Garrett D'Amore wrote:
> I sure do wish you'd generate a real webrev.  The problem with your 
> diffs is that they lack line numbers, which makes it harder to offer up 
> review comments.

I apologize for the diff (it was generated directly from mercurial). 
I've had a hell of a time getting webrev to function properly with 
mercurial.

> Anyway, here's some thoughts so far:
> 
> 1) no need to initialize macp->m_pdata or macp->p_data_size.   (In fact, 
> those fields should probably *not* be touched.)

Gotcha. nemo-design seemed to indicate these needed to be initialized, 
(NULL and 0 respectively IIRC) I'll take this out. I'll double check 
mac_register() to see if there are any other member initializations that 
should be taken out.

> 2) rather than fix the various DNETTRACE cmn_err's, I'd just remove 
> those blocks --- you're already touching the cmn_err, which isn't 
> needed, so better to clean it up this time around.  (But only the ones 
> like DNETTRACE that just show function entry or exit points.  I've not 
> checked for other DNETDEBUG cases.)

Works for me - I'll cut these out. Would this be an appropriate time to 
begin adding dtrace probes in place of cmn_err? What is the policy for 
adding sdt probes to kernel modules?

> 3) Just a question: can tx_interrupt_mask in the dnet_mc_tx() routine 
> just be treated as a constant?  I'm not sure with the removal of the 
> GLD_INTR_WAIT code that there is any point to having a separate variable 
> here.  But I need to look at the whole function to verify that.
> 
> 4) Ditto for the handling in dnet_intr().

I'll look into this as well. If it can be removed, I'll make sure to get 
rid of it.

> The rest of this looks pretty good.

Great. I've started doing some testing, there are a few things that 
still need to be tweaked, but as soon as I have something that has the 
above changes I'll post another patch (hopefully as a webrev next time).

Steve

-- 
Yet magic and hierarchy
arise from the same source,
and this source has a null pointer.

Reference the NULL within NULL,
it is the gateway to all wizardry.
_______________________________________________
driver-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/driver-discuss

Reply via email to