diff --git a/drivers/scsi/qla2xxx/qla_attr.c b/drivers/scsi/qla2xxx/qla_attr.c
index 7836eb0..8100670 100644
--- a/drivers/scsi/qla2xxx/qla_attr.c
+++ b/drivers/scsi/qla2xxx/qla_attr.c
@@ -1788,11 +1788,14 @@
 
 	if ((IS_QLA25XX(ha) || IS_QLA81XX(ha)) && ql2xenabledif) {
 		if (ha->fw_attributes & BIT_4) {
+			int prot = 0;
 			vha->flags.difdix_supported = 1;
 			ql_dbg(ql_dbg_user, vha, 0x7082,
 			    "Registered for DIF/DIX type 1 and 3 protection.\n");
+			if (ql2xenabledif == 1)
+				prot = SHOST_DIX_TYPE0_PROTECTION;
 			scsi_host_set_prot(vha->host,
-			    SHOST_DIF_TYPE1_PROTECTION
+			    prot | SHOST_DIF_TYPE1_PROTECTION
 			    | SHOST_DIF_TYPE2_PROTECTION
 			    | SHOST_DIF_TYPE3_PROTECTION
 			    | SHOST_DIX_TYPE1_PROTECTION
diff --git a/drivers/scsi/qla2xxx/qla_fw.h b/drivers/scsi/qla2xxx/qla_fw.h
index 691783a..aa69486 100644
--- a/drivers/scsi/qla2xxx/qla_fw.h
+++ b/drivers/scsi/qla2xxx/qla_fw.h
@@ -537,6 +537,11 @@
 	/*
 	 * If DIF Error is set in comp_status, these additional fields are
 	 * defined:
+	 *
+	 * !!! NOTE: Firmware sends expected/actual DIF data in big endian
+	 * format; but all of the "data" field gets swab32-d in the beginning
+	 * of qla2x00_status_entry().
+	 *
 	 * &data[10] : uint8_t report_runt_bg[2];	- computed guard
 	 * &data[12] : uint8_t actual_dif[8];		- DIF Data received
 	 * &data[20] : uint8_t expected_dif[8];		- DIF Data computed
diff --git a/drivers/scsi/qla2xxx/qla_inline.h b/drivers/scsi/qla2xxx/qla_inline.h
index d2e904b..c06e5f9 100644
--- a/drivers/scsi/qla2xxx/qla_inline.h
+++ b/drivers/scsi/qla2xxx/qla_inline.h
@@ -102,3 +102,24 @@
 		    fcport->d_id.b.al_pa);
 	}
 }
+
+static inline int
+qla2x00_hba_err_chk_enabled(unsigned char op)
+{
+	switch (op) {
+	case SCSI_PROT_READ_STRIP:
+	case SCSI_PROT_WRITE_INSERT:
+		if (ql2xenablehba_err_chk >= 1)
+			return 1;
+		break;
+	case SCSI_PROT_READ_PASS:
+	case SCSI_PROT_WRITE_PASS:
+		if (ql2xenablehba_err_chk >= 2)
+			return 1;
+		break;
+	case SCSI_PROT_READ_INSERT:
+	case SCSI_PROT_WRITE_STRIP:
+		return 1;
+	}
+	return 0;
+}
diff --git a/drivers/scsi/qla2xxx/qla_iocb.c b/drivers/scsi/qla2xxx/qla_iocb.c
index 49d6906..09ad3ce 100644
--- a/drivers/scsi/qla2xxx/qla_iocb.c
+++ b/drivers/scsi/qla2xxx/qla_iocb.c
@@ -717,12 +717,17 @@
 	unsigned char op = scsi_get_prot_op(cmd);
 
