[SCSI] qla2xxx: Consolidated IOCB processing routines.

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: Chad Dupuis <chad.dupuis@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
diff --git a/drivers/scsi/qla2xxx/qla_iocb.c b/drivers/scsi/qla2xxx/qla_iocb.c
index 64a3075..d2cc0cc 100644
--- a/drivers/scsi/qla2xxx/qla_iocb.c
+++ b/drivers/scsi/qla2xxx/qla_iocb.c
@@ -11,8 +11,6 @@
 
 #include <scsi/scsi_tcq.h>
 
-static void qla2x00_isp_cmd(struct scsi_qla_host *, struct req_que *);
-
 static void qla25xx_set_que(srb_t *, struct rsp_que **);
 /**
  * qla2x00_get_cmd_direction() - Determine control_flag data direction.
@@ -468,6 +466,42 @@
 }
 
 /**
+ * qla2x00_start_iocbs() - Execute the IOCB command
+ */
+static void
+qla2x00_start_iocbs(struct scsi_qla_host *vha, struct req_que *req)
+{
+	struct qla_hw_data *ha = vha->hw;
+	device_reg_t __iomem *reg = ISP_QUE_REG(ha, req->id);
+	struct device_reg_2xxx __iomem *ioreg = &ha->iobase->isp;
+
+	if (IS_QLA82XX(ha)) {
+		qla82xx_start_iocbs(vha);
+	} else {
+		/* Adjust ring index. */
+		req->ring_index++;
+		if (req->ring_index == req->length) {
+			req->ring_index = 0;
+			req->ring_ptr = req->ring;
+		} else
+			req->ring_ptr++;
+
+		/* Set chip new ring index. */
+		if (ha->mqenable) {
+			WRT_REG_DWORD(&reg->isp25mq.req_q_in, req->ring_index);
+			RD_REG_DWORD(&ioreg->hccr);
+		} else if (IS_FWI2_CAPABLE(ha)) {
+			WRT_REG_DWORD(&reg->isp24.req_q_in, req->ring_index);
+			RD_REG_DWORD_RELAXED(&reg->isp24.req_q_in);
+		} else {
+			WRT_REG_WORD(ISP_REQ_Q_IN(ha, &reg->isp),
+				req->ring_index);
+			RD_REG_WORD_RELAXED(ISP_REQ_Q_IN(ha, &reg->isp));
+		}
+	}
+}
+
+/**
  * qla2x00_marker() - Send a marker IOCB to the firmware.
  * @ha: HA context
  * @loop_id: loop ID
@@ -516,7 +550,7 @@
 	}
 	wmb();
 
-	qla2x00_isp_cmd(vha, req);
+	qla2x00_start_iocbs(vha, req);
 
 	return (QLA_SUCCESS);
 }
@@ -537,68 +571,6 @@
 }
 
 /**
- * qla2x00_isp_cmd() - Modify the request ring pointer.
- * @ha: HA context
- *
- * Note: The caller must hold the hardware lock before calling this routine.
- */
-static void
-qla2x00_isp_cmd(struct scsi_qla_host *vha, struct req_que *req)
-{
-	struct qla_hw_data *ha = vha->hw;
-	device_reg_t __iomem *reg = ISP_QUE_REG(ha, req->id);
-	struct device_reg_2xxx __iomem *ioreg = &ha->iobase->isp;
-
-	ql_dbg(ql_dbg_io + ql_dbg_buffer, vha, 0x302d,
-	    "IOCB data:\n");
-	ql_dump_buffer(ql_dbg_io + ql_dbg_buffer, vha, 0x302e,
-	    (uint8_t *)req->ring_ptr, REQUEST_ENTRY_SIZE);
-
-	/* Adjust ring index. */
-	req->ring_index++;
