My gut feeling is that GLDv3 pretty much needs to "require" VLAN support
and link notification (in particular link notification is required for
certain GLDv3 consumers such as NWAM), although I've not looked to see
if Clearview has addressed that. dnet is indeed a very very crufty driver.
I'd be willing to consider that link notification didn't work properly
for SYM phys. Those cards are pretty rare, I think.
-- Garrett
Steven Stallion wrote:
> Garrett D'Amore wrote:
>
>> I had planned on sponsoring you, and I still can -- *however*, right now
>> I'm fairly saturated so it might be a few weeks before I can properly
>> re-review and test this stuff.
>>
>
> No worries, I'm in no rush; I'd rather spend more time testing than not.
>
>
>> If anyone else from Sun wants to step up and help Steve out here, that
>> would be great. Otherwise, there may be a couple of weeks before I can
>> get to it.
>>
>>
>
> I've run into a number of issues updating dnet. It seems that the
> initialization behavior is broken with the current version of dnet
> (tested on both snv_78 and snv_85).
>
> Probably the biggest issue at the moment is the driver does not
> recognize MII on the 21143. I am also noticing that dnet does not
> correctly negotiate the active media block. If the driver is loaded
> while plugged into a 10/100mbps switch, it will negotiate to half
> duplex. If loaded when not plugged into the switch, it will select full
> duplex. DHCP client negotiation also does not work.
>
> (Ben Taylor has lent out a 21140 multi-port Adaptec for testing, I will
> probably switch over to that board for a spell to see if the behavior is
> any different.)
>
> That said, I would like to back out the changes made for handling VLAN
> tags and link notification, and submit the GLDv3 conversion bits only.
>
> The reason I would like to do this is dnet has several deeper issues at
> the moment other than lacking proper link notification and handling VLAN
> tags. The hardware initialization is broken, the soft state does not
> track ifspeed or duplexing, and there are some sizable gaps between MII
> and non-MII PHY's. Making these changes would make the submitted patches
> that much more difficult to read and test.
>
> I would like to work toward modernizing this driver in a second phase by
> fixing the above issues, and by working in Brussels support if at all
> possible.
>
> Thoughts?
>
> Steve
>
>
_______________________________________________
driver-discuss mailing list
driver-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/driver-discuss
