CA_ACA Proposal 97-225r0.DOC
coughlan at star.zko.dec.com
Fri Aug 29 14:44:32 PDT 1997
* From the T10 (formerly SCSI) Reflector (t10 at symbios.com), posted by:
* coughlan at star.zko.dec.com (Tom)
The reflector seems to have failed to forward the second half of my
message, possibly because it was separated from the first half by a
(poor man's) vertical ellipsis, that is, dot<cr>dot<cr>dot<cr>.
Anyway, here is the second half of my comments on CA_ACA Proposal
>7.20.3 Deferred errors
>If it is possible to associate a
>deferred error with a causing initiator and with a particular function or a
>particular subset of data, and the error is either unrecovered or
>be reported by the mode parameters, a deferred error indication shall be
>returned to an application client on the causing initiator. If an
>client on an initiator other than the causing initiator attempts access to
>particular function or subset of data associated with the deferred error,
>command attempting the access shall be responded to according to the
>requirements for a contingent allegiance condition.
This is one area where I think the CA_ACA proposal goes too far.
In this case, the device server has realized that the _data_ that is being
referenced by one initiator is the subject of a deferred error that
is to be returned to another initiator. This should result in a check
condition, because there is (potentially) something wrong with the data.
This is not a case of unwanted initator-to-initiator command
"interference", it is a case of protecting data integrity.
>Changes to SAM-2:
>Notes: 1. The SCSI-2 contingent allegiance condition and extended auto
>contingent allegiance condition have been augmented in SCSI-3 by
>auto contingent allegiance in conjunction with the NACA bit.
First, the way this is worded implies that extended auto contingent
allegiance still exists in SCSI-3. It needs to be made clear that the
SCSI-2 extended auto contingent allegiance is obsolete (illegal?) in
SAM currently only defines ACA, and it says that one of ACA's modes is
SCSI-2 CA. I agree with the proposal to separate CA from ACA, and to
explicitly define both CA and ACA in SCSI-3. This seems to match
people's intuition better, and to reflect the way products have evolved.
I think it is important, however, for SCSI-3 to explicit define the
behavior of CA, in SCSI-3 terms, rather than the current tenuous
references to the "SCSI-2 rules for CA".
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at symbios.com
More information about the T10