qla2xxx: T10-Dif: add T10-PI support

Add support for T10-Dif for Target Mode to qla driver.
The driver will look for firmware attribute that support
this feature.  When the feature is present, the capabilities
will be report to TCM layer.

Add CTIO CRC2 iocb to build T10-Dif commands.
Add support routines to process good & error cases.

Signed-off-by: Quinn Tran <quinn.tran@qlogic.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: Saurav Kashyap <saurav.kashyap@qlogic.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

10 files changed, 851 insertions, 76 deletions

diff --git a/drivers/scsi/qla2xxx/qla_dbg.c b/drivers/scsi/qla2xxx/qla_dbg.c
index cdff3aa380cb..781ca5b09362 100644
--- a/drivers/scsi/qla2xxx/qla_dbg.c
+++ b/drivers/scsi/qla2xxx/qla_dbg.c
@@ -66,7 +66,7 @@
  * |                              |                    | 0xd030-0xd0ff	|
  * |                              |                    | 0xd101-0xd1fe	|
  * |                              |                    | 0xd213-0xd2fe	|
- * | Target Mode		  |	  0xe070       | 0xe021		|
+ * | Target Mode		  |	  0xe078       |		|
  * | Target Mode Management	  |	  0xf072       | 0xf002-0xf003	|
  * |                              |                    | 0xf046-0xf049  |
  * | Target Mode Task Management  |	  0x1000b      |		|
diff --git a/drivers/scsi/qla2xxx/qla_def.h b/drivers/scsi/qla2xxx/qla_def.h
index 5c590d40e676..1fa010448666 100644
--- a/drivers/scsi/qla2xxx/qla_def.h
+++ b/drivers/scsi/qla2xxx/qla_def.h
@@ -1629,10 +1629,20 @@ typedef struct {
 #define PO_MODE_DIF_PASS	2
 #define PO_MODE_DIF_REPLACE	3
 #define PO_MODE_DIF_TCP_CKSUM	6
-#define PO_ENABLE_DIF_BUNDLING	BIT_8
 #define PO_ENABLE_INCR_GUARD_SEED	BIT_3
-#define PO_DISABLE_INCR_REF_TAG	BIT_5
 #define PO_DISABLE_GUARD_CHECK	BIT_4
+#define PO_DISABLE_INCR_REF_TAG	BIT_5
+#define PO_DIS_HEADER_MODE	BIT_7
+#define PO_ENABLE_DIF_BUNDLING	BIT_8
+#define PO_DIS_FRAME_MODE	BIT_9
+#define PO_DIS_VALD_APP_ESC	BIT_10 /* Dis validation for escape tag/ffffh */
+#define PO_DIS_VALD_APP_REF_ESC BIT_11
+
+#define PO_DIS_APP_TAG_REPL	BIT_12 /* disable REG Tag replacement */
+#define PO_DIS_REF_TAG_REPL	BIT_13
+#define PO_DIS_APP_TAG_VALD	BIT_14 /* disable REF Tag validation */
+#define PO_DIS_REF_TAG_VALD	BIT_15
+
 /*
  * ISP queue - 64-Bit addressing, continuation crc entry structure definition.
  */
diff --git a/drivers/scsi/qla2xxx/qla_gbl.h b/drivers/scsi/qla2xxx/qla_gbl.h
index a96efff4146a..d48dea8fab1b 100644
--- a/drivers/scsi/qla2xxx/qla_gbl.h
+++ b/drivers/scsi/qla2xxx/qla_gbl.h
@@ -220,6 +220,13 @@ extern unsigned long qla2x00_get_async_timeout(struct scsi_qla_host *);
 
 extern void *qla2x00_alloc_iocbs(scsi_qla_host_t *, srb_t *);
 extern int qla2x00_issue_marker(scsi_qla_host_t *, int);
+extern int qla24xx_walk_and_build_sglist_no_difb(struct qla_hw_data *, srb_t *,
+	uint32_t *, uint16_t, struct qla_tgt_cmd *);
+extern int qla24xx_walk_and_build_sglist(struct qla_hw_data *, srb_t *,
+	uint32_t *, uint16_t, struct qla_tgt_cmd *);
+extern int qla24xx_walk_and_build_prot_sglist(struct qla_hw_data *, srb_t *,
+	uint32_t *, uint16_t, struct qla_tgt_cmd *);
+
 
 /*
  * Global Function Prototypes in qla_mbx.c source file.
diff --git a/drivers/scsi/qla2xxx/qla_inline.h b/drivers/scsi/qla2xxx/qla_inline.h
index be5b20446e5c..b3b1d6fc2d6c 100644
--- a/drivers/scsi/qla2xxx/qla_inline.h
+++ b/drivers/scsi/qla2xxx/qla_inline.h
@@ -5,6 +5,7 @@
  * See LICENSE.qla2xxx for copyright and licensing details.
  */
 
+#include "qla_target.h"
 /**
  * qla24xx_calc_iocbs() - Determine number of Command Type 3 and
  * Continuation Type 1 IOCBs to allocate.
@@ -128,12 +129,20 @@ qla2x00_clear_loop_id(fc_port_t *fcport) {
 }
 
 static inline void
-qla2x00_clean_dsd_pool(struct qla_hw_data *ha, srb_t *sp)
+qla2x00_clean_dsd_pool(struct qla_hw_data *ha, srb_t *sp,
+	struct qla_tgt_cmd *tc)
 {
 	struct dsd_dma *dsd_ptr, *tdsd_ptr;
 	struct crc_context *ctx;
 
-	ctx = (struct crc_context *)GET_CMD_CTX_SP(sp);
+	if (sp)
+		ctx = (struct crc_context *)GET_CMD_CTX_SP(sp);
+	else if (tc)
+		ctx = (struct crc_context *)tc->ctx;
+	else {
+		BUG();
+		return;
+	}
 
 	/* clean up allocated prev pool */
 	list_for_each_entry_safe(dsd_ptr, tdsd_ptr,
diff --git a/drivers/scsi/qla2xxx/qla_iocb.c b/drivers/scsi/qla2xxx/qla_iocb.c
index af83132141f7..760931529592 100644
--- a/drivers/scsi/qla2xxx/qla_iocb.c
+++ b/drivers/scsi/qla2xxx/qla_iocb.c
@@ -936,9 +936,9 @@ qla24xx_get_one_block_sg(uint32_t blk_sz, struct qla2_sgx *sgx,
 	return 1;
 }
 
-static int
+int
 qla24xx_walk_and_build_sglist_no_difb(struct qla_hw_data *ha, srb_t *sp,
-	uint32_t *dsd, uint16_t tot_dsds)
+	uint32_t *dsd, uint16_t tot_dsds, struct qla_tgt_cmd *tc)
 {
 	void *next_dsd;
 	uint8_t avail_dsds = 0;
@@ -948,21 +948,35 @@ qla24xx_walk_and_build_sglist_no_difb(struct qla_hw_data *ha, srb_t *sp,
 	uint32_t *cur_dsd = dsd;
 	uint16_t	used_dsds = tot_dsds;
 
-	uint32_t	prot_int;
+	uint32_t	prot_int; /* protection interval */
 	uint32_t	partial;
 	struct qla2_sgx sgx;
 	dma_addr_t	sle_dma;
 	uint32_t	sle_dma_len, tot_prot_dma_len = 0;
-	struct scsi_cmnd *cmd = GET_CMD_SP(sp);
-
-	prot_int = cmd->device->sector_size;
+	struct scsi_cmnd *cmd;
+	struct scsi_qla_host *vha;
 
 	memset(&sgx, 0, sizeof(struct qla2_sgx));
-	sgx.tot_bytes = scsi_bufflen(cmd);
-	sgx.cur_sg = scsi_sglist(cmd);
-	sgx.sp = sp;
-
-	sg_prot = scsi_prot_sglist(cmd);
+	if (sp) {
+		vha = sp->fcport->vha;
+		cmd = GET_CMD_SP(sp);
+		prot_int = cmd->device->sector_size;
+
+		sgx.tot_bytes = scsi_bufflen(cmd);
+		sgx.cur_sg = scsi_sglist(cmd);
+		sgx.sp = sp;
+
+		sg_prot = scsi_prot_sglist(cmd);
+	} else if (tc) {
+		vha = tc->vha;
+		prot_int      = tc->blk_sz;
+		sgx.tot_bytes = tc->bufflen;
+		sgx.cur_sg    = tc->sg;
+		sg_prot	      = tc->prot_sg;
+	} else {
+		BUG();
+		return 1;
+	}
 
 	while (qla24xx_get_one_block_sg(prot_int, &sgx, &partial)) {
 
@@ -995,10 +1009,18 @@ alloc_and_fill:
 				return 1;
 			}
 
-			list_add_tail(&dsd_ptr->list,
-			    &((struct crc_context *)sp->u.scmd.ctx)->dsd_list);
+			if (sp) {
+				list_add_tail(&dsd_ptr->list,
+				    &((struct crc_context *)
+					    sp->u.scmd.ctx)->dsd_list);
+
+				sp->flags |= SRB_CRC_CTX_DSD_VALID;
+			} else {
+				list_add_tail(&dsd_ptr->list,
+				    &(tc->ctx->dsd_list));
+				tc->ctx_dsd_alloced = 1;
+			}
 
