Bidirectional transfers in SPI-4

Mark.Heath at Mark.Heath at
Wed Oct 18 07:01:36 PDT 2000

* From the T10 Reflector (t10 at, posted by:
* Mark.Heath at
I noticed in the September ENDL letter that there is some talk of allowing
bidirectional transfers in SPI-4 only in packetized mode.  I know that
somebody mentioned that this would be a problem for application clients
because some commands would only work in bidirectional mode and the
application client may not know (and in my opinion *should not* know) what
transfer mode has been negotiated.

However, I think the problem is even worse than that.  One cannot assume
that transfer mode negotiations only occur when no commands are
outstanding.  For example, the standard permits the initiator or target to
negotiate a new transfer mode *in the middle of* a command.  If certain
commands work only in a bidirectional-capable transfer mode, what is a
target to do if the initiator negotiates out of a bidirectional-capable
transfer mode while the target is operating on such a command?

Wouldn't everyone be better off if we avoid this can of worms by defining
bidirectional transfers such that they operate in all transfer modes?


Mark Heath
Seagate Technology

* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at

More information about the T10 mailing list