Re: [PATCH 30/37] scsi: ibmvscsi_tgt: use 16-bits defined sense codes

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [Critical] Missing `SAM_STAT_CHECK_CONDITION` assignment causes sense data to be ignored on DMA failure, resulting in silent data corruption.
--

commit 9055c36d0afa69c58f35f6c4cc3a76397be71645
Author: Damien Le Moal <dlemoal@kernel.org>

scsi: ibmvscsi_tgt: use 16-bits defined sense codes

This commit refactors the IBM virtual SCSI target driver to use
scsi_set_sense_buffer() and defined sense code enums instead of
hard-coded values. This improves code clarity and self-documentation
without intending any functional changes.

> diff --git a/drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c b/drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c
> index 61f6828007654..e2af90b6da39b 100644
> --- a/drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c
> +++ b/drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c
> @@ -3757,9 +3757,8 @@ static int ibmvscsis_queue_data_in(struct se_cmd *se_cmd)
>  		dev_err(&vscsi->dev, "srp_transfer_data failed: %d\n", rc);
>  		se_cmd->scsi_sense_length = 18;
>  		memset(se_cmd->sense_buffer, 0, se_cmd->scsi_sense_length);
> -		/* Logical Unit Communication Time-out asc/ascq = 0x0801 */
> -		scsi_build_sense_buffer(0, se_cmd->sense_buffer, MEDIUM_ERROR,
> -					0x08, 0x01);
> +		scsi_set_sense_buffer(0, se_cmd->sense_buffer, MEDIUM_ERROR,
> +				      LU_COMMUNICATION_TIMEOUT);
>  	}
>  
>  	srp_build_response(vscsi, cmd, &len);

[Severity: Critical]
This is a pre-existing issue, but does this error path properly notify the
initiator of the failure? 

When srp_transfer_data() fails in ibmvscsis_queue_data_in(), the sense
buffer is built, but it appears se_cmd->scsi_status is never set to
SAM_STAT_CHECK_CONDITION.

Because the status remains 0 (SAM_STAT_GOOD), when srp_build_response()
is subsequently called, the check for rsp->status evaluates to false:

srp_build_response()
    ...
	rsp->status = ibmvscsis_fast_fail(vscsi, cmd);
	if (rsp->status) {
		...
		if (se_cmd->scsi_sense_length && se_cmd->sense_buffer) {
			rsp->sense_data_len = cpu_to_be32(se_cmd->scsi_sense_length);
			...

This causes the built sense data to be skipped, returning a success
response to the initiator despite the DMA transfer failing. Can this lead
to silent data corruption where the initiator incorrectly assumes the read
was successful?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260831020451.585944-1-dlemoal@kernel.org?part=30