Re: [Qemu-devel] [PATCH] eepro100: Add more i825xx devices

[Stefan Weil]

Anthony Liguori schrieb:
> Stefan Weil wrote:
>> Anthony Liguori schrieb:
>>  
>>> Stefan Weil wrote:
>>>    
>>>> Anthony, you asked me to send my maintainer version from
>>>> git://repo.or.cz/qemu/ar7.git
>>>> in small patches. Of course, the final goal is to have devices which
>>>> work.
>>>>
>>>> To have a list of devices which should be supported helps other
>>>> people who want
>>>> to contribute to eepro100.c. They won't write code just for i82557c
>>>> if the same
>>>> code should be applied to i82557[ab] as well, for example.
>>>>         
>>> You can do that with comments instead of exposing broken devices to an
>>> end-user.
>>>
>>> Regards,
>>>
>>> Anthony Liguori
>>>
>>>     
>>
>> 1. The new devices are not exposed to end-users, at least not for the
>> moment.
>>   
>
> They are, or will be soon, thanks to qdev.

I thought so.

>
>> 2. Many qemu devices are more or less "broken". This is quite normal
>> for an
>>     emulation, because developers of those emulations only have limited
>>     documentation / resources / testing capabilities.
>>   
>
> Works in some circumstances and completely untested and expected to
> fail are two separate things.
>
>> So a reasonable way might be to expose many devices to end-users, but
>> to classify them as stable / testing / experimental (like it is done
>> for the
>> host and target support).
>>   
>
> Code that has no chance of working shouldn't be in the tree because
> it's untestable.  Untestable code will rot.  More to the point, what's
> the point of having untested code in the tree when the expectation is
> that it won't work at all?  Who does it benefit?
>
> Regards,
>
> Anthony Liguori
>

There is a misunderstanding: all i825xx belong to the same ethernet
controller family,
and all of them will work with drivers which only use standard (or
basic) features,
(as soon as you have applied all patches which are still needed -
without them,
some very basic features like endianess, multicast or simplified mode
don't work) .

So the emulation will already work basically and provide the correct PCI
ids.
Some very simple fixes needed are already in my code base (which still has
to be integrated into qemu), for example checking for i82557c has to be
extended
to check for i8255a, i8255b, too (in most cases).

Nevertheless I don't claim that everything works out of the box because
I cannot test every device with all applications. Other users who need
a special device can test it, because a large part of the code will work
and the missing parts can be fixed thanks to extensive log messages.

Special features of those devices will only work as far as they are
implemented.
Among those features are things like wake-on-lan or power save modi, things
which none of qemu's ethernet drivers emulates today - so no reason for
panic :-)

Regards
Stefan Weil