GOP at RCHVMP3.VNET.IBM.COM
Wed Mar 22 07:04:30 PST 1995
>I have a question about ACA. If a SCSI-3 target detects a parity
>error during the command phase and returns check condition status,
>how does the target know what ACA processing the initiator requested?
>The target cannot rely on the NACA bit in the control field if the parity
>error was on that byte or on the opcode byte.
>ACA only works if both initiator and target know that ACA is active.
If the target cannot determine the value of the NACA bit then it should
use the defalut value. In this case the default value is 0 which would
mean do SCSI-2 CA. This, of couse, only works if the initiator just
does an ACA tagged request sense with a clear ACA message. Anything more
complicated would cause problems because the Initiator would think the
ACA condition was still active when the Target would have already cleared
the CA condition.
I tend to agree with the group that suggests that if you are getting parity
errors you have big problems. And the issue of CA verses ACA is small in
Bye for now,
More information about the T10