 	switch (scsi_get_prot_type(cmd)) {
-	/* For TYPE 0 protection: no checking */
 	case SCSI_PROT_DIF_TYPE0:
-		pkt->ref_tag_mask[0] = 0x00;
-		pkt->ref_tag_mask[1] = 0x00;
-		pkt->ref_tag_mask[2] = 0x00;
-		pkt->ref_tag_mask[3] = 0x00;
+		/*
+		 * No check for ql2xenablehba_err_chk, as it would be an
+		 * I/O error if hba tag generation is not done.
+		 */
+		pkt->ref_tag = cpu_to_le32((uint32_t)
+		    (0xffffffff & scsi_get_lba(cmd)));
+		pkt->ref_tag_mask[0] = 0xff;
+		pkt->ref_tag_mask[1] = 0xff;
+		pkt->ref_tag_mask[2] = 0xff;
+		pkt->ref_tag_mask[3] = 0xff;
 		break;
 
 	/*
@@ -730,7 +735,7 @@
 	 * match LBA in CDB + N
 	 */
 	case SCSI_PROT_DIF_TYPE2:
-		if (!ql2xenablehba_err_chk)
+		if (!qla2x00_hba_err_chk_enabled(op))
 			break;
 
 		if (scsi_prot_sg_count(cmd)) {
@@ -763,7 +768,7 @@
 	 * 16 bit app tag.
 	 */
 	case SCSI_PROT_DIF_TYPE1:
-		if (!ql2xenablehba_err_chk)
+		if (!qla2x00_hba_err_chk_enabled(op))
 			break;
 
 		if (protcnt && (op == SCSI_PROT_WRITE_STRIP ||
@@ -798,7 +803,161 @@
 	    scsi_get_prot_type(cmd), cmd);
 }
 
+struct qla2_sgx {
+	dma_addr_t		dma_addr;	/* OUT */
+	uint32_t		dma_len;	/* OUT */
 
+	uint32_t		tot_bytes;	/* IN */
+	struct scatterlist	*cur_sg;	/* IN */
+
+	/* for book keeping, bzero on initial invocation */
+	uint32_t		bytes_consumed;
+	uint32_t		num_bytes;
+	uint32_t		tot_partial;
+
+	/* for debugging */
+	uint32_t		num_sg;
+	srb_t			*sp;
+};
+
+static int
+qla24xx_get_one_block_sg(uint32_t blk_sz, struct qla2_sgx *sgx,
+	uint32_t *partial)
+{
+	struct scatterlist *sg;
+	uint32_t cumulative_partial, sg_len;
+	dma_addr_t sg_dma_addr;
+
+	if (sgx->num_bytes == sgx->tot_bytes)
+		return 0;
+
+	sg = sgx->cur_sg;
+	cumulative_partial = sgx->tot_partial;
+
+	sg_dma_addr = sg_dma_address(sg);
+	sg_len = sg_dma_len(sg);
+
+	sgx->dma_addr = sg_dma_addr + sgx->bytes_consumed;
+
+	if ((cumulative_partial + (sg_len - sgx->bytes_consumed)) >= blk_sz) {
+		sgx->dma_len = (blk_sz - cumulative_partial);
+		sgx->tot_partial = 0;
+		sgx->num_bytes += blk_sz;
+		*partial = 0;
+	} else {
+		sgx->dma_len = sg_len - sgx->bytes_consumed;
+		sgx->tot_partial += sgx->dma_len;
+		*partial = 1;
+	}
+
+	sgx->bytes_consumed += sgx->dma_len;
+
+	if (sg_len == sgx->bytes_consumed) {
+		sg = sg_next(sg);
+		sgx->num_sg++;
+		sgx->cur_sg = sg;
+		sgx->bytes_consumed = 0;
+	}
+
+	return 1;
+}
+
+static int
+qla24xx_walk_and_build_sglist_no_difb(struct qla_hw_data *ha, srb_t *sp,
+	uint32_t *dsd, uint16_t tot_dsds)
+{
+	void *next_dsd;
+	uint8_t avail_dsds = 0;
+	uint32_t dsd_list_len;
+	struct dsd_dma *dsd_ptr;
+	struct scatterlist *sg_prot;
+	uint32_t *cur_dsd = dsd;
+	uint16_t	used_dsds = tot_dsds;
+
+	uint32_t	prot_int;
+	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 = sp->cmd;
+
+	prot_int = cmd->device->sector_size;
+
+	memset(&sgx, 0, sizeof(struct qla2_sgx));
+	sgx.tot_bytes = scsi_bufflen(sp->cmd);
+	sgx.cur_sg = scsi_sglist(sp->cmd);
+	sgx.sp = sp;
+
+	sg_prot = scsi_prot_sglist(sp->cmd);
+
+	while (qla24xx_get_one_block_sg(prot_int, &sgx, &partial)) {
+
+		sle_dma = sgx.dma_addr;
+		sle_dma_len = sgx.dma_len;
+alloc_and_fill:
+		/* Allocate additional continuation packets? */
+		if (avail_dsds == 0) {
+			avail_dsds = (used_dsds > QLA_DSDS_PER_IOCB) ?
+					QLA_DSDS_PER_IOCB : used_dsds;
+			dsd_list_len = (avail_dsds + 1) * 12;
+			used_dsds -= avail_dsds;
+
+			/* allocate tracking DS */
+			dsd_ptr = kzalloc(sizeof(struct dsd_dma), GFP_ATOMIC);
+			if (!dsd_ptr)
+				return 1;
+
+			/* allocate new list */
+			dsd_ptr->dsd_addr = next_dsd =
+			    dma_pool_alloc(ha->dl_dma_pool, GFP_ATOMIC,
+				&dsd_ptr->dsd_list_dma);
+
+			if (!next_dsd) {
+				/*
+				 * Need to cleanup only this dsd_ptr, rest
+				 * will be done by sp_free_dma()
+				 */
+				kfree(dsd_ptr);
+				return 1;
+			}
+
+			list_add_tail(&dsd_ptr->list,
+			    &((struct crc_context *)sp->ctx)->dsd_list);
+
+			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));
+			*cur_dsd++ = cpu_to_le32(MSD(dsd_ptr->dsd_list_dma));
+			*cur_dsd++ = dsd_list_len;
+			cur_dsd = (uint32_t *)next_dsd;
+		}
+		*cur_dsd++ = cpu_to_le32(LSD(sle_dma));
+		*cur_dsd++ = cpu_to_le32(MSD(sle_dma));
+		*cur_dsd++ = cpu_to_le32(sle_dma_len);
+		avail_dsds--;
+
+		if (partial == 0) {
+			/* Got a full protection interval */
+			sle_dma = sg_dma_address(sg_prot) + tot_prot_dma_len;
+			sle_dma_len = 8;
+
+			tot_prot_dma_len += sle_dma_len;
+			if (tot_prot_dma_len == sg_dma_len(sg_prot)) {
+				tot_prot_dma_len = 0;
+				sg_prot = sg_next(sg_prot);
+			}
+
+			partial = 1; /* So as to not re-enter this block */
+			goto alloc_and_fill;
+		}
+	}
+	/* Null termination */
+	*cur_dsd++ = 0;
+	*cur_dsd++ = 0;
+	*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)
@@ -981,7 +1140,7 @@
 	struct scsi_cmnd	*cmd;
 	struct scatterlist	*cur_seg;
 	int			sgc;
-	uint32_t		total_bytes;
+	uint32_t		total_bytes = 0;
 	uint32_t		data_bytes;
 	uint32_t		dif_bytes;
 	uint8_t			bundling = 1;
@@ -1023,8 +1182,10 @@
 		    __constant_cpu_to_le16(CF_READ_DATA);
 	}
 
-	tot_prot_dsds = scsi_prot_sg_count(cmd);
-	if (!tot_prot_dsds)
+	if ((scsi_get_prot_op(sp->cmd) == SCSI_PROT_READ_INSERT) ||
+	    (scsi_get_prot_op(sp->cmd) == SCSI_PROT_WRITE_STRIP) ||
+	    (scsi_get_prot_op(sp->cmd) == SCSI_PROT_READ_STRIP) ||
+	    (scsi_get_prot_op(sp->cmd) == SCSI_PROT_WRITE_INSERT))
 		bundling = 0;
 
