Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1380767 > unrolled thread

Re: [PATCH 2/3] ARM: pxa: add pxa25x device-tree support

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [PATCH 2/3] ARM: pxa: add pxa25x device-tree support Arnd Bergmann <arnd@arndb.de> - 2016-04-17 17:30 +0200
    Re: [PATCH 2/3] ARM: pxa: add pxa25x device-tree support Robert Jarzmik <robert.jarzmik@free.fr> - 2016-04-20 08:40 +0200

#1380767 — Re: [PATCH 2/3] ARM: pxa: add pxa25x device-tree support

On Sunday 10 April 2016 21:29:59 Robert Jarzmik wrote:
> +
> +DT_MACHINE_START(PXA_DT, "Marvell PXA3xx (Device Tree Support)")
> +       .map_io         = pxa3xx_map_io,
> +       .init_irq       = pxa3xx_dt_init_irq,
> +       .handle_irq     = pxa3xx_handle_irq,
> +       .restart        = pxa_restart,
> +       .dt_compat      = pxa3xx_dt_board_compat,
> +MACHINE_END
> 

Nothing wrong with your series, it's a straightforward continuation of what
you have for the other platforms, but I have a few comments on the method
overall, and it might be good if you could work on improving those next,
basically eliminating most of the machine descriptor contents in the long
run:

- It would be nice not to call map_io() at all and instead ensure that all
  drivers that have DT bindings use ioremap. The main reason for this is
  that relying on the hardwired mapping makes it easy to get things wrong
  in the bindings, by leaving out required memory ranges.

- The init_irq()/handle_irq() callbacks can probably be replaced with
  a IRQCHIP_DECLARE() statement per irqchip variant, which then goes
  on to initialize the controller and set the handler.

- The restart method is the least important here, but I guess we can
  convert that into a driver, or use an existing one from DT, like
  drivers/power/reset/gpio-restart.c

	Arnd

[toc] | [next] | [standalone]

#1383077

Arnd Bergmann <arnd@arndb.de> writes:

> On Sunday 10 April 2016 21:29:59 Robert Jarzmik wrote:
>> +
>> +DT_MACHINE_START(PXA_DT, "Marvell PXA3xx (Device Tree Support)")
>> +       .map_io         = pxa3xx_map_io,
>> +       .init_irq       = pxa3xx_dt_init_irq,
>> +       .handle_irq     = pxa3xx_handle_irq,
>> +       .restart        = pxa_restart,
>> +       .dt_compat      = pxa3xx_dt_board_compat,
>> +MACHINE_END
>> 
>
> Nothing wrong with your series, it's a straightforward continuation of what
> you have for the other platforms, but I have a few comments on the method
> overall, and it might be good if you could work on improving those next,
> basically eliminating most of the machine descriptor contents in the long
> run:
>
> - It would be nice not to call map_io() at all and instead ensure that all
>   drivers that have DT bindings use ioremap. The main reason for this is
>   that relying on the hardwired mapping makes it easy to get things wrong
>   in the bindings, by leaving out required memory ranges.
Okay, that sounds good, I'll add it to my todo list.
I seem to remember some legacy driver relying on one of these mapping (cpufreq
for pxa relying on memory controller maybe ...), but that's a fix easy enough to
queue. Let's hope I'm not bitten by something else.

> - The init_irq()/handle_irq() callbacks can probably be replaced with
>   a IRQCHIP_DECLARE() statement per irqchip variant, which then goes
>   on to initialize the controller and set the handler.
Okay, I'll verify that, especially that the ordering is ensured, ie. that
interrupts are available at the same time that when it was the machine code
calling the irq init, and that's also going to my todo list.

>
> - The restart method is the least important here, but I guess we can
>   convert that into a driver, or use an existing one from DT, like
>   drivers/power/reset/gpio-restart.c
I think most of the required stuff is already done. The only remaining part is
the reset status clearing specific to pxa, which as you said would ask for a
very tiny driver.

As soon as I'm done with the ac97 rework and if no v4l2 pressure is applied to
pxa, I'll work on this and the MMP to pinctrl conversion.

Thanks for the comments, and cheers.

--
Robert

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web