Re: Re: [PATCH v3] Fix sun7i pin assignment for IRQ's

[Maxime Ripard <maxime.ripard-wi1+55ScJUtKEb57/3fJTNBPR1lH4CV8@public.gmane.org>]

[-- Attachment #1: Type: text/plain, Size: 2360 bytes --]

Hi,

On Sun, Feb 21, 2016 at 11:27:46AM -0800, Henry Paulissen wrote:
> 
> Op zondag 21 februari 2016 18:18:37 UTC+1 schreef Maxime Ripard:
> 
> >
> > Your commit log is going to need some work. Which bugs? What tests did 
> > you make? Why are you making these changes while the datasheet says 
> > otherwise? 
> >
> 
> Its a fix for a not yet existing bug. I was fiddling around with IRQ's and 
> couldn't get them to work.
> I took a dumpster dive into it and found a shitload of contradicting 
> manuals and datasheets.
> 
> 
> Take for example the A20 user manual:
> http://dl.linux-sunxi.org/A20/A20%20user%20manual%20v1.3%2020141010.pdf
> 
> (pin PI14)
> Page 237: EINT26 is on mux *5* in the pin overview.
> Page 288: EINT26 is on mux *6* in the registers.
>  
> Page 233: EINT12 is on pin PC19 mux6 in the pin overview.
> Page 236: EINT12 is on pin PH12 mux6 in the pin overview.
> Page 253: EINT12 is *not* on pin PC19 on the registers.
> Page 281: EINT12 is on pin PH12 mux6 in the registers.

Ok, so i guess you're actually fixing two different things: the first
one is that some interrupts are using the wrong function, while some
others are just not there at all. It would be great if you could make
two different patches for these.

I guess we could also change SUNXI_FUNCTION_IRQ to enforce the mux
value 6, since you're removing the last users of a different value,
but that can be done as a followup

> So manual may say otherwise, but I hope I have proven that the manual isn't 
> to be trusted.
> 
> My patch is based onto testing from both me and Andre (apritzel).
> He with a Banana PI M1 and me with a Cubietruck (both A20 soc).
> 
> We did a basic test by connecting a pulsing signal to a port and configure 
> kernel to use irq.
> 
> e.g.
> echo pin# > /sys/class/gpio/export
> echo in > /sys/class/gpio/gpio#/direction
> echo rising > /sys/class/gpio/gpio#/edge
> 
> and check on /proc/interrupts to see if a irq was attached and if it was 
> receiving.
> 
> Im not sure what andre his pulse source was, but mine was a 1pps coming 
> from a gps.

That's a great explanation overall, it should just be in the commit
log itself. A git commit is easy to find, the discussion that was
triggered by it not so much.

Thanks!
Maxime


-- 
Maxime Ripard, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com