Re: [REGRESSION][BISECTED] Lenovo IdeaPad touchpad does not work when idma64 is present in initramfs

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Wed, Jan 21, 2026 at 04:02:56PM +0100, Mika Westerberg wrote:
> On Wed, Jan 21, 2026 at 04:54:43PM +0200, Andy Shevchenko wrote:
> > > My understanding is that some models touchpad does not work if the idma64
> > > driver is loaded but they do work if it is not. Also there is some sort of
> > > interrupt flood coming from somewhere?
> > > 
> > > When idma64 is not present the toucpad works flawlessly?
> > 
> > On *old* kernels where idma64 blindly acknowledges all the interrupts.
> > Fixed idma64 doesn't affect the IRQ storm because it's just a given
> > fact on these laptops.
> > 
> > > Does it works in Windows?
> > > 
> > > I mean if it works in Windows and we also know it works in Linux without
> > > idma64 then we can go with some sort of quirk (or better yet figure out
> > > what is going on) to keep users touchpads working. I think this is much
> > > better option than asking BIOS update from vendor which there is close to
> > > zero possibility to happen in reality.
> > 
> > What kind of a quirk? Just to create a module that will request the same IRQ
> > and always acknowledges it? Sounds to me like a heavily papering over solution
> > TBH. (Yes, I assume it may be used only on DMI based enumeration only on the
> > affected models, but still...)
> 
> Well I mean if touchpads actually worked prior this idma64 commit and now
> they don't isn't that a regression?

I don't think so, because commit did the right thing and just revealed an issue
that was rather hidden. Reverting is not an option.

> After that one commit touchpads don't
> work anymore. I would much rather make sure users devices keep functional
> even if it turns out to be hack of some sort.

It's one of the ugliest hacks I ever seen. I really don't want to go this way.
In any case, the problem is with interrupts coming from I²C controller. You can
consider adding that hack into there which sounds at least much better place
than idma64.c which has nothing to do with all this.

> For example could be blacklist idma64? Does that help?

No, it won't.

> Yes the interrupt flood still happens but if that does not affect the
> toucpad then that's still better than non-functional touchpad IMO.

So, the only "best" option I see is to add a quirk module (somewhere in PDx86
area) that just registers an interrupt handler to acknowledge it. Very weird.

-- 
With Best Regards,
Andy Shevchenko