 	/* Allocate CRC context from global pool */
@@ -1107,15 +1268,28 @@
 	cmd_pkt->fcp_rsp_dseg_len = 0; /* Let response come in status iocb */
 
 	/* Compute dif len and adjust data len to incude protection */
-	total_bytes = data_bytes;
 	dif_bytes = 0;
 	blk_size = cmd->device->sector_size;
-	if (scsi_get_prot_op(cmd) != SCSI_PROT_NORMAL) {
-		dif_bytes = (data_bytes / blk_size) * 8;
-		total_bytes += dif_bytes;
+	dif_bytes = (data_bytes / blk_size) * 8;
+
+	switch (scsi_get_prot_op(sp->cmd)) {
+	case SCSI_PROT_READ_INSERT:
+	case SCSI_PROT_WRITE_STRIP:
+	    total_bytes = data_bytes;
+	    data_bytes += dif_bytes;
+	    break;
+
+	case SCSI_PROT_READ_STRIP:
+	case SCSI_PROT_WRITE_INSERT:
+	case SCSI_PROT_READ_PASS:
+	case SCSI_PROT_WRITE_PASS:
+	    total_bytes = data_bytes + dif_bytes;
+	    break;
+	default:
+	    BUG();
 	}
 
-	if (!ql2xenablehba_err_chk)
+	if (!qla2x00_hba_err_chk_enabled(scsi_get_prot_op(cmd)))
 		fw_prot_opts |= 0x10; /* Disable Guard tag checking */
 
 	if (!bundling) {
@@ -1151,7 +1325,12 @@
 
 	cmd_pkt->control_flags |=
 	    __constant_cpu_to_le16(CF_DATA_SEG_DESCR_ENABLE);
-	if (qla24xx_walk_and_build_sglist(ha, sp, cur_dsd,
+
+	if (!bundling && tot_prot_dsds) {
+		if (qla24xx_walk_and_build_sglist_no_difb(ha, sp,
+		    cur_dsd, tot_dsds))
+			goto crc_queuing_error;
+	} else if (qla24xx_walk_and_build_sglist(ha, sp, cur_dsd,
 	    (tot_dsds - tot_prot_dsds)))
 		goto crc_queuing_error;
 
@@ -1414,6 +1593,22 @@
 			goto queuing_error;
 		else
 			sp->flags |= SRB_DMA_VALID;
+
+		if ((scsi_get_prot_op(cmd) == SCSI_PROT_READ_INSERT) ||
+		    (scsi_get_prot_op(cmd) == SCSI_PROT_WRITE_STRIP)) {
+			struct qla2_sgx sgx;
+			uint32_t	partial;
+
+			memset(&sgx, 0, sizeof(struct qla2_sgx));
+			sgx.tot_bytes = scsi_bufflen(cmd);
+			sgx.cur_sg = scsi_sglist(cmd);
+			sgx.sp = sp;
+
+			nseg = 0;
+			while (qla24xx_get_one_block_sg(
+			    cmd->device->sector_size, &sgx, &partial))
+				nseg++;
+		}
 	} else
 		nseg = 0;
 
@@ -1428,6 +1623,11 @@
 			goto queuing_error;
 		else
 			sp->flags |= SRB_CRC_PROT_DMA_VALID;
+
+		if ((scsi_get_prot_op(cmd) == SCSI_PROT_READ_INSERT) ||
+		    (scsi_get_prot_op(cmd) == SCSI_PROT_WRITE_STRIP)) {
+			nseg = scsi_bufflen(cmd) / cmd->device->sector_size;
+		}
 	} else {
 		nseg = 0;
 	}
@@ -1454,6 +1654,7 @@
 	/* Build header part of command packet (excluding the OPCODE). */
 	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;
 
