rsnively at Brocade.COM
Mon Mar 4 07:29:10 PST 2002
* From the T10 Reflector (t10 at t10.org), posted by:
* Robert Snively <rsnively at brocade.com>
> Subject: Re: FCP_RESID_OVER question
> I understand your reply in case of a Read Cmd, but my question was
> in fact relative to commands like Inquiry, Receive Diagnostics, Mode
> Sense, .. . In these cases, could you clarify ?
In this case, the length may or may not be known to the initiator.
The initiator will typically choose a data length and an allocation
length in the CDB that are the same. The target will send the
correct number of bytes, and provide good status.
If the allocation length is too long or too short, and assuming my
memory and brief review of the material are correct, this is not an
error for most SPC-2 commands. Data in excess of the allocation
length will not be transmitted. Data allowed by the allocation
length but not available will also not be transmitted. No check
condition will occur, so no incorrect length indication will be
provided. Of course, there are some commands, especially tape
commands, for which an incorrect length indication will be
indicated in the normal manner with a check condition.
In all cases, residuals will be provided in the normal manner
with respect to the FCP specified data lengths.
Hope that is some help.
> Claude Pattein
> Robert Snively wrote:
> > > I wish to get some clarification about RESID_OVER.
> > >
> > > In FCP-2 spec, draft rev.7, I found, in 9.3 FCP_DATA IU, the
> > > sentence :
> > > "If the command requested that data beyond FCP_DL be transferred,
> > > the FCP_RSP_IU shall contain the FCP_RESID_OVER bit set to one"
> > >
> > > In the case of a SCSI CDB containing an insuffisent
> > > Allocation length, but
> > > with FCP_DL equal to this Allocation length, do I get a
> > > FCP_RESID_OVER
> > > Status (due to the troncation of returned data) or do I get a
> > > zero Status
> > > (due to FCP_DL = Allocation length) ?
> > >
> > If I interpret your question correctly, you will not
> > get a residual bit set, but you will have CHECK CONDITION in the
> > FCP, with sense data indicating an incorrect length.
> > > In other words, if FCP_DL always equals Allocation length (of
> > > SCSI CDB),
> > > is it possible to get a FCP_RESID_OVER status ?
> > The allocation length is actually a calculated value based
> > on number of blocks and block size (for disks). If the
> > driver is set up incorrectly, the predicted allocation
> > length (calculated by the driver using the wrong block length) will
> > match the FCP_DL, but will not match the actual
> > allocation length (which is calculated by the disk drive
> > depending on its known properties). In devices where the
> > actual allocation length is exposed, I believe you are
> > correct.
> > Basically FCP_RESID is designed to detect cases where the
> > software that prepares the CDB and the software/firmware that
> > prepares the DMA space are not working from the same assumptions.
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org
More information about the T10