[RFC] OMAP DT i2c aliases

[RFC] OMAP DT i2c aliases

[Ivaylo Dimitrov]

Hi,

patch https://lkml.org/lkml/2013/10/16/744 fixes DT i2c bus ids in case 
of deferred probe and assigns id 0,1 and 2 for i2c buses 1,2 and 3. 
Unfortunately, this breaks Maemo userspace on N900, where board code (in 
case of legacy boot) assigns ids 1, 2 and 3, but with DT boot ids are 
0,1 and 2. I've already send a patch https://lkml.org/lkml/2014/6/1/49 
that will allow me to fix that from board .dts, but I was wondering if 
it is the correct way, or ids should be changed in omap3.dtsi for all 
omap devices.

Regards,
Ivo

Re: [RFC] OMAP DT i2c aliases

[Nishanth Menon]

On 06/01/2014 09:28 AM, Ivaylo Dimitrov wrote:
> Hi,
> 
> patch https://lkml.org/lkml/2013/10/16/744 fixes DT i2c bus ids in case 
> of deferred probe and assigns id 0,1 and 2 for i2c buses 1,2 and 3. 
> Unfortunately, this breaks Maemo userspace on N900, where board code (in 
> case of legacy boot) assigns ids 1, 2 and 3, but with DT boot ids are 

ughh.. missed that :(

> 0,1 and 2. I've already send a patch https://lkml.org/lkml/2014/6/1/49 
> that will allow me to fix that from board .dts, but I was wondering if 
> it is the correct way, or ids should be changed in omap3.dtsi for all 
> omap devices.

Should'nt we retain 0,1,2 as indexing to make this consistent for all
SoCs?


-- 
Regards,
Nishanth Menon

Re: [RFC] OMAP DT i2c aliases

[Ivaylo Dimitrov]

On  2.06.2014 18:42, Nishanth Menon wrote:
> On 06/01/2014 09:28 AM, Ivaylo Dimitrov wrote:
>> Hi,
>>
>> patch https://lkml.org/lkml/2013/10/16/744 fixes DT i2c bus ids in case
>> of deferred probe and assigns id 0,1 and 2 for i2c buses 1,2 and 3.
>> Unfortunately, this breaks Maemo userspace on N900, where board code (in
>> case of legacy boot) assigns ids 1, 2 and 3, but with DT boot ids are
>
> ughh.. missed that :(
>
>> 0,1 and 2. I've already send a patch https://lkml.org/lkml/2014/6/1/49
>> that will allow me to fix that from board .dts, but I was wondering if
>> it is the correct way, or ids should be changed in omap3.dtsi for all
>> omap devices.
>
> Should'nt we retain 0,1,2 as indexing to make this consistent for all
> SoCs?
>
>

I think this is the most sane thing, esp if the "alias replace" patch 
gets accepted(thus allowing us to workaround the problems on N900 and 
N9/50), however I wanted to hear from you on the matter. Esp that 
indexing is different with legacy boot compared to DT boot.

Re: [RFC] OMAP DT i2c aliases

[Nishanth Menon]

On Mon 02 Jun 2014 11:03:24 AM CDT, Ivaylo Dimitrov wrote:
[...]
>>> 0,1 and 2. I've already send a patch https://lkml.org/lkml/2014/6/1/49
>>> that will allow me to fix that from board .dts, but I was wondering if
>>> it is the correct way, or ids should be changed in omap3.dtsi for all
>>> omap devices.
>>
>> Should'nt we retain 0,1,2 as indexing to make this consistent for all
>> SoCs?
>>
>>
>
> I think this is the most sane thing, esp if the "alias replace" patch
> gets accepted(thus allowing us to workaround the problems on N900 and
> N9/50), however I wanted to hear from you on the matter. Esp that
> indexing is different with legacy boot compared to DT boot.

I think that slipped my check list unfortunately. :( But then, if we 
think that it is just n900 that is impacted, then I wonder if we can 
override the alias? just wondering..

--
Regards,
Nishanth Menon

Re: [RFC] OMAP DT i2c aliases

[Ivaylo Dimitrov]

On  2.06.2014 19:19, Nishanth Menon wrote:
>
> I think that slipped my check list unfortunately. :( But then, if we
> think that it is just n900 that is impacted, then I wonder if we can
> override the alias? just wondering..
>

That https://lkml.org/lkml/2014/6/1/49 patch will allow such override, I 
tested it on N900 with Fremantle and it works fine. Ofc I had to add

aliases {
	i2c1 = &i2c1;
	i2c2 = &i2c2;
	i2c3 = &i2c3;
};

to omap3-n900.dts (while keeping omap3.dtsi intact) for it to work.

I checked in some Nemo N9/N950 adaptation kernel and it seems those will 
be broken too(and I bet it is the same in stock Nokia N9/50 kernels):

static void __init rm680_i2c_init(void)
{
	omap3_pmic_get_config(&rm680_twl_data, TWL_COMMON_PDATA_USB,
			      TWL_COMMON_REGULATOR_VDAC |
			      TWL_COMMON_REGULATOR_VPLL2);
	omap_pmic_init(1, 2900, "twl5031", INT_34XX_SYS_NIRQ, &rm680_twl_data);
	omap_register_i2c_bus(2, 400, rm696_peripherals_i2c_board_info_2,
			      ARRAY_SIZE(rm696_peripherals_i2c_board_info_2));
	omap_register_i2c_bus(3, 400, rm696_peripherals_i2c_board_info_3,
			      ARRAY_SIZE(rm696_peripherals_i2c_board_info_3));
}

Again 1,2 and 3 for bus indexes just like on N900.

Anyway, I am fine with the alias override. If the patch makes it to the 
upstream :)

Regards,
Ivo