Re: [PATCH v3 4/4] scsi: hisi_sas: Add support for DIF feature for v3 hw

[John Garry <john.garry@huawei.com>]

On 04/12/2018 04:12, Martin K. Petersen wrote:
>
> John,
>

Hi Martin,

Thanks for checking.

>> +static int hisi_sas_dif_dma_map(struct hisi_hba *hisi_hba,
>> +				int *n_elem_dif, struct sas_task *task)
>> +{
>> +	struct device *dev = hisi_hba->dev;
>> +	struct sas_ssp_task *ssp_task;
>> +	struct scsi_cmnd *scsi_cmnd;
>> +	int rc;
>> +
>> +	if (task->num_scatter) {
>> +		ssp_task = &task->ssp_task;
>> +		scsi_cmnd = ssp_task->cmd;
>> +
>> +		if (scsi_prot_sg_count(scsi_cmnd)) {
>> +			*n_elem_dif = dma_map_sg(dev,
>> +						 scsi_prot_sglist(scsi_cmnd),
>> +						 scsi_prot_sg_count(scsi_cmnd),
>> +						 task->data_dir);
>> +
>
> If you're only supporting DIF there is no DMA mapping or unmapping since
> the PI is generated by or verified by the HBA. No additional data is
> transferred to/from host memory. The protection scatterlist will be
> NULL.

So I was bit rushed in dropping DIX support...

>
>> +	switch (prot_op) {
>> +	case SCSI_PROT_READ_INSERT:
>> +		prot->dw0 |= T10_INSRT_EN_MSK;
>> +		prot->lbrtgv = lbrt_chk_val;
>> +		break;
>> +	case SCSI_PROT_READ_STRIP:
>> +		prot->dw0 |= (T10_RMV_EN_MSK | T10_CHK_EN_MSK);
>> +		prot->lbrtcv = lbrt_chk_val;
>> +		if (prot_type == SCSI_PROT_DIF_TYPE1)
>> +			prot->dw4 |= (0xc << 16);
>> +		else if (prot_type == SCSI_PROT_DIF_TYPE3)
>> +			prot->dw4 |= (0xfc << 16);
>> +		break;
>> +	case SCSI_PROT_READ_PASS:
>> +		prot->dw0 |= T10_CHK_EN_MSK;
>> +		prot->lbrtcv = lbrt_chk_val;
>> +		if (prot_type == SCSI_PROT_DIF_TYPE1)
>> +			prot->dw4 |= (0xc << 16);
>> +		else if (prot_type == SCSI_PROT_DIF_TYPE3)
>> +			prot->dw4 |= (0xfc << 16);
>> +		break;
>> +	case SCSI_PROT_WRITE_INSERT:
>> +		prot->dw0 |= T10_INSRT_EN_MSK;
>> +		prot->lbrtgv = lbrt_chk_val;
>> +		break;
>> +	case SCSI_PROT_WRITE_STRIP:
>> +		prot->dw0 |= (T10_RMV_EN_MSK | T10_CHK_EN_MSK);
>> +		prot->lbrtcv = lbrt_chk_val;
>> +		break;
>> +	case SCSI_PROT_WRITE_PASS:
>> +		prot->dw0 |= T10_CHK_EN_MSK;
>> +		prot->lbrtcv = lbrt_chk_val;
>> +		if (prot_type == SCSI_PROT_DIF_TYPE1)
>> +			prot->dw4 |= (0xc << 16);
>> +		else if (prot_type == SCSI_PROT_DIF_TYPE3)
>> +			prot->dw4 |= (0xfc << 16);
>> +		break;
>> +	default:
>> +		WARN(1, "prot_op(0x%x) is not valid\n", prot_op);
>> +		break;
>> +	}
>
> DIF is WRITE_INSERT/READ_STRIP operations only.
>

as above

>> +		if ((prot_op == SCSI_PROT_READ_INSERT) ||
>> +		    (prot_op == SCSI_PROT_WRITE_INSERT) ||
>> +		    (prot_op == SCSI_PROT_WRITE_PASS) ||
>> +		    (prot_op == SCSI_PROT_READ_PASS)) {
>> +			unsigned int interval = scsi_prot_interval(scsi_cmnd);
>> +			unsigned int ilog2_interval = ilog2(interval);
>> +
>> +			len = (task->total_xfer_len >> ilog2_interval) * 8;
>> +		}
>
> if (scmd->prot_flags & SCSI_PROT_TRANSFER_PI) {
>
>> +	.sg_prot_tablesize	= HISI_SAS_SGE_PAGE_CNT,
>
> Also not required if you're only doing DIF. There is no protection
> scatterlist.

and again.

>
> Anyway. Instead of throwing the DIX stuff away, let's figure out what
> the problem is.
>

So I just did not want to upstream support for a feature which seems to 
not be working.

Are you happy for us to make DIX support in this driver "hisi_sas DIX 
experimental" short term?

Regardless, we can discuss figuring out any SCSI MQ vs DIX issue in 
"DIF/DIX issue related to config CONFIG_SCSI_MQ_DEFAULT".

Cheers,
John