof(phba, sc);

	if (prof == LPFC_PROF_INVALID)
		goto out;

	/* extract some info from the scsi command for PDE1*/
	blksize = lpfc_cmd_blksize(sc);
	lpfc_get_cmd_dif_parms(sc, &apptagmask, &apptagval, &reftag);

	/* setup PDE1 with what we have */
	lpfc_pde_set_bg_parms(pde1, LPFC_PDE1_DESCRIPTOR, prof, blksize,
			BG_EC_STOP_ERR);
	lpfc_pde_set_dif_parms(pde1, apptagmask, apptagval, reftag);

	num_bde++;
	bpl++;

	/* assumption: caller has already run dma_map_sg on command data */
	scsi_for_each_sg(sc, sgde, datasegcnt, i) {
		physaddr = sg_dma_address(sgde);
		bpl->addrLow = le32_to_cpu(putPaddrLow(physaddr));
		bpl->addrHigh = le32_to_cpu(putPaddrHigh(physaddr));
		bpl->tus.f.bdeSize = sg_dma_len(sgde);
		if (datadir == DMA_TO_DEVICE)
			bpl->tus.f.bdeFlags = BUFF_TYPE_BDE_64;
		else
			bpl->tus.f.bdeFlags = BUFF_TYPE_BDE_64I;
		bpl->tus.w = le32_to_cpu(bpl->tus.w);
		bpl++;
		num_bde++;
	}

out:
	return num_bde;
}

/*
 * This function sets up buffer list for protection groups of
 * type LPFC_PG_TYPE_DIF_BUF
 *
 * This is usually used when DIFs are in their own buffers,
 * separate from the data. The HBA can then by instructed
 * to place the DIFs in the outgoing stream.  For read operations,
 * The HBA could extract the DIFs and place it in DIF buffers.
 *
 * The buffer list for this type consists of one or more of the
 * protection groups described below:
 *                                    +-------------------------+
 *   start of first prot group  -->   |          PDE_1          |
 *                                    +-------------------------+
 *                                    |      PDE_3 (Prot BDE)   |
 *                                    +-------------------------+
 *                                    |        Data BDE         |
 *                                    +-------------------------+
 *                                    |more Data BDE's ... (opt)|
 *                                    +-------------------------+
 *   start of new  prot group  -->    |          PDE_1          |
 *                                    +-------------------------+
 *                                    |          ...            |
 *                                    +-------------------------+
 *
 * @sc: pointer to scsi command we're working on
 * @bpl: pointer to buffer list for protection groups
 * @datacnt: number of segments of data that have been dma mapped
 * @protcnt: number of segment of protection data that have been dma mapped
 *
 * Note: It is assumed that both data and protection s/g buffers have been
 *       mapped for DMA
 */
static int
lpfc_bg_setup_bpl_prot(struct lpfc_hba *phba, struct scsi_cmnd *sc,
		struct ulp_bde64 *bpl, int datacnt, int protcnt)
{
	struct scatterlist *sgde = NULL; /* s/g data entry */
	struct scatterlist *sgpe = NULL; /* s/g prot entry */
	struct lpfc_pde *pde1 = NULL;
	struct ulp_bde64 *prot_bde = NULL;
	dma_addr_t dataphysaddr, protphysaddr;
	unsigned short curr_data = 0, curr_prot = 0;
	unsigned int split_offset, protgroup_len;
	unsigned int protgrp_blks, protgrp_bytes;
	unsigned int remainder, subtotal;
	int prof = LPFC_PROF_INVALID;
	int datadir = sc->sc_data_direction;
	unsigned char pgdone = 0, alldone = 0;
	unsigned blksize;
	uint32_t reftag;
	uint16_t apptagmask, apptagval;
	int num_bde = 0;

	sgpe = scsi_prot_sglist(sc);
	sgde = scsi_sglist(sc);

	if (!sgpe || !sgde) {
		lpfc_printf_log(phba, KERN_ERR, LOG_FCP,
				"9020 Invalid s/g entry: data=0x%p prot=0x%p\n",
				sgpe, sgde);
		return 0;
	}

	prof = lpfc_sc_to_sli_prof(phba, sc);
	if (prof == LPFC_PROF_INVALID)
		goto out;