-	if (req->ring_index == req->length) {
-		req->ring_index = 0;
-		req->ring_ptr = req->ring;
-	} else
-		req->ring_ptr++;
-
-	/* Set chip new ring index. */
-	if (IS_QLA82XX(ha)) {
-		uint32_t dbval = 0x04 | (ha->portnum << 5);
-
-		/* write, read and verify logic */
-		dbval = dbval | (req->id << 8) | (req->ring_index << 16);
-		if (ql2xdbwr)
-			qla82xx_wr_32(ha, ha->nxdb_wr_ptr, dbval);
-		else {
-			WRT_REG_DWORD(
-				(unsigned long __iomem *)ha->nxdb_wr_ptr,
-				dbval);
-			wmb();
-			while (RD_REG_DWORD(ha->nxdb_rd_ptr) != dbval) {
-				WRT_REG_DWORD((unsigned long __iomem *)
-					ha->nxdb_wr_ptr, dbval);
-				wmb();
-			}
-		}
-	} else if (ha->mqenable) {
-		/* Set chip new ring index. */
-		WRT_REG_DWORD(&reg->isp25mq.req_q_in, req->ring_index);
-		RD_REG_DWORD(&ioreg->hccr);
-	} else {
-		if (IS_FWI2_CAPABLE(ha)) {
-			WRT_REG_DWORD(&reg->isp24.req_q_in, req->ring_index);
-			RD_REG_DWORD_RELAXED(&reg->isp24.req_q_in);
-		} else {
-			WRT_REG_WORD(ISP_REQ_Q_IN(ha, &reg->isp),
-				req->ring_index);
-			RD_REG_WORD_RELAXED(ISP_REQ_Q_IN(ha, &reg->isp));
-		}
-	}
-
-}
-
-/**
  * qla24xx_calc_iocbs() - Determine number of Command Type 3 and
  * Continuation Type 1 IOCBs to allocate.
  *
@@ -620,6 +592,119 @@
 	return iocbs;
 }
 
+static inline int
+qla24xx_build_scsi_type_6_iocbs(srb_t *sp, struct cmd_type_6 *cmd_pkt,
+	uint16_t tot_dsds)
+{
+	uint32_t *cur_dsd = NULL;
+	scsi_qla_host_t	*vha;
+	struct qla_hw_data *ha;
+	struct scsi_cmnd *cmd;
+	struct	scatterlist *cur_seg;
+	uint32_t *dsd_seg;
+	void *next_dsd;
+	uint8_t avail_dsds;
+	uint8_t first_iocb = 1;
+	uint32_t dsd_list_len;
+	struct dsd_dma *dsd_ptr;
+	struct ct6_dsd *ctx;
+
+	cmd = sp->cmd;
+
+	/* Update entry type to indicate Command Type 3 IOCB */
+	*((uint32_t *)(&cmd_pkt->entry_type)) =
+		__constant_cpu_to_le32(COMMAND_TYPE_6);
+
+	/* No data transfer */
+	if (!scsi_bufflen(cmd) || cmd->sc_data_direction == DMA_NONE) {
+		cmd_pkt->byte_count = __constant_cpu_to_le32(0);
+		return 0;
+	}
+
+	vha = sp->fcport->vha;
+	ha = vha->hw;
+
+	/* Set transfer direction */
+	if (cmd->sc_data_direction == DMA_TO_DEVICE) {
+		cmd_pkt->control_flags =
+		    __constant_cpu_to_le16(CF_WRITE_DATA);
+		ha->qla_stats.output_bytes += scsi_bufflen(cmd);
+	} else if (cmd->sc_data_direction == DMA_FROM_DEVICE) {
+		cmd_pkt->control_flags =
+		    __constant_cpu_to_le16(CF_READ_DATA);
+		ha->qla_stats.input_bytes += scsi_bufflen(cmd);
+	}
+
+	cur_seg = scsi_sglist(cmd);
+	ctx = sp->ctx;
+
+	while (tot_dsds) {
+		avail_dsds = (tot_dsds > QLA_DSDS_PER_IOCB) ?
