[PATCH 4/4] ARM: at91: introduce SAMA5 support

[Nicolas Ferre]

On 03/08/2013 10:27 PM, Joachim Eastwood :
> On 8 March 2013 19:56, Jean-Christophe PLAGNIOL-VILLARD
> <plagnioj at jcrosoft.com> wrote:
>> On 18:18 Fri 08 Mar     , Joachim Eastwood wrote:
>>> On 8 March 2013 17:52, Ludovic Desroches <ludovic.desroches at atmel.com> wrote:
> 
> <snip>
> 
>>>>>
>>>>>> diff --git a/arch/arm/mach-at91/board-dt-sama5.c b/arch/arm/mach-at91/board-dt-sama5.c
>>>>>> new file mode 100644
>>>>>> index 0000000..705305e
>>>>>> --- /dev/null
>>>>>> +++ b/arch/arm/mach-at91/board-dt-sama5.c
>>>>>> @@ -0,0 +1,86 @@
>>>>>> +/*
>>>>>> + *  Setup code for SAMA5 Evaluation Kits with Device Tree support
>>>>>> + *
>>>>>> + *  Copyright (C) 2013 Atmel,
>>>>>> + *                2013 Ludovic Desroches <ludovic.desroches at atmel.com>
>>>>>> + *
>>>>>> + * Licensed under GPLv2 or later.
>>>>>> + */
>>>>>> +
>>>>>> +#include <linux/types.h>
>>>>>> +#include <linux/init.h>
>>>>>> +#include <linux/module.h>
>>>>>> +#include <linux/gpio.h>
>>>>>> +#include <linux/micrel_phy.h>
>>>>>> +#include <linux/of.h>
>>>>>> +#include <linux/of_irq.h>
>>>>>> +#include <linux/of_platform.h>
>>>>>> +#include <linux/phy.h>
>>>>>> +
>>>>>> +#include <asm/setup.h>
>>>>>> +#include <asm/irq.h>
>>>>>> +#include <asm/mach/arch.h>
>>>>>> +#include <asm/mach/map.h>
>>>>>> +#include <asm/mach/irq.h>
>>>>>> +
>>>>>> +#include "at91_aic.h"
>>>>>> +#include "generic.h"
>>>>>> +
>>>>>> +
>>>>>> +static const struct of_device_id irq_of_match[] __initconst = {
>>>>>> +
>>>>>> +       { .compatible = "atmel,sama5d3-aic", .data = at91_aic5_of_init },
>>>>>> +       { /*sentinel*/ }
>>>>>> +};
>>>>>> +
>>>>>> +static void __init at91_dt_init_irq(void)
>>>>>> +{
>>>>>> +       of_irq_init(irq_of_match);
>>>>>> +}
>>>>>> +
>>>>>> +static int ksz9021rn_phy_fixup(struct phy_device *phy)
>>>>>> +{
>>>>>> +       int value;
>>>>>> +
>>>>>> +#define GMII_RCCPSR    260
>>>>>> +#define GMII_RRDPSR    261
>>>>>> +#define GMII_ERCR      11
>>>>>> +#define GMII_ERDWR     12
>>>>>> +
>>>>>> +       /* Set delay values */
>>>>>> +       value = GMII_RCCPSR | 0x8000;
>>>>>> +       phy_write(phy, GMII_ERCR, value);
>>>>>> +       value = 0xF2F4;
>>>>>> +       phy_write(phy, GMII_ERDWR, value);
>>>>>> +       value = GMII_RRDPSR | 0x8000;
>>>>>> +       phy_write(phy, GMII_ERCR, value);
>>>>>> +       value = 0x2222;
>>>>>> +       phy_write(phy, GMII_ERDWR, value);
>>>>>> +
>>>>>> +       return 0;
>>>>>> +}
>>>>>> +
>>>>>> +static void __init sama5_dt_device_init(void)
>>>>>> +{
>>>>>> +       if (of_machine_is_compatible("atmel,sama5d3xcm"))
>>>>>> +               phy_register_fixup_for_uid(PHY_ID_KSZ9021, MICREL_PHY_ID_MASK,
>>>>>> +                       ksz9021rn_phy_fixup);
>>>>>> +
>>>>>> +       of_platform_populate(NULL, of_default_bus_match_table, NULL, NULL);
>>>>>> +}
>>>>>> +
>>>>>> +static const char *sama5_dt_board_compat[] __initdata = {
>>>>>> +       "atmel,sama5",
>>>>>> +       NULL
>>>>>> +};
>>>>>> +
>>>>>> +DT_MACHINE_START(sama5_dt, "Atmel SAMA5 (Device Tree)")
>>>>>> +       /* Maintainer: Atmel */
>>>>>> +       .init_time      = at91sam926x_pit_init,
>>>>>> +       .map_io         = at91_map_io,
>>>>>> +       .handle_irq     = at91_aic5_handle_irq,
>>>>>> +       .init_early     = at91_dt_initialize,
>>>>>> +       .init_irq       = at91_dt_init_irq,
>>>>>> +       .init_machine   = sama5_dt_device_init,
>>>>>> +       .dt_compat      = sama5_dt_board_compat,
>>>>>> +MACHINE_END
>>>>>
>>>>> Do we really need board-dt-sama5.c?
>>>>> Now that we have both irqchip_init and clocksource_of_init it
>>>>> shouldn't be necessary with more than one board-dt.c.
>>>>>
>>>>
>>>> At the beginning, I had the same point of view but we have a new
>>>> architecture, we can't build a single kernel image for both AT91SAM9 and
>>>> SAMA5. So why not splitting it?
>>>
>>> It does add more lines to mach-at91 and I don't see a reason to have
>>> two near identical board-dt files. board-dt could be used for SAMA5
>>> even without irqchip_init and clocksource_of_init so what's the point
>>> in splitting it?
>>>
>>> When the at91 clocksource and irqchip driver are converted/moved we
>>> could support all AT91 SoC from RM9200 to SAMA5 in one board-dt. I
>>> don't see the reason for 3 board-dt files or even 2.
>>>
>>> That we can't build a kernel that supports both ARMv4-5 and ARMv7
>>> doesn't make a difference.
>> my answer is no I do not want to if compatible_is on the c code so no
>> different files
> 
> But you have this already on the above code:
>>>>>> +       if (of_machine_is_compatible("atmel,sama5d3xcm"))
> 
> I don't see the point to have different board-dt files.
> If a board turns out to need lots of specific C code to work I think
> it's fine to have it in its own board file. But right now you are just
> duplicating code unnecessary.

Well, in fact it turns out that we already have quite a few DT tests in
our sama5d3 kernel (github/linux4sam) because all subsystem are not
converted to DT yet. If we put even an extract of this in a common board
file it will soon become a mess.
I also do understand the need for limiting the testing of several
compatibility strings before booting...

Moreover, as discussed with Jean-Christophe, if we join
board-dt-rm9200.c and the sam9 one, we will need to compile
at91rm9200_time.c ST driver, even for sam9s and sama5s which is kind of
useless.

Even if I hate duplicating code, I tend to prefer the "multi-file"
solution as it adds a tiny overhead but leads to a clearer result.

Ludovic, can you please re-spin the whole series with updated patches
and the removal of patch #3 which I plan to queue for 3.9 fixes.

Best regards,
-- 
Nicolas Ferre