On Fri, 2005-07-29 at 15:33 +0200, Arne Redlich wrote:
> Am Freitag, den 29.07.2005, 08:00 -0500 schrieb Brian Wolfe:
> > On Fri, 2005-07-29 at 11:07 +0200, Arne Redlich wrote:
> > > Below's a patch to fix IET's inability of handling separate MaxRxDSLs
> > > for initiator and target.
> > >
> > > I didn't touch the kernel module's sanity check of the MaxRxDSLs,
> > > although I feel it's currently too restrictive. If a too large value (>
> > > 256k) is configured/announced for the target's/initiator's MaxRxDSL, the
> > > minimum value (512) is chosen, which is IMHO a rather hard punishment.
> > > I'd prefer the target to use the maximum supported value in this case.
> > > Any thoughts on this?
> > >
> > > Arne
> >
> > Why not just go back to the default value on request errors like this
> > instead of picking an arbitrary number?
>
> Because we're not talking about request errors, but IET's (arbitrary)
> limit of 256k. If an initiator *declares* via it's MaxRxDSL that it can
> handle larger (valid) values, I think it's not the worst idea to use the
> max. supported data segment length and not the default (8k) or even the
> min. value (512). So at least when sending to the initiator I'm strongly
> in favour of this approach.
>
agree. that punishment is too huge. largest supported value is a good
fit here.
> Arne
ming
-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO September
19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
|
