This message is from the T13 list server.


There is no limit on how long it may take to transfer data, and so no upper
limit on pausing once you have entered that state  (there may be command
timeouts used at the higher levels of the protocol, but nothing at this
level).  I think this is fine - SCSI works the same way - and I have not
seen any practical problems (indeed, introducing a timeout would introduce
more problems, like what action is taken when you hit the timeout?)

However, the sender has to take action based on the not ready signal from
the receiver to pause the transfer within a defined period of time (which is
one element that leads to the receiver having to make provision to receive
up to 3 more words after it initially asserts the not ready signal).

Technically the PAUSE action is always undertaken by the sender.  In one
case the sender initiates the pause.  In another case, the receiver
indicates that it is not ready by asserting the not ready signal, the sender
sees this, and then pauses.

And I agree that the UDMA section could stand some cleanup.  The wording is
largely intact from the original proposals, but since then the document
itself has changed quite a bit.  So rewording some sections might give the
overall document a better flow.  As you know, care must be taken in any such
effort to make sure that substantive changes do not creep in.

Jim


-----Original Message-----
From: Hale Landis [mailto:[EMAIL PROTECTED]]
Sent: Tuesday, April 09, 2002 10:31 PM
To: T13 List Server
Subject: RE: RE: [t13] UDMA Bursts - Pause versus Termination


This message is from the T13 list server.


On Tue, 9 Apr 2002 16:57:05 -0700, McGrath, Jim wrote:
>This message is from the T13 list server.
>PAUSE works quite simply. [...]

Thanks! Yes, I agree. But to continue my followup based on Pat's
questions...

As you said, U-DMA defines a "pause" protocol that can be used by the
"receiver" of the data. When the data receiver uses this protocol the
sender shall stop sending data. Questions... Is there an upper limit
on how long the receiver should hold the interface in this "paused"
state? Should there be at least some recommendation for an upper time
limit? I'll add here that I don't see a need for such a
recommendation.

And, as you said, if the "sender" of the U-DMA data wants to slow
down or even stop the data transfer  it may just stoping sending data
(stop the strobe signal toggles). This is not the same as the "pause"
protocol but it has the same effect: data stops moving across the
interface. But here we may have a terminology problem (and that may
be confusing some people?). Questions... Does this action by the
sender have a name? Does it need a name? How many times is this
action used only by the data sender confused with the "pause"
protocol used only by the data receiver?

And to continue... What part of the ATA/ATAPI UDMA protocol
description says that it is valid for the data sender to just stop
sending data? Is it solely the fact that there is no "max" on the
t2cyc or tcyc values in the U-DMA timing table? ... And the answer to
that question is: No, in clause 6.x.2.2 Data transfer phase rules we
find this statement "3) A sender pauses an Ultra DMA burst by not
generating STROBE edges...". OK, but now I have another question: Why
are basic U-DMA protocol rules in clause 6.x? These rules should be
in U-DMA protocol clause (10.2.x) along side the timing diagrams and
timing tables.



*** Hale Landis *** www.ata-atapi.com ***


Reply via email to