diff --git a/drivers/scsi/qla2xxx/qla_isr.c b/drivers/scsi/qla2xxx/qla_isr.c
index b16b772..53339f1 100644
--- a/drivers/scsi/qla2xxx/qla_isr.c
+++ b/drivers/scsi/qla2xxx/qla_isr.c
@@ -1435,25 +1435,27 @@
  * ASC/ASCQ fields in the sense buffer with ILLEGAL_REQUEST
  * to indicate to the kernel that the HBA detected error.
  */
-static inline void
+static inline int
 qla2x00_handle_dif_error(srb_t *sp, struct sts_entry_24xx *sts24)
 {
 	struct scsi_qla_host *vha = sp->fcport->vha;
 	struct scsi_cmnd *cmd = sp->cmd;
-	struct scsi_dif_tuple	*ep =
-			(struct scsi_dif_tuple *)&sts24->data[20];
-	struct scsi_dif_tuple	*ap =
-			(struct scsi_dif_tuple *)&sts24->data[12];
+	uint8_t		*ap = &sts24->data[12];
+	uint8_t		*ep = &sts24->data[20];
 	uint32_t	e_ref_tag, a_ref_tag;
 	uint16_t	e_app_tag, a_app_tag;
 	uint16_t	e_guard, a_guard;
 
-	e_ref_tag = be32_to_cpu(ep->ref_tag);
-	a_ref_tag = be32_to_cpu(ap->ref_tag);
-	e_app_tag = be16_to_cpu(ep->app_tag);
-	a_app_tag = be16_to_cpu(ap->app_tag);
-	e_guard = be16_to_cpu(ep->guard);
-	a_guard = be16_to_cpu(ap->guard);
+	/*
+	 * swab32 of the "data" field in the beginning of qla2x00_status_entry()
+	 * would make guard field appear at offset 2
+	 */
+	a_guard   = le16_to_cpu(*(uint16_t *)(ap + 2));
+	a_app_tag = le16_to_cpu(*(uint16_t *)(ap + 0));
+	a_ref_tag = le32_to_cpu(*(uint32_t *)(ap + 4));
+	e_guard   = le16_to_cpu(*(uint16_t *)(ep + 2));
+	e_app_tag = le16_to_cpu(*(uint16_t *)(ep + 0));
+	e_ref_tag = le32_to_cpu(*(uint32_t *)(ep + 4));
 
 	ql_dbg(ql_dbg_io, vha, 0x3023,
 	    "iocb(s) %p Returned STATUS.\n", sts24);
@@ -1465,6 +1467,63 @@
 	    cmd->cmnd[0], (u64)scsi_get_lba(cmd), 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) &&
+	    ((scsi_get_prot_type(cmd) != SCSI_PROT_DIF_TYPE3) ||
+	     (a_ref_tag == 0xffffffff))) {
+		uint32_t blocks_done, resid;
+		sector_t lba_s = scsi_get_lba(cmd);
+
+		/* 2TB boundary case covered automatically with this */
+		blocks_done = e_ref_tag - (uint32_t)lba_s + 1;
+
+		resid = scsi_bufflen(cmd) - (blocks_done *
+		    cmd->device->sector_size);
+
+		scsi_set_resid(cmd, resid);
+		cmd->result = DID_OK << 16;
+
+		/* Update protection tag */
+		if (scsi_prot_sg_count(cmd)) {
+			uint32_t i, j = 0, k = 0, num_ent;
+			struct scatterlist *sg;
+			struct sd_dif_tuple *spt;
+
+			/* Patch the corresponding protection tags */
+			scsi_for_each_prot_sg(cmd, sg,
+			    scsi_prot_sg_count(cmd), 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) {
+				qla_printk(KERN_WARNING, sp->fcport->vha->hw,
+				    "unexpected tag values tag:lba=%x:%lx)\n",
+				    e_ref_tag, lba_s);
+				return 1;
+			}
+
+			spt = page_address(sg_page(sg)) + sg->offset;
+			spt += j;
+
+			spt->app_tag = 0xffff;
+			if (scsi_get_prot_type(cmd) == SCSI_PROT_DIF_TYPE3)
+				spt->ref_tag = 0xffffffff;
+		}
+
+		return 0;
+	}
+
 	/* check guard */
 	if (e_guard != a_guard) {
 		scsi_build_sense_buffer(1, cmd->sense_buffer, ILLEGAL_REQUEST,
@@ -1472,7 +1531,7 @@
 		set_driver_byte(cmd, DRIVER_SENSE);
 		set_host_byte(cmd, DID_ABORT);
 		cmd->result |= SAM_STAT_CHECK_CONDITION << 1;
-		return;
+		return 1;
 	}
 