-			sp->flags |= SRB_CRC_CTX_DSD_VALID;
 
 			/* add new list to cmd iocb or last list */
 			*cur_dsd++ = cpu_to_le32(LSD(dsd_ptr->dsd_list_dma));
@@ -1033,21 +1055,35 @@ alloc_and_fill:
 	return 0;
 }
 
-static int
+int
 qla24xx_walk_and_build_sglist(struct qla_hw_data *ha, srb_t *sp, uint32_t *dsd,
-	uint16_t tot_dsds)
+	uint16_t tot_dsds, struct qla_tgt_cmd *tc)
 {
 	void *next_dsd;
 	uint8_t avail_dsds = 0;
 	uint32_t dsd_list_len;
 	struct dsd_dma *dsd_ptr;
-	struct scatterlist *sg;
+	struct scatterlist *sg, *sgl;
 	uint32_t *cur_dsd = dsd;
 	int	i;
 	uint16_t	used_dsds = tot_dsds;
-	struct scsi_cmnd *cmd = GET_CMD_SP(sp);
+	struct scsi_cmnd *cmd;
+	struct scsi_qla_host *vha;
+
+	if (sp) {
+		cmd = GET_CMD_SP(sp);
+		sgl = scsi_sglist(cmd);
+		vha = sp->fcport->vha;
+	} else if (tc) {
+		sgl = tc->sg;
+		vha = tc->vha;
+	} else {
+		BUG();
+		return 1;
+	}
 
-	scsi_for_each_sg(cmd, sg, tot_dsds, i) {
+
+	for_each_sg(sgl, sg, tot_dsds, i) {
 		dma_addr_t	sle_dma;
 
 		/* Allocate additional continuation packets? */
@@ -1076,10 +1112,17 @@ qla24xx_walk_and_build_sglist(struct qla_hw_data *ha, srb_t *sp, uint32_t *dsd,
 				return 1;
 			}
 
-			list_add_tail(&dsd_ptr->list,
-			    &((struct crc_context *)sp->u.scmd.ctx)->dsd_list);
+			if (sp) {
+				list_add_tail(&dsd_ptr->list,
+				    &((struct crc_context *)
+					    sp->u.scmd.ctx)->dsd_list);
 
-			sp->flags |= SRB_CRC_CTX_DSD_VALID;
+				sp->flags |= SRB_CRC_CTX_DSD_VALID;
+			} else {
+				list_add_tail(&dsd_ptr->list,
+				    &(tc->ctx->dsd_list));
+				tc->ctx_dsd_alloced = 1;
+			}
 
 			/* add new list to cmd iocb or last list */
 			*cur_dsd++ = cpu_to_le32(LSD(dsd_ptr->dsd_list_dma));
@@ -1102,23 +1145,37 @@ qla24xx_walk_and_build_sglist(struct qla_hw_data *ha, srb_t *sp, uint32_t *dsd,
 	return 0;
 }
 
-static int
+int
 qla24xx_walk_and_build_prot_sglist(struct qla_hw_data *ha, srb_t *sp,
-							uint32_t *dsd,
-	uint16_t tot_dsds)
+	uint32_t *dsd, uint16_t tot_dsds, struct qla_tgt_cmd *tc)
 {
 	void *next_dsd;
 	uint8_t avail_dsds = 0;
 	uint32_t dsd_list_len;
 	struct dsd_dma *dsd_ptr;
-	struct scatterlist *sg;
+	struct scatterlist *sg, *sgl;
 	int	i;
 	struct scsi_cmnd *cmd;
 	uint32_t *cur_dsd = dsd;
-	uint16_t	used_dsds = tot_dsds;
+	uint16_t used_dsds = tot_dsds;
+	struct scsi_qla_host *vha;
+
+	if (sp) {
+		cmd = GET_CMD_SP(sp);
+		sgl = scsi_prot_sglist(cmd);
+		vha = sp->fcport->vha;
+	} else if (tc) {
+		vha = tc->vha;
+		sgl = tc->prot_sg;
+	} else {
+		BUG();
+		return 1;
+	}
 
-	cmd = GET_CMD_SP(sp);
-	scsi_for_each_prot_sg(cmd, sg, tot_dsds, i) {
+	ql_dbg(ql_dbg_tgt, vha, 0xe021,
+		"%s: enter\n", __func__);
+
+	for_each_sg(sgl, sg, tot_dsds, i) {
 		dma_addr_t	sle_dma;
 
 		/* Allocate additional continuation packets? */
@@ -1147,10 +1204,17 @@ qla24xx_walk_and_build_prot_sglist(struct qla_hw_data *ha, srb_t *sp,
 				return 1;
 			}
 
-			list_add_tail(&dsd_ptr->list,
-			    &((struct crc_context *)sp->u.scmd.ctx)->dsd_list);
+			if (sp) {
+				list_add_tail(&dsd_ptr->list,
+				    &((struct crc_context *)
+					    sp->u.scmd.ctx)->dsd_list);
 
-			sp->flags |= SRB_CRC_CTX_DSD_VALID;
+				sp->flags |= SRB_CRC_CTX_DSD_VALID;
+			} else {
+				list_add_tail(&dsd_ptr->list,
+				    &(tc->ctx->dsd_list));
+				tc->ctx_dsd_alloced = 1;
+			}
 
 			/* add new list to cmd iocb or last list */
 			*cur_dsd++ = cpu_to_le32(LSD(dsd_ptr->dsd_list_dma));
@@ -1386,10 +1450,10 @@ qla24xx_build_scsi_crc_2_iocbs(srb_t *sp, struct cmd_type_crc_2 *cmd_pkt,
 
 	if (!bundling && tot_prot_dsds) {
 		if (qla24xx_walk_and_build_sglist_no_difb(ha, sp,
-		    cur_dsd, tot_dsds))
+			cur_dsd, tot_dsds, NULL))
 			goto crc_queuing_error;
 	} else if (qla24xx_walk_and_build_sglist(ha, sp, cur_dsd,
-	    (tot_dsds - tot_prot_dsds)))
+			(tot_dsds - tot_prot_dsds), NULL))
 		goto crc_queuing_error;
 
 	if (bundling && tot_prot_dsds) {
@@ -1398,7 +1462,7 @@ qla24xx_build_scsi_crc_2_iocbs(srb_t *sp, struct cmd_type_crc_2 *cmd_pkt,
 			__constant_cpu_to_le16(CF_DIF_SEG_DESCR_ENABLE);
 		cur_dsd = (uint32_t *) &crc_ctx_pkt->u.bundling.dif_address;
 		if (qla24xx_walk_and_build_prot_sglist(ha, sp, cur_dsd,
-		    tot_prot_dsds))
+				tot_prot_dsds, NULL))
 			goto crc_queuing_error;
 	}
 	return QLA_SUCCESS;
diff --git a/drivers/scsi/qla2xxx/qla_isr.c b/drivers/scsi/qla2xxx/qla_isr.c
index 014f8c310b31..a56825c73c31 100644
--- a/drivers/scsi/qla2xxx/qla_isr.c
+++ b/drivers/scsi/qla2xxx/qla_isr.c
@@ -2474,12 +2474,14 @@ void qla24xx_process_response_queue(struct scsi_qla_host *vha,
 		if (pkt->entry_status != 0) {
 			qla2x00_error_entry(vha, rsp, (sts_entry_t *) pkt);
 
-			(void)qlt_24xx_process_response_error(vha, pkt);
+			if (qlt_24xx_process_response_error(vha, pkt))
+				goto process_err;
 