+		    QLA_DSDS_PER_IOCB : tot_dsds;
+		tot_dsds -= avail_dsds;
+		dsd_list_len = (avail_dsds + 1) * QLA_DSD_SIZE;
+
+		dsd_ptr = list_first_entry(&ha->gbl_dsd_list,
+		    struct dsd_dma, list);
+		next_dsd = dsd_ptr->dsd_addr;
+		list_del(&dsd_ptr->list);
+		ha->gbl_dsd_avail--;
+		list_add_tail(&dsd_ptr->list, &ctx->dsd_list);
+		ctx->dsd_use_cnt++;
+		ha->gbl_dsd_inuse++;
+
+		if (first_iocb) {
+			first_iocb = 0;
+			dsd_seg = (uint32_t *)&cmd_pkt->fcp_data_dseg_address;
+			*dsd_seg++ = cpu_to_le32(LSD(dsd_ptr->dsd_list_dma));
+			*dsd_seg++ = cpu_to_le32(MSD(dsd_ptr->dsd_list_dma));
+			cmd_pkt->fcp_data_dseg_len = cpu_to_le32(dsd_list_len);
+		} else {
+			*cur_dsd++ = cpu_to_le32(LSD(dsd_ptr->dsd_list_dma));
+			*cur_dsd++ = cpu_to_le32(MSD(dsd_ptr->dsd_list_dma));
+			*cur_dsd++ = cpu_to_le32(dsd_list_len);
+		}
+		cur_dsd = (uint32_t *)next_dsd;
+		while (avail_dsds) {
+			dma_addr_t	sle_dma;
+
+			sle_dma = sg_dma_address(cur_seg);
+			*cur_dsd++ = cpu_to_le32(LSD(sle_dma));
+			*cur_dsd++ = cpu_to_le32(MSD(sle_dma));
+			*cur_dsd++ = cpu_to_le32(sg_dma_len(cur_seg));
+			cur_seg = sg_next(cur_seg);
+			avail_dsds--;
+		}
+	}
+
+	/* Null termination */
+	*cur_dsd++ =  0;
+	*cur_dsd++ = 0;
+	*cur_dsd++ = 0;
+	cmd_pkt->control_flags |= CF_DATA_SEG_DESCR_ENABLE;
+	return 0;
+}
+
+/*
+ * qla24xx_calc_dsd_lists() - Determine number of DSD list required
+ * for Command Type 6.
+ *
+ * @dsds: number of data segment decriptors needed
+ *
+ * Returns the number of dsd list needed to store @dsds.
+ */
+inline uint16_t
+qla24xx_calc_dsd_lists(uint16_t dsds)
+{
+	uint16_t dsd_lists = 0;
+
+	dsd_lists = (dsds/QLA_DSDS_PER_IOCB);
+	if (dsds % QLA_DSDS_PER_IOCB)
+		dsd_lists++;
+	return dsd_lists;
+}
+
+
 /**
  * qla24xx_build_scsi_iocbs() - Build IOCB command utilizing Command Type 7
  * IOCB types.
@@ -946,6 +1031,7 @@
 	*cur_dsd++ = 0;
 	return 0;
 }
+
 static int
 qla24xx_walk_and_build_sglist(struct qla_hw_data *ha, srb_t *sp, uint32_t *dsd,
 	uint16_t tot_dsds)
@@ -1794,42 +1880,6 @@
 }
 
 static void
-qla2x00_start_iocbs(srb_t *sp)
-{
-	struct qla_hw_data *ha = sp->fcport->vha->hw;
-	struct req_que *req = ha->req_q_map[0];
-	device_reg_t __iomem *reg = ISP_QUE_REG(ha, req->id);
-	struct device_reg_2xxx __iomem *ioreg = &ha->iobase->isp;
-
-	if (IS_QLA82XX(ha)) {
-		qla82xx_start_iocbs(sp);
-	} else {
-		/* Adjust ring index. */
-		req->ring_index++;
-		if (req->ring_index == req->length) {
-			req->ring_index = 0;
-			req->ring_ptr = req->ring;
-		} else
-			req->ring_ptr++;
-
-		/* Set chip new ring index. */
-		if (ha->mqenable) {
-			WRT_REG_DWORD(&reg->isp25mq.req_q_in, req->ring_index);
-			RD_REG_DWORD(&ioreg->hccr);
-		} else if (IS_QLA82XX(ha)) {
-			qla82xx_start_iocbs(sp);
-		} else if (IS_FWI2_CAPABLE(ha)) {
-			WRT_REG_DWORD(&reg->isp24.req_q_in, req->ring_index);
-			RD_REG_DWORD_RELAXED(&reg->isp24.req_q_in);
-		} else {
-			WRT_REG_WORD(ISP_REQ_Q_IN(ha, &reg->isp),
-				req->ring_index);
-			RD_REG_WORD_RELAXED(ISP_REQ_Q_IN(ha, &reg->isp));
-		}
-	}
-}
-
-static void
 qla24xx_login_iocb(srb_t *sp, struct logio_entry_24xx *logio)
 {
 	struct srb_ctx *ctx = sp->ctx;
@@ -2161,6 +2211,372 @@
         ct_iocb->entry_count = entry_count;
 }
 
+/*
+ * qla82xx_start_scsi() - Send a SCSI command to the ISP
+ * @sp: command to send to the ISP
+ *
+ * Returns non-zero if a failure occurred, else zero.