 	/* check appl tag */
@@ -1482,7 +1541,7 @@
 		set_driver_byte(cmd, DRIVER_SENSE);
 		set_host_byte(cmd, DID_ABORT);
 		cmd->result |= SAM_STAT_CHECK_CONDITION << 1;
-		return;
+		return 1;
 	}
 
 	/* check ref tag */
@@ -1492,8 +1551,9 @@
 		set_driver_byte(cmd, DRIVER_SENSE);
 		set_host_byte(cmd, DID_ABORT);
 		cmd->result |= SAM_STAT_CHECK_CONDITION << 1;
-		return;
+		return 1;
 	}
+	return 1;
 }
 
 /**
@@ -1767,7 +1827,7 @@
 		break;
 
 	case CS_DIF_ERROR:
-		qla2x00_handle_dif_error(sp, sts24);
+		logit = qla2x00_handle_dif_error(sp, sts24);
 		break;
 	default:
 		cp->result = DID_ERROR << 16;
diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c
index e02df27..d65a300 100644
--- a/drivers/scsi/qla2xxx/qla_os.c
+++ b/drivers/scsi/qla2xxx/qla_os.c
@@ -106,17 +106,21 @@
 		"Maximum queue depth to report for target devices.");
 
 /* Do not change the value of this after module load */
-int ql2xenabledif = 1;
+int ql2xenabledif = 0;
 module_param(ql2xenabledif, int, S_IRUGO|S_IWUSR);
 MODULE_PARM_DESC(ql2xenabledif,
 		" Enable T10-CRC-DIF "
-		" Default is 0 - No DIF Support. 1 - Enable it");
+		" Default is 0 - No DIF Support. 1 - Enable it"
+		", 2 - Enable DIF for all types, except Type 0.");
 
-int ql2xenablehba_err_chk;
+int ql2xenablehba_err_chk = 2;
 module_param(ql2xenablehba_err_chk, int, S_IRUGO|S_IWUSR);
 MODULE_PARM_DESC(ql2xenablehba_err_chk,
-		" Enable T10-CRC-DIF Error isolation by HBA"
-		" Default is 0 - Error isolation disabled, 1 - Enable it");
+		" Enable T10-CRC-DIF Error isolation by HBA:\n"
+		" Default is 1.\n"
+		"  0 -- Error isolation disabled\n"
+		"  1 -- Error isolation enabled only for DIX Type 0\n"
+		"  2 -- Error isolation enabled for all Types\n");
 
 int ql2xiidmaenable=1;
 module_param(ql2xiidmaenable, int, S_IRUGO);
@@ -2380,11 +2384,14 @@
 
 	if ((IS_QLA25XX(ha) || IS_QLA81XX(ha)) && ql2xenabledif) {
 		if (ha->fw_attributes & BIT_4) {
+			int prot = 0;
 			base_vha->flags.difdix_supported = 1;
 			ql_dbg(ql_dbg_init, base_vha, 0x00f1,
 			    "Registering for DIF/DIX type 1 and 3 protection.\n");
+			if (ql2xenabledif == 1)
+				prot = SHOST_DIX_TYPE0_PROTECTION;
 			scsi_host_set_prot(host,
-			    SHOST_DIF_TYPE1_PROTECTION
+			    prot | SHOST_DIF_TYPE1_PROTECTION
 			    | SHOST_DIF_TYPE2_PROTECTION
 			    | SHOST_DIF_TYPE3_PROTECTION
 			    | SHOST_DIX_TYPE1_PROTECTION