 			((response_t *)pkt)->signature = RESPONSE_PROCESSED;
 			wmb();
 			continue;
 		}
+process_err:
 
 		switch (pkt->entry_type) {
 		case STATUS_TYPE:
@@ -2496,10 +2498,10 @@ void qla24xx_process_response_queue(struct scsi_qla_host *vha,
 			qla24xx_logio_entry(vha, rsp->req,
 			    (struct logio_entry_24xx *)pkt);
 			break;
-                case CT_IOCB_TYPE:
+		case CT_IOCB_TYPE:
 			qla24xx_els_ct_entry(vha, rsp->req, pkt, CT_IOCB_TYPE);
 			break;
-                case ELS_IOCB_TYPE:
+		case ELS_IOCB_TYPE:
 			qla24xx_els_ct_entry(vha, rsp->req, pkt, ELS_IOCB_TYPE);
 			break;
 		case ABTS_RECV_24XX:
@@ -2508,6 +2510,7 @@ void qla24xx_process_response_queue(struct scsi_qla_host *vha,
 		case ABTS_RESP_24XX:
 		case CTIO_TYPE7:
 		case NOTIFY_ACK_TYPE:
+		case CTIO_CRC2:
 			qlt_response_pkt_all_vps(vha, (response_t *)pkt);
 			break;
 		case MARKER_TYPE:
diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c
index ef6332b61c18..5a70e24d9fe3 100644
--- a/drivers/scsi/qla2xxx/qla_os.c
+++ b/drivers/scsi/qla2xxx/qla_os.c
@@ -616,7 +616,7 @@ qla2x00_sp_free_dma(void *vha, void *ptr)
 
 	if (sp->flags & SRB_CRC_CTX_DSD_VALID) {
 		/* List assured to be having elements */
-		qla2x00_clean_dsd_pool(ha, sp);
+		qla2x00_clean_dsd_pool(ha, sp, NULL);
 		sp->flags &= ~SRB_CRC_CTX_DSD_VALID;
 	}
 
diff --git a/drivers/scsi/qla2xxx/qla_target.c b/drivers/scsi/qla2xxx/qla_target.c
index 0cb73074c199..f24d44cdecc4 100644
--- a/drivers/scsi/qla2xxx/qla_target.c
+++ b/drivers/scsi/qla2xxx/qla_target.c
@@ -182,6 +182,11 @@ struct scsi_qla_host *qlt_find_host_by_vp_idx(struct scsi_qla_host *vha,
 void qlt_24xx_atio_pkt_all_vps(struct scsi_qla_host *vha,
 	struct atio_from_isp *atio)
 {
+	ql_dbg(ql_dbg_tgt, vha, 0xe072,
+		"%s: qla_target(%d): type %x ox_id %04x\n",
+		__func__, vha->vp_idx, atio->u.raw.entry_type,
+		be16_to_cpu(atio->u.isp24.fcp_hdr.ox_id));
+
 	switch (atio->u.raw.entry_type) {
 	case ATIO_TYPE7:
 	{
@@ -236,6 +241,10 @@ void qlt_24xx_atio_pkt_all_vps(struct scsi_qla_host *vha,
 void qlt_response_pkt_all_vps(struct scsi_qla_host *vha, response_t *pkt)
 {
 	switch (pkt->entry_type) {
+	case CTIO_CRC2:
+		ql_dbg(ql_dbg_tgt, vha, 0xe073,
+			"qla_target(%d):%s: CRC2 Response pkt\n",
+			vha->vp_idx, __func__);
 	case CTIO_TYPE7:
 	{
 		struct ctio7_from_24xx *entry = (struct ctio7_from_24xx *)pkt;
@@ -1350,13 +1359,42 @@ static int qlt_pci_map_calc_cnt(struct qla_tgt_prm *prm)
 
 	prm->cmd->sg_mapped = 1;
 
-	/*
-	 * If greater than four sg entries then we need to allocate
-	 * the continuation entries
-	 */
-	if (prm->seg_cnt > prm->tgt->datasegs_per_cmd)
-		prm->req_cnt += DIV_ROUND_UP(prm->seg_cnt -
-		    prm->tgt->datasegs_per_cmd, prm->tgt->datasegs_per_cont);
+	if (cmd->se_cmd.prot_op == TARGET_PROT_NORMAL) {
+		/*
+		 * If greater than four sg entries then we need to allocate
+		 * the continuation entries
+		 */
+		if (prm->seg_cnt > prm->tgt->datasegs_per_cmd)
+			prm->req_cnt += DIV_ROUND_UP(prm->seg_cnt -
+			prm->tgt->datasegs_per_cmd,
+			prm->tgt->datasegs_per_cont);
+	} else {
+		/* DIF */
+		if ((cmd->se_cmd.prot_op == TARGET_PROT_DIN_INSERT) ||
+		    (cmd->se_cmd.prot_op == TARGET_PROT_DOUT_STRIP)) {
+			prm->seg_cnt = DIV_ROUND_UP(cmd->bufflen, cmd->blk_sz);
+			prm->tot_dsds = prm->seg_cnt;
+		} else
+			prm->tot_dsds = prm->seg_cnt;
+
+		if (cmd->prot_sg_cnt) {
+			prm->prot_sg      = cmd->prot_sg;
+			prm->prot_seg_cnt = pci_map_sg(prm->tgt->ha->pdev,
+				cmd->prot_sg, cmd->prot_sg_cnt,
+				cmd->dma_data_direction);
+			if (unlikely(prm->prot_seg_cnt == 0))
+				goto out_err;
+
+			if ((cmd->se_cmd.prot_op == TARGET_PROT_DIN_INSERT) ||
+			    (cmd->se_cmd.prot_op == TARGET_PROT_DOUT_STRIP)) {
+				/* Dif Bundling not support here */
+				prm->prot_seg_cnt = DIV_ROUND_UP(cmd->bufflen,
+								cmd->blk_sz);
+				prm->tot_dsds += prm->prot_seg_cnt;
+			} else
+				prm->tot_dsds += prm->prot_seg_cnt;
+		}
+	}
 
 	ql_dbg(ql_dbg_tgt, prm->cmd->vha, 0xe009, "seg_cnt=%d, req_cnt=%d\n",
 	    prm->seg_cnt, prm->req_cnt);
@@ -1377,6 +1415,16 @@ static inline void qlt_unmap_sg(struct scsi_qla_host *vha,
 	BUG_ON(!cmd->sg_mapped);
 	pci_unmap_sg(ha->pdev, cmd->sg, cmd->sg_cnt, cmd->dma_data_direction);
 	cmd->sg_mapped = 0;
+
+	if (cmd->prot_sg_cnt)
+		pci_unmap_sg(ha->pdev, cmd->prot_sg, cmd->prot_sg_cnt,
+			cmd->dma_data_direction);
+
+	if (cmd->ctx_dsd_alloced)
+		qla2x00_clean_dsd_pool(ha, NULL, cmd);
+
+	if (cmd->ctx)
+		dma_pool_free(ha->dl_dma_pool, cmd->ctx, cmd->ctx->crc_ctx_dma);
 }
 
 static int qlt_check_reserve_free_req(struct scsi_qla_host *vha,
@@ -1665,8 +1713,9 @@ static int qlt_pre_xmit_response(struct qla_tgt_cmd *cmd,
 		return QLA_TGT_PRE_XMIT_RESP_CMD_ABORTED;
 	}
 
-	ql_dbg(ql_dbg_tgt, vha, 0xe011, "qla_target(%d): tag=%u\n",
-	    vha->vp_idx, cmd->tag);
+	ql_dbg(ql_dbg_tgt, vha, 0xe011, "qla_target(%d): tag=%u ox_id %04x\n",
+		vha->vp_idx, cmd->tag,
+		be16_to_cpu(cmd->atio.u.isp24.fcp_hdr.ox_id));
 
 	prm->cmd = cmd;
 	prm->tgt = tgt;
@@ -1902,6 +1951,323 @@ skip_explict_conf:
 	/* Sense with len > 24, is it possible ??? */
 }
 
