On Thu, Jun 06, 2002 at 01:11:06PM -0700, David Brownell wrote:
> Matthew Dharm wrote:
> > My only concern is that the behavior be _very_ well defined for this.  I
> > can think of all sorts of corner-cases (URBs about to complete, endpoints
> > halted, zero-length packets, etc) which all need to be covered before we
> > make a change like this.
> 
> I'm not clear on what you're saying:
> 
> - Are you implying that this change could do more than just
>    remove some unneeded failure modes?  (If so, what?)

I'm concerned that those failure modes are not 'unneeded'.  Likely they
are, but the only way to be sure is to change it and see what happens.

> - Or are you instead implying that the current behavior for
>    queued URBs is not defined well enough to suit you?

This is my big concern.  While, as you point out, races and other such
problems must be addressed, it is critical that they be addressed in a
_completely_consistent_ manner across all HCDs.

This is the primary reason I've avoided using this feature until now.

My guess is that there really is no "correct" behavior in some of these
cases, which means that we're going to have to make some serious design
decisions.  One case I'm specifically worried about is what happens when an
URB in the middle of a series of queued URBs terminates with a short
packet?  How does that affect the URBs that come after it, if there is any
affect at all?  I'd like to see all those URBs terminated immediately, but
I'm sure others would like to see them all executed as if nothing had
happened.

(This, BTW, is why I wanted URBs to accept scatter-gather lists, so the
driver author could distinguish between these cases.)

It's thinks like this which concern me greatly.  I really don't care which
direction we go in, but I want to make sure that _all_ HCDs will respond
_exactly_ the same way.

Matt

-- 
Matthew Dharm                              Home: [EMAIL PROTECTED] 
Maintainer, Linux USB Mass Storage Driver

I'm a pink gumdrop! How can anything be worse?!!
                                        -- Erwin
User Friendly, 10/4/1998

Attachment: msg07081/pgp00000.pgp
Description: PGP signature

Reply via email to