Felix Feng wrote:
> Darren J Moffat ??:
>> Frank Che wrote:
>>> I'm sponsoring this case for Felix Feng. It is to provide a new
>>> wireless driver, uath(7D), for Atheros AR5523 USB wireless devices.
>>> The requested release binding is patch/micro.
>>>
>>> I have marked it as "closed approved automatic" since the change is
>>> straightforward and the interface is clear. If there is request, I'd be
>>> happy to upgrade it to a fast track and set the timer.
>>
>> Why is a different driver required just because this is USB rather
>> than PCI attached ?
>>
> Not just the difference of interface. Please see the details below.
> 1. The uath driver supports atheros5k USB chipset. The ath driver
> supports ath5k PCI/PCI-E chipset. They are two different series.
> 2. The uath driver need to download firmware to initial hardware. But
> ath driver do not need to do so. And the hardware operation is different.
> 3. The whole working mechanism is different between uath driver and
> ath driver. The uath driver depends on USBA. But the ath driver relies
> on PCI configuration.
> 4. For the same part of wifi driver, it has been abstracted as
> net80211 module.
In general, for any two parts, one PCI and one USB, you will need a
different driver. The interfaces between USB and PCI are so totally
different that while it might be theoretically possible to build a
single driver, you'd wind up with a single driver containing most of the
functionality of two separate drivers within it.
(USB drivers communicate with devices using the STREAMs mblks and USBA.
PCI drivers on the other hand use mapped registers and possibly DMA.
Even the way interrupt handling works is totally different between USB
and PCI.)
This isn't a good design choice. Better choice is separate drivers, as
this project team has done.
-- Garrett
>
> ------
> Thanks,
> Felix