+
+
+/* diff  */
+static inline int
+qlt_hba_err_chk_enabled(struct se_cmd *se_cmd)
+{
+	/*
+	 * Uncomment when corresponding SCSI changes are done.
+	 *
+	 if (!sp->cmd->prot_chk)
+	 return 0;
+	 *
+	 */
+	switch (se_cmd->prot_op) {
+	case TARGET_PROT_DOUT_INSERT:
+	case TARGET_PROT_DIN_STRIP:
+		if (ql2xenablehba_err_chk >= 1)
+			return 1;
+		break;
+	case TARGET_PROT_DOUT_PASS:
+	case TARGET_PROT_DIN_PASS:
+		if (ql2xenablehba_err_chk >= 2)
+			return 1;
+		break;
+	case TARGET_PROT_DIN_INSERT:
+	case TARGET_PROT_DOUT_STRIP:
+		return 1;
+	default:
+		break;
+	}
+	return 0;
+}
+
+/*
+ * qla24xx_set_t10dif_tags_from_cmd - Extract Ref and App tags from SCSI command
+ *
+ */
+static inline void
+qlt_set_t10dif_tags(struct se_cmd *se_cmd, struct crc_context *ctx)
+{
+	uint32_t lba = 0xffffffff & se_cmd->t_task_lba;
+
+	/* wait til Mode Sense/Select cmd, modepage Ah, subpage 2
+	 * have been immplemented by TCM, before AppTag is avail.
+	 * Look for modesense_handlers[]
+	 */
+	ctx->app_tag = __constant_cpu_to_le16(0);
+	ctx->app_tag_mask[0] = 0x0;
+	ctx->app_tag_mask[1] = 0x0;
+
+	switch (se_cmd->prot_type) {
+	case TARGET_DIF_TYPE0_PROT:
+		/*
+		 * No check for ql2xenablehba_err_chk, as it would be an
+		 * I/O error if hba tag generation is not done.
+		 */
+		ctx->ref_tag = cpu_to_le32(lba);
+
+		if (!qlt_hba_err_chk_enabled(se_cmd))
+			break;
+
+		/* enable ALL bytes of the ref tag */
+		ctx->ref_tag_mask[0] = 0xff;
+		ctx->ref_tag_mask[1] = 0xff;
+		ctx->ref_tag_mask[2] = 0xff;
+		ctx->ref_tag_mask[3] = 0xff;
+		break;
+	/*
+	 * For TYpe 1 protection: 16 bit GUARD tag, 32 bit REF tag, and
+	 * 16 bit app tag.
+	 */
+	case TARGET_DIF_TYPE1_PROT:
+		ctx->ref_tag = cpu_to_le32(lba);
+
+		if (!qlt_hba_err_chk_enabled(se_cmd))
+			break;
+
+		/* enable ALL bytes of the ref tag */
+		ctx->ref_tag_mask[0] = 0xff;
+		ctx->ref_tag_mask[1] = 0xff;
+		ctx->ref_tag_mask[2] = 0xff;
+		ctx->ref_tag_mask[3] = 0xff;
+		break;
+	/*
+	 * For TYPE 2 protection: 16 bit GUARD + 32 bit REF tag has to
+	 * match LBA in CDB + N
+	 */
+	case TARGET_DIF_TYPE2_PROT:
+		ctx->ref_tag = cpu_to_le32(lba);
+
+		if (!qlt_hba_err_chk_enabled(se_cmd))
+			break;
+
+		/* enable ALL bytes of the ref tag */
+		ctx->ref_tag_mask[0] = 0xff;
+		ctx->ref_tag_mask[1] = 0xff;
+		ctx->ref_tag_mask[2] = 0xff;
+		ctx->ref_tag_mask[3] = 0xff;
+		break;
+
+	/* For Type 3 protection: 16 bit GUARD only */
+	case TARGET_DIF_TYPE3_PROT:
+		ctx->ref_tag_mask[0] = ctx->ref_tag_mask[1] =
+			ctx->ref_tag_mask[2] = ctx->ref_tag_mask[3] = 0x00;
+		break;
+	}
+}
+
+
+static inline int
+qlt_build_ctio_crc2_pkt(struct qla_tgt_prm *prm, scsi_qla_host_t *vha)
+{
+	uint32_t		*cur_dsd;
+	int			sgc;
+	uint32_t		transfer_length = 0;
+	uint32_t		data_bytes;
+	uint32_t		dif_bytes;
+	uint8_t			bundling = 1;
+	uint8_t			*clr_ptr;
+	struct crc_context	*crc_ctx_pkt = NULL;
+	struct qla_hw_data	*ha;
+	struct ctio_crc2_to_fw	*pkt;
+	dma_addr_t		crc_ctx_dma;
+	uint16_t		fw_prot_opts = 0;
+	struct qla_tgt_cmd	*cmd = prm->cmd;
+	struct se_cmd		*se_cmd = &cmd->se_cmd;
+	uint32_t h;
+	struct atio_from_isp *atio = &prm->cmd->atio;
+
+	sgc = 0;
+	ha = vha->hw;
+
+	pkt = (struct ctio_crc2_to_fw *)vha->req->ring_ptr;
+	prm->pkt = pkt;
+	memset(pkt, 0, sizeof(*pkt));
+
+	ql_dbg(ql_dbg_tgt, vha, 0xe071,
+		"qla_target(%d):%s: se_cmd[%p] CRC2 prot_op[0x%x] cmd prot sg:cnt[%p:%x] lba[%llu]\n",
+		vha->vp_idx, __func__, se_cmd, se_cmd->prot_op,
+		prm->prot_sg, prm->prot_seg_cnt, se_cmd->t_task_lba);
+
+	if ((se_cmd->prot_op == TARGET_PROT_DIN_INSERT) ||
+	    (se_cmd->prot_op == TARGET_PROT_DOUT_STRIP))
+		bundling = 0;
+
+	/* Compute dif len and adjust data len to incude protection */
+	data_bytes = cmd->bufflen;
+	dif_bytes  = (data_bytes / cmd->blk_sz) * 8;
+
+	switch (se_cmd->prot_op) {
+	case TARGET_PROT_DIN_INSERT:
+	case TARGET_PROT_DOUT_STRIP:
+		transfer_length = data_bytes;
+		data_bytes += dif_bytes;
+		break;
+
+	case TARGET_PROT_DIN_STRIP:
+	case TARGET_PROT_DOUT_INSERT:
+	case TARGET_PROT_DIN_PASS:
+	case TARGET_PROT_DOUT_PASS:
+		transfer_length = data_bytes + dif_bytes;
+		break;
+
+	default:
+		BUG();
+		break;
+	}
+
+	if (!qlt_hba_err_chk_enabled(se_cmd))
+		fw_prot_opts |= 0x10; /* Disable Guard tag checking */
+	/* HBA error checking enabled */
+	else if (IS_PI_UNINIT_CAPABLE(ha)) {
+		if ((se_cmd->prot_type == TARGET_DIF_TYPE1_PROT) ||
+		    (se_cmd->prot_type == TARGET_DIF_TYPE2_PROT))
+			fw_prot_opts |= PO_DIS_VALD_APP_ESC;
+		else if (se_cmd->prot_type == TARGET_DIF_TYPE3_PROT)
+			fw_prot_opts |= PO_DIS_VALD_APP_REF_ESC;
+	}
+
+	switch (se_cmd->prot_op) {
+	case TARGET_PROT_DIN_INSERT:
+	case TARGET_PROT_DOUT_INSERT:
+		fw_prot_opts |= PO_MODE_DIF_INSERT;
+		break;
+	case TARGET_PROT_DIN_STRIP:
+	case TARGET_PROT_DOUT_STRIP:
+		fw_prot_opts |= PO_MODE_DIF_REMOVE;
+		break;
+	case TARGET_PROT_DIN_PASS:
+	case TARGET_PROT_DOUT_PASS:
+		fw_prot_opts |= PO_MODE_DIF_PASS;
+		/* FUTURE: does tcm require T10CRC<->IPCKSUM conversion? */
+		break;
+	default:/* Normal Request */
+		fw_prot_opts |= PO_MODE_DIF_PASS;
+		break;
+	}
+
+
+	/* ---- PKT ---- */
+	/* Update entry type to indicate Command Type CRC_2 IOCB */
+	pkt->entry_type  = CTIO_CRC2;
+	pkt->entry_count = 1;
+	pkt->vp_index = vha->vp_idx;
+
+	h = qlt_make_handle(vha);
+	if (unlikely(h == QLA_TGT_NULL_HANDLE)) {
+		/*
+		 * CTIO type 7 from the firmware doesn't provide a way to
+		 * know the initiator's LOOP ID, hence we can't find
+		 * the session and, so, the command.
+		 */
+		return -EAGAIN;
+	} else
+		ha->tgt.cmds[h-1] = prm->cmd;
+
+
+	pkt->handle  = h | CTIO_COMPLETION_HANDLE_MARK;
+	pkt->nport_handle = prm->cmd->loop_id;
+	pkt->timeout = __constant_cpu_to_le16(QLA_TGT_TIMEOUT);
+	pkt->initiator_id[0] = atio->u.isp24.fcp_hdr.s_id[2];
+	pkt->initiator_id[1] = atio->u.isp24.fcp_hdr.s_id[1];
+	pkt->initiator_id[2] = atio->u.isp24.fcp_hdr.s_id[0];
+	pkt->exchange_addr   = atio->u.isp24.exchange_addr;
+	pkt->ox_id  = swab16(atio->u.isp24.fcp_hdr.ox_id);
+	pkt->flags |= (atio->u.isp24.attr << 9);
+	pkt->relative_offset = cpu_to_le32(prm->cmd->offset);
+
+	/* Set transfer direction */
+	if (cmd->dma_data_direction == DMA_TO_DEVICE)
+		pkt->flags = __constant_cpu_to_le16(CTIO7_FLAGS_DATA_IN);
+	else if (cmd->dma_data_direction == DMA_FROM_DEVICE)
+		pkt->flags = __constant_cpu_to_le16(CTIO7_FLAGS_DATA_OUT);
+
+
+	pkt->dseg_count = prm->tot_dsds;
+	/* Fibre channel byte count */
+	pkt->transfer_length = cpu_to_le32(transfer_length);
+
+
+	/* ----- CRC context -------- */
+
+	/* Allocate CRC context from global pool */
+	crc_ctx_pkt = cmd->ctx =
+	    dma_pool_alloc(ha->dl_dma_pool, GFP_ATOMIC, &crc_ctx_dma);
+
+	if (!crc_ctx_pkt)
+		goto crc_queuing_error;
+
+	/* Zero out CTX area. */
+	clr_ptr = (uint8_t *)crc_ctx_pkt;
+	memset(clr_ptr, 0, sizeof(*crc_ctx_pkt));
+
+	crc_ctx_pkt->crc_ctx_dma = crc_ctx_dma;
+	INIT_LIST_HEAD(&crc_ctx_pkt->dsd_list);
+
+	/* Set handle */
+	crc_ctx_pkt->handle = pkt->handle;
+
+	qlt_set_t10dif_tags(se_cmd, crc_ctx_pkt);
+
+	pkt->crc_context_address[0] = cpu_to_le32(LSD(crc_ctx_dma));
+	pkt->crc_context_address[1] = cpu_to_le32(MSD(crc_ctx_dma));
+	pkt->crc_context_len = CRC_CONTEXT_LEN_FW;
+
+
+	if (!bundling) {
+		cur_dsd = (uint32_t *) &crc_ctx_pkt->u.nobundling.data_address;
+	} else {
+		/*
+		 * Configure Bundling if we need to fetch interlaving
+		 * protection PCI accesses
+		 */
+		fw_prot_opts |= PO_ENABLE_DIF_BUNDLING;
+		crc_ctx_pkt->u.bundling.dif_byte_count = cpu_to_le32(dif_bytes);
+		crc_ctx_pkt->u.bundling.dseg_count =
+			cpu_to_le16(prm->tot_dsds - prm->prot_seg_cnt);
+		cur_dsd = (uint32_t *) &crc_ctx_pkt->u.bundling.data_address;
+	}
+
+	/* Finish the common fields of CRC pkt */
+	crc_ctx_pkt->blk_size   = cpu_to_le16(cmd->blk_sz);
+	crc_ctx_pkt->prot_opts  = cpu_to_le16(fw_prot_opts);
+	crc_ctx_pkt->byte_count = cpu_to_le32(data_bytes);
+	crc_ctx_pkt->guard_seed = __constant_cpu_to_le16(0);
+
+
+	/* Walks data segments */
+	pkt->flags |= __constant_cpu_to_le16(CTIO7_FLAGS_DSD_PTR);
+
+	if (!bundling && prm->prot_seg_cnt) {
+		if (qla24xx_walk_and_build_sglist_no_difb(ha, NULL, cur_dsd,
+			prm->tot_dsds, cmd))
+			goto crc_queuing_error;
+	} else if (qla24xx_walk_and_build_sglist(ha, NULL, cur_dsd,
+		(prm->tot_dsds - prm->prot_seg_cnt), cmd))
+		goto crc_queuing_error;
+
+	if (bundling && prm->prot_seg_cnt) {
+		/* Walks dif segments */
+		pkt->add_flags |=
+			__constant_cpu_to_le16(CTIO_CRC2_AF_DIF_DSD_ENA);
+
+		cur_dsd = (uint32_t *) &crc_ctx_pkt->u.bundling.dif_address;
+		if (qla24xx_walk_and_build_prot_sglist(ha, NULL, cur_dsd,
+			prm->prot_seg_cnt, cmd))
+			goto crc_queuing_error;
+	}
+	return QLA_SUCCESS;
+
+crc_queuing_error:
+	/* Cleanup will be performed by the caller */
+
+	return QLA_FUNCTION_FAILED;
+}
+
+
 /*
  * Callback to setup response of xmit_type of QLA_TGT_XMIT_DATA and *
  * QLA_TGT_XMIT_STATUS for >= 24xx silicon
@@ -1921,9 +2287,10 @@ int qlt_xmit_response(struct qla_tgt_cmd *cmd, int xmit_type,
 	qlt_check_srr_debug(cmd, &xmit_type);
 
