Re: Requestirq() = freeze for touchscreendriver

[Kristoffer Ericson <kristoffer.ericson@gmail.com>]

Greetings,

It does indeed freeze at request_irq() (added some printk's). I've tried to conclude if its due to any IRQ3 changes, but feels doubtful since it worked in the past. Also, when it was a device driver it atleast booted even though the driver didn't work.

Test IRQ3 in all modes and nada (low level, falling, rising). Also tried to remove most of interrupt content but nothing changed.
The interrupt of the touchscreen just loops forever, so any printk's I add will just do the same.

My best guess atm is that the inits (hp6xx / irq / touchscreen) somehow collide and thats why it atleast booted when it was an device driver.

What would happen if request IRQ would try to grab IRQ3 before plat_irq_setup_pins had started? Could it look anything like this?



On Fri, 8 Feb 2008 11:57:07 +0900
"Magnus Damm" <magnus.damm@gmail.com> wrote:

> Hi Kristoffer,
> 
> On Feb 7, 2008 10:24 PM, Kristoffer Ericson
> <kristoffer.ericson@gmail.com> wrote:
> > Greetings,
> >
> > These patches are against vanilla 2.6.24, but should apply cleanly to linux-2.6 also. hp6xx_ts_input.patch creates platform_driver for the touchscreen, not much is kept from the old.
> >
> > The booting comes to "waiting to aquire IRQ" and then freezes. So it happens in one of these lines :
> >
> >         printk(KERN_INFO "ts: waiting to request IRQ\n");
> >         error = request_irq(HP680_TS_IRQ, hp680_ts_interrupt,
> >                         IRQF_DISABLED, "HP6XX Touchscreen Driver", pdev);
> 
> You need to track down exactly why the system is freezing. Some recommendations:
> 
> - Figure out how the device is hooked up to the processor. IRQn? IRLn? PINTn?
> - Verify that you are using the correct IRQ number by calculating it
> from the vector number.
> - Add a printk in the interrupt handler (hp680_ts_interrupt), see if
> you get any printouts.
> - Is the interrupt shared with someone else? Disable that user and try again.
> - Add/modify the irq demuxer in the machvec (.mv_irq_demux) to print
> out all irqs.
> - Remove as many working devices as possible from your kernel and try again.
> - Check that the interrupt sensing is correct. Is it level triggered?
> Active high? Edge triggered?
> - Make sure the mask and priority bit fields are correct in the
> cpu-specific INTC tables.
> 
> Good luck!
> 
> / magnus