Question on ata_gen_passthru_sense interpretation of fixed format sense buffer

[Akshat Jain <akshatzen@google.com>]

Hello Jens Tejun,
I have a question regarding the ata_gen_passthru_sense function (libata-scsi.c).

This function while generating fixed format sense blocks, puts the
INFORMATION field at offset 8 and COMMAND-SPECIFIC-INFORMATION at
offset 16.
While as per SCSI Primary commands - 4 specification, section 4.5.3
Fixed format sense data Table 53, the INFORMATION field is at offset 3
and COMMAND-SPECIFIC-INFORMATION is at offset 8.

Code snippet:

static void ata_gen_passthru_sense(struct ata_queued_cmd *qc)
{
struct scsi_cmnd *cmd = qc->scsicmd;
struct ata_taskfile *tf = &qc->result_tf;
unsigned char *sb = cmd->sense_buffer;
unsigned char *desc = sb + 8;
….
    if ((cmd->sense_buffer[0] & 0x7f) >= 0x72) {
    ....
    } else {
        /* Fixed sense format */
       desc[0] = tf->feature;
       desc[1] = tf->command; /* status */
       desc[2] = tf->device;
       desc[3] = tf->nsect;
       desc[7] = 0;
       if (tf->flags & ATA_TFLAG_LBA48)  {
           desc[8] |= 0x80;
           if (tf->hob_nsect)
               desc[8] |= 0x40;
           if (tf->hob_lbal || tf->hob_lbam || tf->hob_lbah)
               desc[8] |= 0x20;
       }
       desc[9] = tf->lbal;
       desc[10] = tf->lbam;
       desc[11] = tf->lbah;
    }
}

Link to spec: https://www.t10.org/cgi-bin/ac.pl?t=f&f=spc4r37a.pdf

My team has confirmed this observation. Do you believe our
interpretation of the specification is correct and if yes does this
need to be corrected?

Regards,
Akshat Jain