 	ql_dbg(ql_dbg_tgt, cmd->vha, 0xe018,
-	    "is_send_status=%d, cmd->bufflen=%d, cmd->sg_cnt=%d, "
-	    "cmd->dma_data_direction=%d\n", (xmit_type & QLA_TGT_XMIT_STATUS) ?
-	    1 : 0, cmd->bufflen, cmd->sg_cnt, cmd->dma_data_direction);
+	    "is_send_status=%d, cmd->bufflen=%d, cmd->sg_cnt=%d, cmd->dma_data_direction=%d se_cmd[%p]\n",
+	    (xmit_type & QLA_TGT_XMIT_STATUS) ?
+	    1 : 0, cmd->bufflen, cmd->sg_cnt, cmd->dma_data_direction,
+	    &cmd->se_cmd);
 
 	res = qlt_pre_xmit_response(cmd, &prm, xmit_type, scsi_status,
 	    &full_req_cnt);
@@ -1941,7 +2308,10 @@ int qlt_xmit_response(struct qla_tgt_cmd *cmd, int xmit_type,
 	if (unlikely(res))
 		goto out_unmap_unlock;
 
-	res = qlt_24xx_build_ctio_pkt(&prm, vha);
+	if (cmd->se_cmd.prot_op && (xmit_type & QLA_TGT_XMIT_DATA))
+		res = qlt_build_ctio_crc2_pkt(&prm, vha);
+	else
+		res = qlt_24xx_build_ctio_pkt(&prm, vha);
 	if (unlikely(res != 0))
 		goto out_unmap_unlock;
 
@@ -1953,7 +2323,8 @@ int qlt_xmit_response(struct qla_tgt_cmd *cmd, int xmit_type,
 		    __constant_cpu_to_le16(CTIO7_FLAGS_DATA_IN |
 			CTIO7_FLAGS_STATUS_MODE_0);
 
-		qlt_load_data_segments(&prm, vha);
+		if (cmd->se_cmd.prot_op == TARGET_PROT_NORMAL)
+			qlt_load_data_segments(&prm, vha);
 
 		if (prm.add_status_pkt == 0) {
 			if (xmit_type & QLA_TGT_XMIT_STATUS) {
@@ -1983,8 +2354,14 @@ int qlt_xmit_response(struct qla_tgt_cmd *cmd, int xmit_type,
 			ql_dbg(ql_dbg_tgt, vha, 0xe019,
 			    "Building additional status packet\n");
 
+			/*
+			 * T10Dif: ctio_crc2_to_fw overlay ontop of
+			 * ctio7_to_24xx
+			 */
 			memcpy(ctio, pkt, sizeof(*ctio));
+			/* reset back to CTIO7 */
 			ctio->entry_count = 1;
+			ctio->entry_type = CTIO_TYPE7;
 			ctio->dseg_count = 0;
 			ctio->u.status1.flags &= ~__constant_cpu_to_le16(
 			    CTIO7_FLAGS_DATA_IN);
@@ -1993,6 +2370,11 @@ int qlt_xmit_response(struct qla_tgt_cmd *cmd, int xmit_type,
 			pkt->handle |= CTIO_INTERMEDIATE_HANDLE_MARK;
 			pkt->u.status0.flags |= __constant_cpu_to_le16(
 			    CTIO7_FLAGS_DONT_RET_CTIO);
+
+			/* qlt_24xx_init_ctio_to_isp will correct
+			 * all neccessary fields that's part of CTIO7.
+			 * There should be no residual of CTIO-CRC2 data.
+			 */
 			qlt_24xx_init_ctio_to_isp((struct ctio7_to_24xx *)ctio,
 			    &prm);
 			pr_debug("Status CTIO7: %p\n", ctio);
@@ -2041,8 +2423,10 @@ int qlt_rdy_to_xfer(struct qla_tgt_cmd *cmd)
 	if (qlt_issue_marker(vha, 0) != QLA_SUCCESS)
 		return -EIO;
 
-	ql_dbg(ql_dbg_tgt, vha, 0xe01b, "CTIO_start: vha(%d)",
-	    (int)vha->vp_idx);
+	ql_dbg(ql_dbg_tgt, vha, 0xe01b,
+		"%s: CTIO_start: vha(%d) se_cmd %p ox_id %04x\n",
+		__func__, (int)vha->vp_idx, &cmd->se_cmd,
+		be16_to_cpu(cmd->atio.u.isp24.fcp_hdr.ox_id));
 
 	/* Calculate number of entries and segments required */
 	if (qlt_pci_map_calc_cnt(&prm) != 0)
@@ -2054,14 +2438,19 @@ int qlt_rdy_to_xfer(struct qla_tgt_cmd *cmd)
 	res = qlt_check_reserve_free_req(vha, prm.req_cnt);
 	if (res != 0)
 		goto out_unlock_free_unmap;
+	if (cmd->se_cmd.prot_op)
+		res = qlt_build_ctio_crc2_pkt(&prm, vha);
+	else
+		res = qlt_24xx_build_ctio_pkt(&prm, vha);
 
