Re: [PATCH 0/2] UIO: Optional custom threaded IRQ handler

From: Greg KH <gregkh@linuxfoundation.org>
To: Kevin Bube <k.bube@web.de>
Cc: kernelnewbies@kernelnewbies.org
Subject: Re: [PATCH 0/2] UIO: Optional custom threaded IRQ handler
Date: Mon, 5 Jan 2026 13:45:41 +0100	[thread overview]
Message-ID: <2026010503-vitamins-slam-fa50@gregkh> (raw)
In-Reply-To: <87ikdgqm2m.fsf@web.de>

On Mon, Jan 05, 2026 at 01:29:53PM +0100, Kevin Bube wrote:
> Greg KH <gregkh@linuxfoundation.org> writes:
> > On Mon, Jan 05, 2026 at 11:21:48AM +0100, Kevin Bube wrote:
> >> Greg KH <gregkh@linuxfoundation.org> writes:
> >> > On Sat, Jan 03, 2026 at 09:06:29PM +0100, Kevin Bube wrote:
> >> >> Add an optional threaded IRQ handler to the generic UIO driver. This allows
> >> >> driver developers to keep the fast IRQ handler small and move tasks to the IRQ
> >> >> thread. We also update the documentation accordingly.
> >> >> 
> >> >> Kevin Bube (2):
> >> >>   UIO: Add optional IRQ thread callback
> >> >>   Documentation: Document threaded IRQ handler for UIO
> >> >> 
> >> >>  Documentation/driver-api/uio-howto.rst | 8 ++++++++
> >> >>  drivers/uio/uio.c                      | 6 +++++-
> >> >>  include/linux/uio_driver.h             | 2 ++
> >> >>  3 files changed, 15 insertions(+), 1 deletion(-)
> >> >
> >> > But as no in-kernel driver now uses this, it will just be removed again
> >> > :(
> >> >
> >> > Please always provide real users for api changes, otherwise we can't
> >> > determine if they are actually working or not.
> >> 
> >> I see. Thanks for taking a look.
> >
> > How was this tested if you don't have an actual user for it?  Why add it
> > if you don't have a user for it?
> 
> We have special hardware with a FPGA that requires its own customized
> driver at my dayjob. While the driver is GPL, the FPGA image is not (not
> my decision). So it makes little sense to include the driver in
> mainline.

There's no reason not to include it, so why not submit it?

> The patch is just a part of the whole change we make for our device. I
> sent it because I think it may be useful for others, too. But I
> understand that it is bad to add an API without knowing its users and
> not being able to test it. My misunderstanding was that the generic UIO
> driver is more of a framework to derive custom drivers.

We can't ever support out-of-tree drivers for obvious reasons, so just
submit your driver and then it should be fine.

thanks,

greg k-h

_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
https://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies