Re: Linux irq subsys i2c interaction question

[Thomas Gleixner]

On Tue, 7 Mar 2017, Hans de Goede wrote:
> On 07-03-17 10:18, Thomas Gleixner wrote:
> > I'm a bit worried about this being hardcoded for that particular use
> > case. That also means that you cannot use the generic regmap irq handling
> > stuff and need to have your own irq magic there.
> 
> Yeah, although the amount of irqchip code needed is pretty minimal
> for this (otherwise) simple device.
> 
> > Wouldn't it be smarter to
> > mark that interrupt in some way and have that as a generic option?
> 
> Such a generic option is what I was looking for before going this
> route, so yeah +1 for that. I'm not familiar enough with the
> irq code to feel comfortable adding that, but if you can provide
> a patch, I can test it.
> 
> > Can you point me at the relevant drivers/patches?
> 
> Here is my current wip stuff (I've pushed this to a for-irq-discussion
> branch so that the commits won't go away when I rebase my wip
> branch).
> 
> mfd patches:
> https://github.com/jwrdegoede/linux-sunxi/commit/a0b9ee4ba7b8206880415351a53d5fbbd5969ba6
> https://github.com/jwrdegoede/linux-sunxi/commit/a2bff50785c31239ec699a8b43e827c40089ee1f
> https://github.com/jwrdegoede/linux-sunxi/commit/708985e8f33471976359c9ff52e2db9964e3ddf4
> 
> i2c-adapter patch:
> https://github.com/jwrdegoede/linux-sunxi/commit/88a671b3543ffd1f5f4251f4028e4a5d67d88fda
> 
> Note the mfd driver has its own irqchip using the regmap-irq.c stuff (second
> commit)
> and the trouble some irqchip driver is nested from that in the i2c-adapter
> driver
> (the last commit).

As this is anyway using it's own interrupt handler and not using the
generic regmap one, this should be fine.

Thanks,

	tglx