-	res = qlt_24xx_build_ctio_pkt(&prm, vha);
 	if (unlikely(res != 0))
 		goto out_unlock_free_unmap;
 	pkt = (struct ctio7_to_24xx *)prm.pkt;
 	pkt->u.status0.flags |= __constant_cpu_to_le16(CTIO7_FLAGS_DATA_OUT |
 	    CTIO7_FLAGS_STATUS_MODE_0);
-	qlt_load_data_segments(&prm, vha);
+
+	if (cmd->se_cmd.prot_op == TARGET_PROT_NORMAL)
+		qlt_load_data_segments(&prm, vha);
 
 	cmd->state = QLA_TGT_STATE_NEED_DATA;
 
@@ -2079,6 +2468,143 @@ out_unlock_free_unmap:
 }
 EXPORT_SYMBOL(qlt_rdy_to_xfer);
 
+
+/*
+ * Checks the guard or meta-data for the type of error
+ * detected by the HBA.
+ */
+static inline int
+qlt_handle_dif_error(struct scsi_qla_host *vha, struct qla_tgt_cmd *cmd,
+		struct ctio_crc_from_fw *sts)
+{
+	uint8_t		*ap = &sts->actual_dif[0];
+	uint8_t		*ep = &sts->expected_dif[0];
+	uint32_t	e_ref_tag, a_ref_tag;
+	uint16_t	e_app_tag, a_app_tag;
+	uint16_t	e_guard, a_guard;
+	uint64_t	lba = cmd->se_cmd.t_task_lba;
+
+	a_guard   = be16_to_cpu(*(uint16_t *)(ap + 0));
+	a_app_tag = be16_to_cpu(*(uint16_t *)(ap + 2));
+	a_ref_tag = be32_to_cpu(*(uint32_t *)(ap + 4));
+
+	e_guard   = be16_to_cpu(*(uint16_t *)(ep + 0));
+	e_app_tag = be16_to_cpu(*(uint16_t *)(ep + 2));
+	e_ref_tag = be32_to_cpu(*(uint32_t *)(ep + 4));
+
+	ql_dbg(ql_dbg_tgt, vha, 0xe075,
+	    "iocb(s) %p Returned STATUS.\n", sts);
+
+	ql_dbg(ql_dbg_tgt, vha, 0xf075,
+	    "dif check TGT cdb 0x%x lba 0x%llu: [Actual|Expected] Ref Tag[0x%x|0x%x], App Tag [0x%x|0x%x], Guard [0x%x|0x%x]\n",
+	    cmd->atio.u.isp24.fcp_cmnd.cdb[0], lba,
+	    a_ref_tag, e_ref_tag, a_app_tag, e_app_tag, a_guard, e_guard);
+
+	/*
+	 * Ignore sector if:
+	 * For type     3: ref & app tag is all 'f's
+	 * For type 0,1,2: app tag is all 'f's
+	 */
+	if ((a_app_tag == 0xffff) &&
+	    ((cmd->se_cmd.prot_type != TARGET_DIF_TYPE3_PROT) ||
+	     (a_ref_tag == 0xffffffff))) {
+		uint32_t blocks_done;
+
+		/* 2TB boundary case covered automatically with this */
+		blocks_done = e_ref_tag - (uint32_t)lba + 1;
+		cmd->se_cmd.bad_sector = e_ref_tag;
+		cmd->se_cmd.pi_err = 0;
+		ql_dbg(ql_dbg_tgt, vha, 0xf074,
+			"need to return scsi good\n");
+
+		/* Update protection tag */
+		if (cmd->prot_sg_cnt) {
+			uint32_t i, j = 0, k = 0, num_ent;
+			struct scatterlist *sg, *sgl;
+
+
+			sgl = cmd->prot_sg;
+
+			/* Patch the corresponding protection tags */
+			for_each_sg(sgl, sg, cmd->prot_sg_cnt, i) {
+				num_ent = sg_dma_len(sg) / 8;
+				if (k + num_ent < blocks_done) {
+					k += num_ent;
+					continue;
+				}
+				j = blocks_done - k - 1;
+				k = blocks_done;
+				break;
+			}
+
+			if (k != blocks_done) {
+				ql_log(ql_log_warn, vha, 0xf076,
+				    "unexpected tag values tag:lba=%u:%llu)\n",
+				    e_ref_tag, (unsigned long long)lba);
+				goto out;
+			}
+
+#if 0
+			struct sd_dif_tuple *spt;
+			/* TODO:
+			 * This section came from initiator. Is it valid here?
+			 * should ulp be override with actual val???
+			 */
+			spt = page_address(sg_page(sg)) + sg->offset;
+			spt += j;
+
+			spt->app_tag = 0xffff;
+			if (cmd->se_cmd.prot_type == SCSI_PROT_DIF_TYPE3)
+				spt->ref_tag = 0xffffffff;
+#endif
+		}
+
+		return 0;
+	}
+
+	/* check guard */
+	if (e_guard != a_guard) {
+		cmd->se_cmd.pi_err = TCM_LOGICAL_BLOCK_GUARD_CHECK_FAILED;
+		cmd->se_cmd.bad_sector = cmd->se_cmd.t_task_lba;
+
+		ql_log(ql_log_warn, vha, 0xe076,
+		    "Guard ERR: cdb 0x%x lba 0x%llx: [Actual|Expected] Ref Tag[0x%x|0x%x], App Tag [0x%x|0x%x], Guard [0x%x|0x%x] cmd=%p\n",
+		    cmd->atio.u.isp24.fcp_cmnd.cdb[0], lba,
+		    a_ref_tag, e_ref_tag, a_app_tag, e_app_tag,
+		    a_guard, e_guard, cmd);
+		goto out;
+	}
+
+	/* check ref tag */
+	if (e_ref_tag != a_ref_tag) {
+		cmd->se_cmd.pi_err = TCM_LOGICAL_BLOCK_REF_TAG_CHECK_FAILED;
+		cmd->se_cmd.bad_sector = e_ref_tag;
+
+		ql_log(ql_log_warn, vha, 0xe077,
+			"Ref Tag ERR: cdb 0x%x lba 0x%llx: [Actual|Expected] Ref Tag[0x%x|0x%x], App Tag [0x%x|0x%x], Guard [0x%x|0x%x] cmd=%p\n",
+			cmd->atio.u.isp24.fcp_cmnd.cdb[0], lba,
+			a_ref_tag, e_ref_tag, a_app_tag, e_app_tag,
+			a_guard, e_guard, cmd);
+		goto out;
+	}
+
+	/* check appl tag */
+	if (e_app_tag != a_app_tag) {
+		cmd->se_cmd.pi_err = TCM_LOGICAL_BLOCK_APP_TAG_CHECK_FAILED;
+		cmd->se_cmd.bad_sector = cmd->se_cmd.t_task_lba;
+
+		ql_log(ql_log_warn, vha, 0xe078,
+			"App Tag ERR: cdb 0x%x lba 0x%llx: [Actual|Expected] Ref Tag[0x%x|0x%x], App Tag [0x%x|0x%x], Guard [0x%x|0x%x] cmd=%p\n",
+			cmd->atio.u.isp24.fcp_cmnd.cdb[0], lba,
+			a_ref_tag, e_ref_tag, a_app_tag, e_app_tag,
+			a_guard, e_guard, cmd);
+		goto out;
+	}
+out:
+	return 1;
+}
+
+
 /* If hardware_lock held on entry, might drop it, then reaquire */
 /* This function sends the appropriate CTIO to ISP 2xxx or 24xx */
 static int __qlt_send_term_exchange(struct scsi_qla_host *vha,
@@ -2159,14 +2685,20 @@ done:
 		if (!ha_locked && !in_interrupt())
 			msleep(250); /* just in case */
 
+		if (cmd->sg_mapped)
+			qlt_unmap_sg(vha, cmd);
 		vha->hw->tgt.tgt_ops->free_cmd(cmd);
 	}
 }
 
 void qlt_free_cmd(struct qla_tgt_cmd *cmd)
 {
-	BUG_ON(cmd->sg_mapped);
+	ql_dbg(ql_dbg_tgt, cmd->vha, 0xe074,
+	    "%s: se_cmd[%p] ox_id %04x\n",
+	    __func__, &cmd->se_cmd,
+	    be16_to_cpu(cmd->atio.u.isp24.fcp_hdr.ox_id));
 