	/* extract some info from the scsi command for PDE1*/
	blksize = lpfc_cmd_blksize(sc);
	lpfc_get_cmd_dif_parms(sc, &apptagmask, &apptagval, &reftag);

	split_offset = 0;
	do {
		/* setup the first PDE_1 */
		pde1 = (struct lpfc_pde *) bpl;

		lpfc_pde_set_bg_parms(pde1, LPFC_PDE1_DESCRIPTOR, prof, blksize,
				BG_EC_STOP_ERR);
		lpfc_pde_set_dif_parms(pde1, apptagmask, apptagval, reftag);

		num_bde++;
		bpl++;

		/* setup the first BDE that points to protection buffer */
		prot_bde = (struct ulp_bde64 *) bpl;
		protphysaddr = sg_dma_address(sgpe);
		prot_bde->addrLow = le32_to_cpu(putPaddrLow(protphysaddr));
		prot_bde->addrHigh = le32_to_cpu(putPaddrHigh(protphysaddr));
		protgroup_len = sg_dma_len(sgpe);


		/* must be integer multiple of the DIF block length */
		BUG_ON(protgroup_len % 8);

		protgrp_blks = protgroup_len / 8;
		protgrp_bytes = protgrp_blks * blksize;

		prot_bde->tus.f.bdeSize = protgroup_len;
		if (datadir == DMA_TO_DEVICE)
			prot_bde->tus.f.bdeFlags = BUFF_TYPE_BDE_64;
		else
			prot_bde->tus.f.bdeFlags = BUFF_TYPE_BDE_64I;
		prot_bde->tus.w = le32_to_cpu(bpl->tus.w);

		curr_prot++;
		num_bde++;

		/* setup BDE's for data blocks associated with DIF data */
		pgdone = 0;
		subtotal = 0; /* total bytes processed for current prot grp */
		while (!pgdone) {
			if (!sgde) {
				lpfc_printf_log(phba, KERN_ERR, LOG_BG,
					"9065 BLKGRD:%s Invalid data segment\n",
						__func__);
				return 0;
			}
			bpl++;
			dataphysaddr = sg_dma_address(sgde) + split_offset;
			bpl->addrLow = le32_to_cpu(putPaddrLow(dataphysaddr));
			bpl->addrHigh = le32_to_cpu(putPaddrHigh(dataphysaddr));

			remainder = sg_dma_len(sgde) - split_offset;

			if ((subtotal + remainder) <= protgrp_bytes) {
				/* we can use this whole buffer */
				bpl->tus.f.bdeSize = remainder;
				split_offset = 0;

				if ((subtotal + remainder) == protgrp_bytes)
					pgdone = 1;
			} else {
				/* must split this buffer with next prot grp */
				bpl->tus.f.bdeSize = protgrp_bytes - subtotal;
				split_offset += bpl->tus.f.bdeSize;
			}

			subtotal += bpl->tus.f.bdeSize;

			if (datadir == DMA_TO_DEVICE)
				bpl->tus.f.bdeFlags = BUFF_TYPE_BDE_64;
			else
				bpl->tus.f.bdeFlags = BUFF_TYPE_BDE_64I;
			bpl->tus.w = le32_to_cpu(bpl->tus.w);

			num_bde++;
			curr_data++;

			if (split_offset)
				break;

			/* Move to the next s/g segment if possible */
			sgde = sg_next(sgde);
		}

		/* are we done ? */
		if (curr_prot == protcnt) {
			alldone = 1;
		} else if (curr_prot < protcnt) {
			/* advance to next prot buffer */
			sgpe = sg_next(sgpe);
			bpl++;

			/* update the reference tag */
			reftag += protgrp_blks;
		} else {
			/* if we're here, we have a bug */
			lpfc_printf_log(phba, KERN_ERR, LOG_BG,
				"9054 BLKGRD: bug in %s\n", __func__);
		}