+ */
+int
+qla82xx_start_scsi(srb_t *sp)
+{
+	int		ret, nseg;
+	unsigned long   flags;
+	struct scsi_cmnd *cmd;
+	uint32_t	*clr_ptr;
+	uint32_t        index;
+	uint32_t	handle;
+	uint16_t	cnt;
+	uint16_t	req_cnt;
+	uint16_t	tot_dsds;
+	struct device_reg_82xx __iomem *reg;
+	uint32_t dbval;
+	uint32_t *fcp_dl;
+	uint8_t additional_cdb_len;
+	struct ct6_dsd *ctx;
+	struct scsi_qla_host *vha = sp->fcport->vha;
+	struct qla_hw_data *ha = vha->hw;
+	struct req_que *req = NULL;
+	struct rsp_que *rsp = NULL;
+	char		tag[2];
+
+	/* Setup device pointers. */
+	ret = 0;
+	reg = &ha->iobase->isp82;
+	cmd = sp->cmd;
+	req = vha->req;
+	rsp = ha->rsp_q_map[0];
+
+	/* So we know we haven't pci_map'ed anything yet */
+	tot_dsds = 0;
+
+	dbval = 0x04 | (ha->portnum << 5);
+
+	/* Send marker if required */
+	if (vha->marker_needed != 0) {
+		if (qla2x00_marker(vha, req,
+			rsp, 0, 0, MK_SYNC_ALL) != QLA_SUCCESS) {
+			ql_log(ql_log_warn, vha, 0x300c,
+			    "qla2x00_marker failed for cmd=%p.\n", cmd);
+			return QLA_FUNCTION_FAILED;
+		}
+		vha->marker_needed = 0;
+	}
+
+	/* Acquire ring specific lock */
+	spin_lock_irqsave(&ha->hardware_lock, flags);
+
+	/* Check for room in outstanding command list. */
+	handle = req->current_outstanding_cmd;
+	for (index = 1; index < MAX_OUTSTANDING_COMMANDS; index++) {
+		handle++;
+		if (handle == MAX_OUTSTANDING_COMMANDS)
+			handle = 1;
+		if (!req->outstanding_cmds[handle])
+			break;
+	}
+	if (index == MAX_OUTSTANDING_COMMANDS)
+		goto queuing_error;
+
+	/* Map the sg table so we have an accurate count of sg entries needed */
+	if (scsi_sg_count(cmd)) {
+		nseg = dma_map_sg(&ha->pdev->dev, scsi_sglist(cmd),
+		    scsi_sg_count(cmd), cmd->sc_data_direction);
+		if (unlikely(!nseg))
+			goto queuing_error;
+	} else
+		nseg = 0;
+
+	tot_dsds = nseg;
+
+	if (tot_dsds > ql2xshiftctondsd) {
+		struct cmd_type_6 *cmd_pkt;
+		uint16_t more_dsd_lists = 0;
+		struct dsd_dma *dsd_ptr;
+		uint16_t i;
+
+		more_dsd_lists = qla24xx_calc_dsd_lists(tot_dsds);
+		if ((more_dsd_lists + ha->gbl_dsd_inuse) >= NUM_DSD_CHAIN) {
+			ql_dbg(ql_dbg_io, vha, 0x300d,
+			    "Num of DSD list %d is than %d for cmd=%p.\n",
+			    more_dsd_lists + ha->gbl_dsd_inuse, NUM_DSD_CHAIN,
+			    cmd);
+			goto queuing_error;
+		}
+
+		if (more_dsd_lists <= ha->gbl_dsd_avail)
+			goto sufficient_dsds;
+		else
+			more_dsd_lists -= ha->gbl_dsd_avail;
+
+		for (i = 0; i < more_dsd_lists; i++) {
+			dsd_ptr = kzalloc(sizeof(struct dsd_dma), GFP_ATOMIC);
+			if (!dsd_ptr) {
+				ql_log(ql_log_fatal, vha, 0x300e,
+				    "Failed to allocate memory for dsd_dma "
+				    "for cmd=%p.\n", cmd);
+				goto queuing_error;
+			}
+
+			dsd_ptr->dsd_addr = dma_pool_alloc(ha->dl_dma_pool,
+				GFP_ATOMIC, &dsd_ptr->dsd_list_dma);
+			if (!dsd_ptr->dsd_addr) {
+				kfree(dsd_ptr);
+				ql_log(ql_log_fatal, vha, 0x300f,
+				    "Failed to allocate memory for dsd_addr "
+				    "for cmd=%p.\n", cmd);