+	BUG_ON(cmd->sg_mapped);
 	if (unlikely(cmd->free_sg))
 		kfree(cmd->sg);
 	kmem_cache_free(qla_tgt_cmd_cachep, cmd);
@@ -2404,10 +2936,40 @@ static void qlt_do_ctio_completion(struct scsi_qla_host *vha, uint32_t handle,
 			else
 				return;
 
+		case CTIO_DIF_ERROR: {
+			struct ctio_crc_from_fw *crc =
+				(struct ctio_crc_from_fw *)ctio;
+			ql_dbg(ql_dbg_tgt_mgt, vha, 0xf073,
+			    "qla_target(%d): CTIO with DIF_ERROR status %x received (state %x, se_cmd %p) actual_dif[0x%llx] expect_dif[0x%llx]\n",
+			    vha->vp_idx, status, cmd->state, se_cmd,
+			    *((u64 *)&crc->actual_dif[0]),
+			    *((u64 *)&crc->expected_dif[0]));
+
+			if (qlt_handle_dif_error(vha, cmd, ctio)) {
+				if (cmd->state == QLA_TGT_STATE_NEED_DATA) {
+					/* scsi Write/xfer rdy complete */
+					goto skip_term;
+				} else {
+					/* scsi read/xmit respond complete
+					 * call handle dif to send scsi status
+					 * rather than terminate exchange.
+					 */
+					cmd->state = QLA_TGT_STATE_PROCESSED;
+					ha->tgt.tgt_ops->handle_dif_err(cmd);
+					return;
+				}
+			} else {
+				/* Need to generate a SCSI good completion.
+				 * because FW did not send scsi status.
+				 */
+				status = 0;
+				goto skip_term;
+			}
+			break;
+		}
 		default:
 			ql_dbg(ql_dbg_tgt_mgt, vha, 0xf05b,
-			    "qla_target(%d): CTIO with error status "
-			    "0x%x received (state %x, se_cmd %p\n",
+			    "qla_target(%d): CTIO with error status 0x%x received (state %x, se_cmd %p\n",
 			    vha->vp_idx, status, cmd->state, se_cmd);
 			break;
 		}
@@ -2416,6 +2978,7 @@ static void qlt_do_ctio_completion(struct scsi_qla_host *vha, uint32_t handle,
 			if (qlt_term_ctio_exchange(vha, ctio, cmd, status))
 				return;
 	}
+skip_term:
 
 	if (cmd->state == QLA_TGT_STATE_PROCESSED) {
 		ql_dbg(ql_dbg_tgt, vha, 0xe01f, "Command %p finished\n", cmd);
@@ -2563,8 +3126,9 @@ static void qlt_do_work(struct work_struct *work)
 	    atio->u.isp24.fcp_cmnd.add_cdb_len]));
 
 	ql_dbg(ql_dbg_tgt, vha, 0xe022,
-	    "qla_target: START qla command: %p lun: 0x%04x (tag %d)\n",
-	    cmd, cmd->unpacked_lun, cmd->tag);
+		"qla_target: START qla cmd: %p se_cmd %p lun: 0x%04x (tag %d) len(%d) ox_id %x\n",
+		cmd, &cmd->se_cmd, cmd->unpacked_lun, cmd->tag, data_length,
+		cmd->atio.u.isp24.fcp_hdr.ox_id);
 
 	ret = vha->hw->tgt.tgt_ops->handle_cmd(vha, cmd, cdb, data_length,
 	    fcp_task_attr, data_dir, bidi);
@@ -3527,11 +4091,11 @@ static void qlt_24xx_atio_pkt(struct scsi_qla_host *vha,
 	switch (atio->u.raw.entry_type) {
 	case ATIO_TYPE7:
 		ql_dbg(ql_dbg_tgt, vha, 0xe02d,
-		    "ATIO_TYPE7 instance %d, lun %Lx, read/write %d/%d, "
-		    "add_cdb_len %d, data_length %04x, s_id %x:%x:%x\n",
+		    "ATIO_TYPE7 instance %d, lun %Lx, read/write %d/%d, cdb %x, add_cdb_len %x, data_length %04x, s_id %02x%02x%02x\n",
 		    vha->vp_idx, atio->u.isp24.fcp_cmnd.lun,
 		    atio->u.isp24.fcp_cmnd.rddata,
 		    atio->u.isp24.fcp_cmnd.wrdata,
+		    atio->u.isp24.fcp_cmnd.cdb[0],
 		    atio->u.isp24.fcp_cmnd.add_cdb_len,
 		    be32_to_cpu(get_unaligned((uint32_t *)
 			&atio->u.isp24.fcp_cmnd.add_cdb[
@@ -3629,11 +4193,13 @@ static void qlt_response_pkt(struct scsi_qla_host *vha, response_t *pkt)
 	tgt->irq_cmd_count++;
 
 	switch (pkt->entry_type) {
+	case CTIO_CRC2:
 	case CTIO_TYPE7:
 	{
 		struct ctio7_from_24xx *entry = (struct ctio7_from_24xx *)pkt;
-		ql_dbg(ql_dbg_tgt, vha, 0xe030, "CTIO_TYPE7: instance %d\n",
-		    vha->vp_idx);
+		ql_dbg(ql_dbg_tgt, vha, 0xe030,
+			"CTIO[0x%x] 12/CTIO7 7A/CRC2: instance %d\n",
+			entry->entry_type, vha->vp_idx);
 		qlt_do_ctio_completion(vha, entry->handle,
 		    le16_to_cpu(entry->status)|(pkt->entry_status << 16),
 		    entry);
@@ -4768,6 +5334,7 @@ qlt_24xx_process_response_error(struct scsi_qla_host *vha,
 	case ABTS_RESP_24XX:
 	case CTIO_TYPE7:
 	case NOTIFY_ACK_TYPE:
+	case CTIO_CRC2:
 		return 1;
 	default:
 		return 0;
diff --git a/drivers/scsi/qla2xxx/qla_target.h b/drivers/scsi/qla2xxx/qla_target.h
index ce33d8c26406..f873e10451d2 100644
--- a/drivers/scsi/qla2xxx/qla_target.h
+++ b/drivers/scsi/qla2xxx/qla_target.h
@@ -293,6 +293,7 @@ struct ctio_to_2xxx {
 #define CTIO_ABORTED			0x02
 #define CTIO_INVALID_RX_ID		0x08
 #define CTIO_TIMEOUT			0x0B
+#define CTIO_DIF_ERROR			0x0C     /* DIF error detected  */
 #define CTIO_LIP_RESET			0x0E
 #define CTIO_TARGET_RESET		0x17
 #define CTIO_PORT_UNAVAILABLE		0x28
@@ -498,11 +499,12 @@ struct ctio7_from_24xx {
 #define CTIO7_FLAGS_DONT_RET_CTIO	BIT_8
 #define CTIO7_FLAGS_STATUS_MODE_0	0
 #define CTIO7_FLAGS_STATUS_MODE_1	BIT_6
+#define CTIO7_FLAGS_STATUS_MODE_2	BIT_7
 #define CTIO7_FLAGS_EXPLICIT_CONFORM	BIT_5
 #define CTIO7_FLAGS_CONFIRM_SATISF	BIT_4
 #define CTIO7_FLAGS_DSD_PTR		BIT_2
-#define CTIO7_FLAGS_DATA_IN		BIT_1
-#define CTIO7_FLAGS_DATA_OUT		BIT_0
+#define CTIO7_FLAGS_DATA_IN		BIT_1 /* data to initiator */
+#define CTIO7_FLAGS_DATA_OUT		BIT_0 /* data from initiator */
 
 #define ELS_PLOGI			0x3
 #define ELS_FLOGI			0x4
@@ -514,6 +516,68 @@ struct ctio7_from_24xx {
 #define ELS_ADISC			0x52
 
