Subject: Re: disabling tagged queueing for SCSI1 devices
To: Greywolf <email@example.com>
From: Matthew Jacob <firstname.lastname@example.org>
Date: 02/10/2001 12:57:02
On Sat, 10 Feb 2001, Greywolf wrote:
> On Sat, 10 Feb 2001, Matthew Jacob wrote:
> # Sort of, because many CCS (SCSI-1, but not SCSI-2) drives support T'qing fine.
> Pardon my ignorance, but what's CCS?
"Common Command Set"
An extra agreement between vendors after the 1986 publication of the original
SCSI specification about what would be interesting ot actually support.
> # What issue are you trying to address here?
> I think the issue is that stuff that doesn't support tagged queueing
> tends to badly munge the system during startup.
> I have a better(?) idea: If the device does not report success during a
> tagged queueing check, disable it dynamically.
Actually, the preferred mechanism, which is sort of what we have now, is that
all HBAs should *not* have wide, sync, or tagged by default enabled.
When the inquiry data is scanned, it can report support for these options if
this is SCSI-2 or better.
There is enabling ioctls for the hbas for all of this.
The sd driver can enable tagged q'ing as it sees fit for SCSI-1 drives.
In other words, there shouldn't even be a need to 'disable' t-qing for SCSI-1
because it really shouldn't be enabled.
This is probably a thorpej-scsi branch issue.