+				goto queuing_error;
+			}
+			list_add_tail(&dsd_ptr->list, &ha->gbl_dsd_list);
+			ha->gbl_dsd_avail++;
+		}
+
+sufficient_dsds:
+		req_cnt = 1;
+
+		if (req->cnt < (req_cnt + 2)) {
+			cnt = (uint16_t)RD_REG_DWORD_RELAXED(
+				&reg->req_q_out[0]);
+			if (req->ring_index < cnt)
+				req->cnt = cnt - req->ring_index;
+			else
+				req->cnt = req->length -
+					(req->ring_index - cnt);
+		}
+
+		if (req->cnt < (req_cnt + 2))
+			goto queuing_error;
+
+		ctx = sp->ctx = mempool_alloc(ha->ctx_mempool, GFP_ATOMIC);
+		if (!sp->ctx) {
+			ql_log(ql_log_fatal, vha, 0x3010,
+			    "Failed to allocate ctx for cmd=%p.\n", cmd);
+			goto queuing_error;
+		}
+		memset(ctx, 0, sizeof(struct ct6_dsd));
+		ctx->fcp_cmnd = dma_pool_alloc(ha->fcp_cmnd_dma_pool,
+			GFP_ATOMIC, &ctx->fcp_cmnd_dma);
+		if (!ctx->fcp_cmnd) {
+			ql_log(ql_log_fatal, vha, 0x3011,
+			    "Failed to allocate fcp_cmnd for cmd=%p.\n", cmd);
+			goto queuing_error_fcp_cmnd;
+		}
+
+		/* Initialize the DSD list and dma handle */
+		INIT_LIST_HEAD(&ctx->dsd_list);
+		ctx->dsd_use_cnt = 0;
+
+		if (cmd->cmd_len > 16) {
+			additional_cdb_len = cmd->cmd_len - 16;
+			if ((cmd->cmd_len % 4) != 0) {
+				/* SCSI command bigger than 16 bytes must be
+				 * multiple of 4
+				 */
+				ql_log(ql_log_warn, vha, 0x3012,
+				    "scsi cmd len %d not multiple of 4 "
+				    "for cmd=%p.\n", cmd->cmd_len, cmd);
+				goto queuing_error_fcp_cmnd;
+			}
+			ctx->fcp_cmnd_len = 12 + cmd->cmd_len + 4;
+		} else {
+			additional_cdb_len = 0;
+			ctx->fcp_cmnd_len = 12 + 16 + 4;
+		}
+
+		cmd_pkt = (struct cmd_type_6 *)req->ring_ptr;
+		cmd_pkt->handle = MAKE_HANDLE(req->id, handle);
+
+		/* Zero out remaining portion of packet. */
+		/*    tagged queuing modifier -- default is TSK_SIMPLE (0). */
+		clr_ptr = (uint32_t *)cmd_pkt + 2;
+		memset(clr_ptr, 0, REQUEST_ENTRY_SIZE - 8);
+		cmd_pkt->dseg_count = cpu_to_le16(tot_dsds);
+
+		/* Set NPORT-ID and LUN number*/
+		cmd_pkt->nport_handle = cpu_to_le16(sp->fcport->loop_id);
+		cmd_pkt->port_id[0] = sp->fcport->d_id.b.al_pa;
+		cmd_pkt->port_id[1] = sp->fcport->d_id.b.area;
+		cmd_pkt->port_id[2] = sp->fcport->d_id.b.domain;
+		cmd_pkt->vp_index = sp->fcport->vp_idx;
+
+		/* Build IOCB segments */
+		if (qla24xx_build_scsi_type_6_iocbs(sp, cmd_pkt, tot_dsds))
+			goto queuing_error_fcp_cmnd;
+
+		int_to_scsilun(sp->cmd->device->lun, &cmd_pkt->lun);
+		host_to_fcp_swap((uint8_t *)&cmd_pkt->lun, sizeof(cmd_pkt->lun));
+
+		/* build FCP_CMND IU */
+		memset(ctx->fcp_cmnd, 0, sizeof(struct fcp_cmnd));
+		int_to_scsilun(sp->cmd->device->lun, &ctx->fcp_cmnd->lun);
+		ctx->fcp_cmnd->additional_cdb_len = additional_cdb_len;
+
+		if (cmd->sc_data_direction == DMA_TO_DEVICE)
+			ctx->fcp_cmnd->additional_cdb_len |= 1;
+		else if (cmd->sc_data_direction == DMA_FROM_DEVICE)
+			ctx->fcp_cmnd->additional_cdb_len |= 2;
+
+		/*
+		 * Update tagged queuing modifier -- default is TSK_SIMPLE (0).
