lkml.org 
[lkml]   [2014]   [Mar]   [4]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH V3 1/5] genirq: define flag IRQ_SRC_DST_INVERTED, and accessors
On 03/04/2014 03:04 AM, Thomas Gleixner wrote:
> On Mon, 3 Mar 2014, Stephen Warren wrote:
>
>> From: Stephen Warren <swarren@nvidia.com>
>>
>> Some devices have configurable IRQ output polarities. Software might
>> use IRQ_TYPE_* to determine how to configure such a device's IRQ
>> output polarity in order to match how the IRQ controller input is
>> configured. If the board or SoC inverts the signal between the
>> device's IRQ output and controller's IRQ output, software must be
>> aware of this fact, in order to program the IRQ output to the correct
>> (i.e. opposite) polarity. This flag provides that information.
>
> So what you're saying is:
>
> Device IRQ output --> [Optional Inverter Logic] --> IRQ controller input.
>
> And you're storing the information about the presence of the inverter
> logic in the irq itself, but the core does not make any use of it and
> you let the device driver deal with the outcome.
>
> This sucks as all affected drivers have to implement the same sanity
> logic for this.
>
> Why don't you implement a core function which tells the driver which
> polarity to select? That requires a few more changes, but I think it's
> worth it for other reasons.
>
> Right now the set_type logic requires the irq chip drivers to
> implement sanity checking and default selections for TYPE_NONE. We can
> be more clever about that and add this information to the irq chip
> flags.

I don't see any such checking in drivers/irqchip/irq-gic.c; it rejects
any type other than IRQ_TYPE_LEVEL_HIGH or IRQ_TYPE_EDGE_RISING, and I
don't see any mention of TYPE_NONE in that file. Is the driver incomplete?

Instead of adding all this extra logic to the core, what do you think of
simply telling each driver that has a configurable interrupt output
polarity exactly which polarity to use. This information would come from
device tree or platform data. This is what I implemented in V1/V2 of
this series:

http://www.spinics.net/lists/devicetree/msg23648.html

Is that any better, or do you definitely want the IRQ core to manage this?


\
 
 \ /
  Last update: 2014-03-04 17:41    [W:0.148 / U:1.224 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site