Re: [Qemu-devel] Large USB patch

[nix . wie . weg]

Hello Fabrice,

I can understand, that you are not so happy with such an huge patch. But
sorry it can't be done with just patching the state machine. The main
problem is, that the cvs source needs the usbhub. But this hub is not
operational at its current state. So in fact whenever we try to find a
usb problem, we have to look in at least 4 completly different USB
components (uhci controller, usb state machine, usb hub, usb device). It
is a great help to reduce these components by one. Now if you go so far,
you have to touch all these 15 files and then it makes no sence not to
implement a better, and more flexible api. I have done this work and I
was successful, as in the last days we (Lonnie Mendez and I) have found
a lot of bugs. Some bugs were surely introduced by the new api, but we
have also fixed bugs which were simply not detected earlier because the
USB system was never bevor working so good. Today I have even detected
why my usb scanner was not recognised. I will add the patch later on.

So for the foregoing reasons, I will clearly reject to produce a patch
which will only cover the state machine. I have worked very hard on that
patch and some times you have to take what you get :)

But as I said before I can understand, that it is a problem to add such
a huge patch. So I want to submit the following offer: I and maybe also
Lonnie Mendez (hey Lonnie this is a request for some more of your
wonderful help :) ) will test the patch one week longer and get it a
little more stable. On saturday I will make another comprehensive patch
and you will please add it to cvs. Is that an offer?

Thanks for your time and your work so far,
Tino H. Seifert

PS: Your feature request is noted and I will look into it after we have
successfully dealt with the issues above. At the moment I'm not sure if
it is possible at all.

Fabrice Bellard wrote:
> Hi,
>
> Could you make a small patch containing just the bug fixes of the
> state machine ?
>
> Concerning your API changes, I am relunctant to include them now, but
> my mind can evolve. An API change that I would consider as very useful
> could be to be able to make asynchronous USB I/Os to avoid blocking
> QEMU while doing host USB I/Os.
>
> Regards,
>
> Fabrice.
>
>
>