Re: [Qemu-devel] [PATCH] Add rtl8139b

["Igor Kovalenko" <igor.v.kovalenko@gmail.com>]

On Wed, Dec 17, 2008 at 1:12 AM, Igor Kovalenko
<igor.v.kovalenko@gmail.com> wrote:
> On Wed, Dec 17, 2008 at 12:36 AM, Andreas Färber <andreas.faerber@web.de> wrote:
>>
>> Am 16.12.2008 um 21:42 schrieb Jamie Lokier:
>>
>>> Igor Kovalenko wrote:
>>>>
>>>> On Tue, Dec 16, 2008 at 10:19 PM, Jamie Lokier <jamie@shareable.org>
>>>> wrote:
>>>>>
>>>>> takasi-y@ops.dti.ne.jp wrote:
>>>>>>
>>>>>> +                        " ne2k_pci pcnet rtl8139 rtl8139b e1000\n");
>>>>>
>>>>> On seeing this in qemu output, and with no documentation, it suggests
>>>>> "rtl8139b" is a later device revision than "rtl8139", and preferable
>>>>> to use for new VMs.
>>>>>
>>>>> Imho it would be better to write " rtl8139b rtl8139 " and of course
>>>>> document that rtl8139b is an older device than rtl8139.
>>>>>
>>>>
>>>> -net nic,model=rtl8139[,quirks=8139cplus|8139b] default to 8139cplus
>>>
>>> Nice!  I'd change to [,revision=8139cplus|8139b], as it's not a quirk.
>>
>> In that case [,revision=cplus|b] would be even simpler. :)
>>
>
> Well, no, C+ is a mode of operation supported on 8139 rev C chip, and
> probably some later ones. Not "all older chips starting with rev b do
> not have it". Extended mode can cause problems with some buggy C+ mode
> drivers so it is sometimes handy to disable C+ mode and revert to the
> other "compatible" but somewhat slower driver.
>
> Maybe original poster's rationale is required to continue :)
>

FWIW RTL8139C board had different revs. Some were exposing PCI ids of
less than 0x20 which now means "no c+ mode supported".

-- 
Kind regards,
Igor V. Kovalenko