Logical unit groups and optional target reset
Elliott, Robert
Robert.Elliott at COMPAQ.com
Thu Dec 21 08:57:31 PST 2000
* From the T10 Reflector (t10 at t10.org), posted by:
* "Elliott, Robert" <Robert.Elliott at compaq.com>
*
Today all devices on all transports are required to support
TARGET RESET, so the target design cannot choose to ignore
it. We could change SAM-2 to let command sets dictate,
so new command sets like SBC-2 and SSC-2 could mark it as
optional. However, command set standards don't normally
deal with task management features.
A device can certainly implement a vendor-unique target reset,
even if the transport doesn't support it. This would prevent
standard software from using it (e.g. to clear reservations)
but leave it accessible for for lab debug.
---
PC: Robert.Elliott at compaq.com
UNIX: relliott at unixmail.compaq.com
> -----Original Message-----
> From: Tom Coughlan [mailto:coughlan at missioncriticallinux.com]
> Sent: Thursday, December 21, 2000 8:55 AM
> Subject: Re: Logical unit groups and optional target reset
>
> I agree with the goal of encouraging the implementation and use of
> Logical Unit Reset. But, I'm not sure that dropping support for Target
> Reset in future interconnects is a good idea. Target Reset is the
> medium weight hammer, between Logical Unit Reset, and resetting the
> whole environment by, say, resetting the host bus adapter.
>
> I accept that there may be some target designs where resetting all the
> logical units does all there is to do, and there is no point in
> implementing a Target Reset. But isn't it reasonable to
> expect that, in
> some target designs, a Target Reset will clear a hung
> something-or-other, that is not cleared by Logical Unit Resets?
>
> Shouldn't the decision to implement Target Reset be made based on the
> design of the target, rather than the interconnect?
> --
> Tom Coughlan
> Mission Critical Linux
> tel: 978-606-0262
> coughlan at missioncriticallinux.com
>
> "Elliott, Robert" wrote:
> > 01-015r0 Making Target Reset optional in SAM-2
> >
> > This would let new protocol standards like SRP and iSCSI
> drop support for
> > Target Reset. Target Reset is very disruptive in a SAN
> environment where
> > different logical units on one target device are being used
> by different
> > initiators. Logical Unit Reset is a better tool.
> >
> > SRP and iSCSI drivers can map legacy Target Reset requests
> into Logical Unit
> > Resets for every logical unit the initiator is using. This avoids
> > disrupting other initiators using other logical units. Its
> absence from
> > future standards may help wean software from issuing Target Resets.
*
* 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
mailing list