+		 */
+		if (scsi_populate_tag_msg(cmd, tag)) {
+			switch (tag[0]) {
+			case HEAD_OF_QUEUE_TAG:
+				ctx->fcp_cmnd->task_attribute =
+				    TSK_HEAD_OF_QUEUE;
+				break;
+			case ORDERED_QUEUE_TAG:
+				ctx->fcp_cmnd->task_attribute =
+				    TSK_ORDERED;
+				break;
+			}
+		}
+
+		memcpy(ctx->fcp_cmnd->cdb, cmd->cmnd, cmd->cmd_len);
+
+		fcp_dl = (uint32_t *)(ctx->fcp_cmnd->cdb + 16 +
+		    additional_cdb_len);
+		*fcp_dl = htonl((uint32_t)scsi_bufflen(cmd));
+
+		cmd_pkt->fcp_cmnd_dseg_len = cpu_to_le16(ctx->fcp_cmnd_len);
+		cmd_pkt->fcp_cmnd_dseg_address[0] =
+		    cpu_to_le32(LSD(ctx->fcp_cmnd_dma));
+		cmd_pkt->fcp_cmnd_dseg_address[1] =
+		    cpu_to_le32(MSD(ctx->fcp_cmnd_dma));
+
+		sp->flags |= SRB_FCP_CMND_DMA_VALID;
+		cmd_pkt->byte_count = cpu_to_le32((uint32_t)scsi_bufflen(cmd));
+		/* Set total data segment count. */
+		cmd_pkt->entry_count = (uint8_t)req_cnt;
+		/* Specify response queue number where
+		 * completion should happen
+		 */
+		cmd_pkt->entry_status = (uint8_t) rsp->id;
+	} else {
+		struct cmd_type_7 *cmd_pkt;
+		req_cnt = qla24xx_calc_iocbs(vha, tot_dsds);
+		if (req->cnt < (req_cnt + 2)) {
+			cnt = (uint16_t)RD_REG_DWORD_RELAXED(
+			    &reg->req_q_out[0]);
+			if (req->ring_index < cnt)
+				req->cnt = cnt - req->ring_index;
+			else
+				req->cnt = req->length -
+					(req->ring_index - cnt);
+		}
+		if (req->cnt < (req_cnt + 2))
+			goto queuing_error;
+
+		cmd_pkt = (struct cmd_type_7 *)req->ring_ptr;
+		cmd_pkt->handle = MAKE_HANDLE(req->id, handle);
+
+		/* Zero out remaining portion of packet. */
+		/* tagged queuing modifier -- default is TSK_SIMPLE (0).*/
+		clr_ptr = (uint32_t *)cmd_pkt + 2;
+		memset(clr_ptr, 0, REQUEST_ENTRY_SIZE - 8);
+		cmd_pkt->dseg_count = cpu_to_le16(tot_dsds);
+
+		/* Set NPORT-ID and LUN number*/
+		cmd_pkt->nport_handle = cpu_to_le16(sp->fcport->loop_id);
+		cmd_pkt->port_id[0] = sp->fcport->d_id.b.al_pa;
+		cmd_pkt->port_id[1] = sp->fcport->d_id.b.area;
+		cmd_pkt->port_id[2] = sp->fcport->d_id.b.domain;
+		cmd_pkt->vp_index = sp->fcport->vp_idx;
+
+		int_to_scsilun(sp->cmd->device->lun, &cmd_pkt->lun);
+		host_to_fcp_swap((uint8_t *)&cmd_pkt->lun,
+			sizeof(cmd_pkt->lun));
+
+		/*
+		 * Update tagged queuing modifier -- default is TSK_SIMPLE (0).
