Re: Contributions: Adding New Devices

[Federico Vaga]

Thanks for your answers

On Fri, Jul 02, 2021 at 12:07:49AM +0200, Philippe Mathieu-Daudé wrote:
> On 7/1/21 9:48 PM, Connor Kuehl wrote:
> > On 6/30/21 7:01 AM, Federico Vaga wrote:
> > > Hello,
> > >
> > > I can't find this information on the website, so here I am.
> > >
> > > I developed a QEMU device that virtualises a PCI card that we widely use at 
> > > CERN.
>
> What kind of device is it? You might be surprise to see later someone
> else interested in what your specific device does and reuse part of it.

It's really specific to our particle accelerators facility. This card receives
packets over a custom network, which payloads contain information about the
accelerator "setup", or synchronization information.

> Is the datasheet/documentation public?
>
> Can you provide test? (so it doesn't bitrot while we do code changes
> unrelated to your device code)

Ideally, with some work, yes.

> > > But this card is only used at CERN.
> > >
> > > Clearly, having CERN specific devices in QEMU does not help much the qemu
> > > community, hence I maintain an internal QEMU fork.
> > >
> > > But, I was wondering what is the QEMU policy about contributions that are known 
> > > to be
> > > used only by a handful of people (one organization in this case)? Are they 
> > > accepted?
> >
> > Your first instinct is correct that it's unlikely that the community
> > will be able to maintain a device if it's really so niche as to only
> > be used at your organization.
> >
> > However, if you do decide to try to upstream it, it could only help
> > your chances if you or some of your colleagues agreed to maintain it
> > for the QEMU community. This mainly involves adding an entry to the
> > MAINTAINERS file where, if accepted, the expectation is that you'll
> > be reachable within reason to review patches, make pull requests,
> > help discuss bugs in the subsystem, etc.
>
> As Connor said (although you probably won't have to worry about
> pull request burden, other will help you while you ramp up).
> Beside, code only usable by you shouldn't have a lot of traffic,
> so maintenance shouldn't take too much.

I could try to submit it, at least to get a review. Then, for us it's not a
problem to apply a patch on to of QEMU to have our virtual device.