Re: [PATCH] scsi/libata: Support variable-length cdb of ata pass-thru(32).

[Bart Van Assche <Bart.VanAssche@wdc.com>]

On Sat, 2017-06-24 at 01:50 +0900, Minwoo Im wrote:
> On Thu, Jun 22, 2017 at 3:52 AM, Bart Van Assche <Bart.VanAssche@wdc.com> wrote:
> > > @@ -4385,7 +4463,12 @@ int ata_scsi_add_hosts(struct ata_host *host, struct scsi_host_template *sht)
> > >               shost->max_id = 16;
> > >               shost->max_lun = 1;
> > >               shost->max_channel = 1;
> > > -             shost->max_cmd_len = 16;
> > > +             /*
> > > +              * SPC-3, SPC-4: Definition of CDB
> > > +              * A CDB may have a fixed length of up to 16 bytes or
> > > +              * variable length of between 12 and 260 bytes.
> > > +              */
> > > +             shost->max_cmd_len = 260;
> > 
> > Does ATA pass-through really support 260-byte CDBs or is the maximum CDB length
> > that is supported by ATA_32 perhaps 32 bytes?
> 
> Here's my opinion about your question.
> In perspective of SCSI host, I guess the max cmd len should be 260 bytes.
> Because SPC says that, in case of variable-length command, it could be
> from 12 to 260 bytes.
> That's why I have applied 260 value to max_cmd_len of scsi host.
> Please feel free to give any opinions about it.

Hello Minwoo,

Table 172 in document sat4r06.pdf shows that the CDB of the ATA PASS-THROUGH (32)
command is exactly 32 bytes long. My conclusion from analyzing __ata_scsi_queuecmd()
is that the current version of that function does not support CDBs of more than
16 bytes. Since your patch adds support for a single command with 32-byte CDB I
am convinced that max_cmd_len should be set to 32 in ata_scsi_add_hosts().

Bart.