	} while (!alldone);

out:


	return num_bde;
}
/*
 * Given a SCSI command that supports DIF, determine composition of protection
 * groups involved in setting up buffer lists
 *
 * Returns:
 *			      for DIF (for both read and write)
 * */
static int
lpfc_prot_group_type(struct lpfc_hba *phba, struct scsi_cmnd *sc)
{
	int ret = LPFC_PG_TYPE_INVALID;
	unsigned char op = scsi_get_prot_op(sc);

	switch (op) {
	case SCSI_PROT_READ_STRIP:
	case SCSI_PROT_WRITE_INSERT:
		ret = LPFC_PG_TYPE_NO_DIF;
		break;
	case SCSI_PROT_READ_INSERT:
	case SCSI_PROT_WRITE_STRIP:
	case SCSI_PROT_READ_PASS:
	case SCSI_PROT_WRITE_PASS:
		ret = LPFC_PG_TYPE_DIF_BUF;
		break;
	default:
		lpfc_printf_log(phba, KERN_ERR, LOG_FCP,
				"9021 Unsupported protection op:%d\n", op);
		break;
	}

	return ret;
}

/*
 * This is the protection/DIF aware version of
 * lpfc_scsi_prep_dma_buf(). It may be a good idea to combine the
 * two functions eventually, but for now, it's here
 */
static int
lpfc_bg_scsi_prep_dma_buf(struct lpfc_hba *phba,
		struct lpfc_scsi_buf *lpfc_cmd)
{
	struct scsi_cmnd *scsi_cmnd = lpfc_cmd->pCmd;
	struct fcp_cmnd *fcp_cmnd = lpfc_cmd->fcp_cmnd;
	struct ulp_bde64 *bpl = lpfc_cmd->fcp_bpl;
	IOCB_t *iocb_cmd = &lpfc_cmd->cur_iocbq.iocb;
	uint32_t num_bde = 0;
	int datasegcnt, protsegcnt, datadir = scsi_cmnd->sc_data_direction;
	int prot_group_type = 0;
	int diflen, fcpdl;
	unsigned blksize;

	/*
	 * Start the lpfc command prep by bumping the bpl beyond fcp_cmnd
	 *  fcp_rsp regions to the first data bde entry
	 */
	bpl += 2;
	if (scsi_sg_count(scsi_cmnd)) {
		/*
		 * The driver stores the segment count returned from pci_map_sg
		 * because this a count of dma-mappings used to map the use_sg
		 * pages.  They are not guaranteed to be the same for those
		 * architectures that implement an IOMMU.
		 */
		datasegcnt = dma_map_sg(&phba->pcidev->dev,
					scsi_sglist(scsi_cmnd),
					scsi_sg_count(scsi_cmnd), datadir);
		if (unlikely(!datasegcnt))
			return 1;

		lpfc_cmd->seg_cnt = datasegcnt;
		if (lpfc_cmd->seg_cnt > phba->cfg_sg_seg_cnt) {
			lpfc_printf_log(phba, KERN_ERR, LOG_BG,
					"9067 BLKGRD: %s: Too many sg segments"
					" from dma_map_sg.  Config %d, seg_cnt"
					" %d\n",
					__func__, phba->cfg_sg_seg_cnt,
					lpfc_cmd->seg_cnt);
			scsi_dma_unmap(scsi_cmnd);
			return 1;
		}

		prot_group_type = lpfc_prot_group_type(phba, scsi_cmnd);

		switch (prot_group_type) {
		case LPFC_PG_TYPE_NO_DIF:
			num_bde = lpfc_bg_setup_bpl(phba, scsi_cmnd, bpl,
					datasegcnt);
			/* we should have 2 or more entries in buffer list */
			if (num_bde < 2)
				goto err;
			break;
		case LPFC_PG_TYPE_DIF_BUF:{
			/*
			 * This type indicates that protection buffers are
			 * passed to the driver, so that needs to be prepared
			 * for DMA
			 */
			protsegcnt = dma_map_sg(&phba->pcidev->dev,
					scsi_prot_sglist(scsi_cmnd),
					scsi_prot_sg_count(scsi_cmnd), datadir);
			if (unlikely(!protsegcnt)) {
				scsi_dma_unmap(scsi_cmnd);
				return 1;
			}

			lpfc_cmd->prot_seg_cnt = protsegcnt;
			if (lpfc_cmd->prot_seg_cnt
			    > phba->cfg_prot_sg_seg_cnt) {
				lpfc_printf_log(phba, KERN_ERR, LOG_BG,
					"9068 BLKGRD: %s: Too many prot sg "
					"segments from dma_map_sg.  Config %d,"
						"prot_seg_cnt %d\n", __func__,…