 /*
+ *CTIO Type CRC_2 IOCB
+ */
+struct ctio_crc2_to_fw {
+	uint8_t entry_type;		/* Entry type. */
+#define CTIO_CRC2 0x7A
+	uint8_t entry_count;		/* Entry count. */
+	uint8_t sys_define;		/* System defined. */
+	uint8_t entry_status;		/* Entry Status. */
+
+	uint32_t handle;		/* System handle. */
+	uint16_t nport_handle;		/* N_PORT handle. */
+	uint16_t timeout;		/* Command timeout. */
+
+	uint16_t dseg_count;		/* Data segment count. */
+	uint8_t  vp_index;
+	uint8_t  add_flags;		/* additional flags */
+#define CTIO_CRC2_AF_DIF_DSD_ENA BIT_3
+
+	uint8_t  initiator_id[3];	/* initiator ID */
+	uint8_t  reserved1;
+	uint32_t exchange_addr;		/* rcv exchange address */
+	uint16_t reserved2;
+	uint16_t flags;			/* refer to CTIO7 flags values */
+	uint32_t residual;
+	uint16_t ox_id;
+	uint16_t scsi_status;
+	uint32_t relative_offset;
+	uint32_t reserved5;
+	uint32_t transfer_length;		/* total fc transfer length */
+	uint32_t reserved6;
+	uint32_t crc_context_address[2];/* Data segment address. */
+	uint16_t crc_context_len;	/* Data segment length. */
+	uint16_t reserved_1;		/* MUST be set to 0. */
+} __packed;
+
+/* CTIO Type CRC_x Status IOCB */
+struct ctio_crc_from_fw {
+	uint8_t entry_type;		/* Entry type. */
+	uint8_t entry_count;		/* Entry count. */
+	uint8_t sys_define;		/* System defined. */
+	uint8_t entry_status;		/* Entry Status. */
+
+	uint32_t handle;		/* System handle. */
+	uint16_t status;
+	uint16_t timeout;		/* Command timeout. */
+	uint16_t dseg_count;		/* Data segment count. */
+	uint32_t reserved1;
+	uint16_t state_flags;
+#define CTIO_CRC_SF_DIF_CHOPPED BIT_4
+
+	uint32_t exchange_address;	/* rcv exchange address */
+	uint16_t reserved2;
+	uint16_t flags;
+	uint32_t resid_xfer_length;
+	uint16_t ox_id;
+	uint8_t  reserved3[12];
+	uint16_t runt_guard;		/* reported runt blk guard */
+	uint8_t  actual_dif[8];
+	uint8_t  expected_dif[8];
+} __packed;
+
+/*
  * ISP queue - ABTS received/response entries structure definition for 24xx.
  */
 #define ABTS_RECV_24XX		0x54 /* ABTS received (for 24xx) */
@@ -641,6 +705,7 @@ struct qla_tgt_func_tmpl {
 	int (*handle_cmd)(struct scsi_qla_host *, struct qla_tgt_cmd *,
 			unsigned char *, uint32_t, int, int, int);
 	void (*handle_data)(struct qla_tgt_cmd *);
+	void (*handle_dif_err)(struct qla_tgt_cmd *);
 	int (*handle_tmr)(struct qla_tgt_mgmt_cmd *, uint32_t, uint8_t,
 			uint32_t);
 	void (*free_cmd)(struct qla_tgt_cmd *);
@@ -829,9 +894,9 @@ struct qla_tgt_sess {
 };
 
 struct qla_tgt_cmd {
+	struct se_cmd se_cmd;
 	struct qla_tgt_sess *sess;
 	int state;
-	struct se_cmd se_cmd;
 	struct work_struct free_work;
 	struct work_struct work;
 	/* Sense buffer that will be mapped into outgoing status */
@@ -843,6 +908,7 @@ struct qla_tgt_cmd {
 	unsigned int free_sg:1;
 	unsigned int aborted:1; /* Needed in case of SRR */
 	unsigned int write_data_transferred:1;
+	unsigned int ctx_dsd_alloced:1;
 
 	struct scatterlist *sg;	/* cmd data buffer SG vector */
 	int sg_cnt;		/* SG segments count */
@@ -857,6 +923,12 @@ struct qla_tgt_cmd {
 	struct scsi_qla_host *vha;
 
 	struct atio_from_isp atio;
+	/* t10dif */
+	struct scatterlist *prot_sg;
+	uint32_t prot_sg_cnt;
+	uint32_t blk_sz;
+	struct crc_context *ctx;
+
 };
 
 struct qla_tgt_sess_work_param {
@@ -901,6 +973,10 @@ struct qla_tgt_prm {
 	int sense_buffer_len;
 	int residual;
 	int add_status_pkt;
+	/* dif */
+	struct scatterlist *prot_sg;
+	uint16_t prot_seg_cnt;
+	uint16_t tot_dsds;
 };
 
 struct qla_tgt_srr_imm {
@@ -976,6 +1052,8 @@ extern void qlt_24xx_atio_pkt_all_vps(struct scsi_qla_host *,
 extern void qlt_response_pkt_all_vps(struct scsi_qla_host *, response_t *);
 extern int qlt_rdy_to_xfer(struct qla_tgt_cmd *);
 extern int qlt_xmit_response(struct qla_tgt_cmd *, int, uint8_t);
+extern int qlt_rdy_to_xfer_dif(struct qla_tgt_cmd *);
+extern int qlt_xmit_response_dif(struct qla_tgt_cmd *, int, uint8_t);
 extern void qlt_xmit_tm_rsp(struct qla_tgt_mgmt_cmd *);
 extern void qlt_free_mcmd(struct qla_tgt_mgmt_cmd *);
 extern void qlt_free_cmd(struct qla_tgt_cmd *cmd);
diff --git a/drivers/scsi/qla2xxx/tcm_qla2xxx.c b/drivers/scsi/qla2xxx/tcm_qla2xxx.c
index 68fb66fdb757..896cb23adb77 100644
--- a/drivers/scsi/qla2xxx/tcm_qla2xxx.c
+++ b/drivers/scsi/qla2xxx/tcm_qla2xxx.c
@@ -472,6 +472,11 @@ static int tcm_qla2xxx_write_pending(struct se_cmd *se_cmd)
 	cmd->sg_cnt = se_cmd->t_data_nents;
 	cmd->sg = se_cmd->t_data_sg;
 
+	cmd->prot_sg_cnt = se_cmd->t_prot_nents;
+	cmd->prot_sg = se_cmd->t_prot_sg;
+	cmd->blk_sz  = se_cmd->se_dev->dev_attrib.block_size;
+	se_cmd->pi_err = 0;
+
 	/*
 	 * qla_target.c:qlt_rdy_to_xfer() will call pci_map_sg() to setup
 	 * the SGL mappings into PCIe memory for incoming FCP WRITE data.
@@ -567,8 +572,13 @@ static void tcm_qla2xxx_handle_data_work(struct work_struct *work)
 			return;
 		}
 
-		transport_generic_request_failure(&cmd->se_cmd,
-						  TCM_CHECK_CONDITION_ABORT_CMD);
+		if (cmd->se_cmd.pi_err)
+			transport_generic_request_failure(&cmd->se_cmd,
+				cmd->se_cmd.pi_err);
+		else
+			transport_generic_request_failure(&cmd->se_cmd,
+				TCM_CHECK_CONDITION_ABORT_CMD);
+
 		return;
 	}
 
@@ -584,6 +594,27 @@ static void tcm_qla2xxx_handle_data(struct qla_tgt_cmd *cmd)
 	queue_work(tcm_qla2xxx_free_wq, &cmd->work);
 }
 
+static void tcm_qla2xxx_handle_dif_work(struct work_struct *work)
+{
+	struct qla_tgt_cmd *cmd = container_of(work, struct qla_tgt_cmd, work);
+
+	/* take an extra kref to prevent cmd free too early.
+	 * need to wait for SCSI status/check condition to
+	 * finish responding generate by transport_generic_request_failure.
+	 */
+	kref_get(&cmd->se_cmd.cmd_kref);
+	transport_generic_request_failure(&cmd->se_cmd, cmd->se_cmd.pi_err);
+}
+
+/*
+ * Called from qla_target.c:qlt_do_ctio_completion()
+ */
+static void tcm_qla2xxx_handle_dif_err(struct qla_tgt_cmd *cmd)
+{
+	INIT_WORK(&cmd->work, tcm_qla2xxx_handle_dif_work);
+	queue_work(tcm_qla2xxx_free_wq, &cmd->work);
+}
+
 /*
  * Called from qla_target.c:qlt_issue_task_mgmt()
  */
@@ -610,6 +641,11 @@ static int tcm_qla2xxx_queue_data_in(struct se_cmd *se_cmd)
 	cmd->sg = se_cmd->t_data_sg;
 	cmd->offset = 0;
 
+	cmd->prot_sg_cnt = se_cmd->t_prot_nents;
+	cmd->prot_sg = se_cmd->t_prot_sg;
+	cmd->blk_sz  = se_cmd->se_dev->dev_attrib.block_size;
+	se_cmd->pi_err = 0;
+
 	/*
 	 * Now queue completed DATA_IN the qla2xxx LLD and response ring
 	 */
@@ -1600,6 +1636,7 @@ static void tcm_qla2xxx_update_sess(struct qla_tgt_sess *sess, port_id_t s_id,
 static struct qla_tgt_func_tmpl tcm_qla2xxx_template = {
 	.handle_cmd		= tcm_qla2xxx_handle_cmd,
 	.handle_data		= tcm_qla2xxx_handle_data,
+	.handle_dif_err		= tcm_qla2xxx_handle_dif_err,
 	.handle_tmr		= tcm_qla2xxx_handle_tmr,
 	.free_cmd		= tcm_qla2xxx_free_cmd,
 	.free_mcmd		= tcm_qla2xxx_free_mcmd,