+		 */
+		if (scsi_populate_tag_msg(cmd, tag)) {
+			switch (tag[0]) {
+			case HEAD_OF_QUEUE_TAG:
+				cmd_pkt->task = TSK_HEAD_OF_QUEUE;
+				break;
+			case ORDERED_QUEUE_TAG:
+				cmd_pkt->task = TSK_ORDERED;
+				break;
+			}
+		}
+
+		/* Load SCSI command packet. */
+		memcpy(cmd_pkt->fcp_cdb, cmd->cmnd, cmd->cmd_len);
+		host_to_fcp_swap(cmd_pkt->fcp_cdb, sizeof(cmd_pkt->fcp_cdb));
+
+		cmd_pkt->byte_count = cpu_to_le32((uint32_t)scsi_bufflen(cmd));
+
+		/* Build IOCB segments */
+		qla24xx_build_scsi_iocbs(sp, cmd_pkt, tot_dsds);
+
+		/* Set total data segment count. */
+		cmd_pkt->entry_count = (uint8_t)req_cnt;
+		/* Specify response queue number where
+		 * completion should happen.
+		 */
+		cmd_pkt->entry_status = (uint8_t) rsp->id;
+
+	}
+	/* Build command packet. */
+	req->current_outstanding_cmd = handle;
+	req->outstanding_cmds[handle] = sp;
+	sp->handle = handle;
+	sp->cmd->host_scribble = (unsigned char *)(unsigned long)handle;
+	req->cnt -= req_cnt;
+	wmb();
+
+	/* Adjust ring index. */
+	req->ring_index++;
+	if (req->ring_index == req->length) {
+		req->ring_index = 0;
+		req->ring_ptr = req->ring;
+	} else
+		req->ring_ptr++;
+
+	sp->flags |= SRB_DMA_VALID;
+
+	/* Set chip new ring index. */
+	/* write, read and verify logic */
+	dbval = dbval | (req->id << 8) | (req->ring_index << 16);
+	if (ql2xdbwr)
+		qla82xx_wr_32(ha, ha->nxdb_wr_ptr, dbval);
+	else {
+		WRT_REG_DWORD(
+			(unsigned long __iomem *)ha->nxdb_wr_ptr,
+			dbval);
+		wmb();
+		while (RD_REG_DWORD(ha->nxdb_rd_ptr) != dbval) {
+			WRT_REG_DWORD(
+				(unsigned long __iomem *)ha->nxdb_wr_ptr,
+				dbval);
+			wmb();
+		}
+	}
+
+	/* Manage unprocessed RIO/ZIO commands in response queue. */
+	if (vha->flags.process_response_queue &&
+	    rsp->ring_ptr->signature != RESPONSE_PROCESSED)
+		qla24xx_process_response_queue(vha, rsp);
+
+	spin_unlock_irqrestore(&ha->hardware_lock, flags);
+	return QLA_SUCCESS;
+
+queuing_error_fcp_cmnd:
+	dma_pool_free(ha->fcp_cmnd_dma_pool, ctx->fcp_cmnd, ctx->fcp_cmnd_dma);
+queuing_error:
+	if (tot_dsds)
+		scsi_dma_unmap(cmd);
+
+	if (sp->ctx) {
+		mempool_free(sp->ctx, ha->ctx_mempool);
+		sp->ctx = NULL;
+	}
+	spin_unlock_irqrestore(&ha->hardware_lock, flags);
+
+	return QLA_FUNCTION_FAILED;
+}
+
 int
 qla2x00_start_sp(srb_t *sp)
 {
@@ -2213,7 +2629,7 @@
 	}
 
 	wmb();
-	qla2x00_start_iocbs(sp);
+	qla2x00_start_iocbs(sp->fcport->vha, ha->req_q_map[0]);
 done:
 	spin_unlock_irqrestore(&ha->hardware_lock, flags);
 	return rval;