Re: [PATCH 1/2] ARM: dts: imx6q: extend support for the cm-fx6

[Igor Grinberg <grinberg@compulab.co.il>]

Hi Christopher,

On 06/06/2016 01:37 PM, Christopher Spinrath wrote:
> Hi Igor,
> 
> On 06/06/2016 08:49 AM, Igor Grinberg wrote:
>> Hi Christopher,
>>
>> On 06/05/2016 08:32 PM, Christopher Spinrath wrote:
>>> Hi Igor,
>>>
>>> On 06/05/2016 06:43 PM, Igor Grinberg wrote:
>>>> Hi Christopher,
>>>>
>>>> On 05/26/2016 02:37 PM, Igor Grinberg wrote:
>>>>> On 05/26/2016 12:50 PM, Lucas Stach wrote:
>>>>>> Hi Igor,
>>>>>>
>>>>>> Am Donnerstag, den 26.05.2016, 11:50 +0300 schrieb Igor Grinberg:
>>>>>>> Hi Lucas,
>>>>>>>
>>>>>>> Thanks for reviewing the patch(es).
>>>>>>>
>>>>>>> On 05/23/2016 12:03 PM, Lucas Stach wrote:
>>>>>>>> Am Montag, den 23.05.2016, 00:47 +0200 schrieb
>>>>>>>> christopher.spinrath@rwth-aachen.de:
>>>>>>>>> From: Christopher Spinrath <christopher.spinrath@rwth-aachen.de>
>>>>>>>
>>>>>>> [...]
>>>>>>>
>>>>>>>>> +&ecspi1 {
>>>>>>>>> +	fsl,spi-num-chipselects = <2>;
>>>>>>>>> +	cs-gpios = <&gpio2 30 0>, <&gpio3 19 0>;
>>>>>>>>> +	pinctrl-names = "default";
>>>>>>>>> +	pinctrl-0 = <&pinctrl_ecspi1>;
>>>>>>>>> +	status = "okay";
>>>>>>>>> +
>>>>>>>>> +	flash: m25p80@0 {
>>>>>>>>> +		#address-cells = <1>;
>>>>>>>>> +		#size-cells = <1>;
>>>>>>>>> +		compatible = "st,m25p", "jedec,spi-nor";
>>>>>>>>> +		spi-max-frequency = <20000000>;
>>>>>>>>> +		reg = <0>;
>>>>>>>>> +
>>>>>>>>> +		partition@0 {
>>>>>>>>> +			label = "uboot";
>>>>>>>>> +			reg = <0x0 0xc0000>;
>>>>>>>>> +		};
>>>>>>>>> +
>>>>>>>>> +		partition@c0000 {
>>>>>>>>> +			label = "uboot environment";
>>>>>>>>> +			reg = <0xc0000 0x40000>;
>>>>>>>>> +		};
>>>>>>>>> +
>>>>>>>>> +		partition@100000 {
>>>>>>>>> +			label = "reserved";
>>>>>>>>> +			reg = <0x100000 0x100000>;
>>>>>>>>> +		};
>>>>>>>>
>>>>>>>> Partition layouts don't belong in the upstream DT, as it a device
>>>>>>>> configuration thing. Please kep them in the bootloader/firmware and make
>>>>>>>> this one pass the partition layout to the kernel.
>>>>>>>
>>>>>>> I don't completely agree with this.
>>>>>>> We have lots of partition layouts in the upstream DT.
>>>>>>
>>>>>> No, we don't. At least not for the i.MX6. There are some for the earlier
>>>>>> i.MX boards, but IMO it's wrong to put device configuration into the
>>>>>> upstream DT. Let's not start doing this again.
>>>>>
>>>>> Why not?
>>>>> For i.MX6 there are 2 boards that have the partitioning scheme.
>>>>> I'm not considering this a device configuration, but rather
>>>>> a default partitioning layout/scheme.
>>>>> Current case is for the firmware storage device that is not likely
>>>>> to change.
>>>>> Moreover, a DT is not really a part of the kernel, but lays along the kernel
>>>>> sources for convenience and simplicity (at least IIRC as it was decided
>>>>> about 5 years ago). It is more a part of the firmware for a device, than
>>>>> an upstream kernel source code.
>>>>> I think it is only a meter of time when Linus will decide that he does not
>>>>> want it inside the kernel anymore...
>>>>>
>>>>>>
>>>>>>> Moreover, this is the default layout and changing it, will
>>>>>>> result in incompatibilities and also might result in device "bricking".
>>>>>>> Those can be changed from the boot loader in case you need those
>>>>>>> the other way around.
>>>>>>> Another question of mine is, why should you?
>>>>>>>
>>>>>> Partition layout is device configuration, which is governed by the
>>>>>> device firmware.
>>>>>
>>>>> Yet again, DT is a part of device firmware.
>>>>> Moreover, the firmware (in that case U-Boot), can be configured
>>>>> using the very same DT code, so not having this in, might force
>>>>> various w/a and hacks.
>>>>>
>>>>>> By not having the partition layout in the upstream DT
>>>>>> people are forced to set it from the firmware, which is exactly the
>>>>>> right thing to do, weather or not you plan to change it at any time.
>>>>>
>>>>> I might be ignorant, sorry for that.
>>>>> Why? Why is it right and why would you want to force people to do that?
>>>>>
>>>>>
>>>>
>>>> No answer?
>>>> I think it is worth keeping this as a default firmware layout.
>>>>
>>>
>>> I was not happy removing the layout at first, too. However, they are
>>> added rarely (at least recently) and the binding documentation [1]
>>> states that they shall only be added if there are "strong conventions".
>>
>> Well, not exactly, it says:
>>
>> "Partitions can be represented by sub-nodes of an mtd device. This can be used
>> on platforms which have strong conventions about which portions of a flash are
>> used for what purposes, but which don't use an on-flash partition table such
>> as RedBoot."
>>
>> In my understanding, "can" is _not_ exactly "shall only".
>>
> 
> Definitely, "state" was the wrong word here - sorry about that. But it's
> how I interpret the documentation. It is the only one I have seen that
> specifies an allowed use case. So for me it is more like "deny all,
> allow ...".

That moves the discussion to what "strong conventions" are?
But you have already removed the layout, so this is meaningless.

> 
>>>
>>> Moreover, it is really easy to pass the layout via firmware/u-boot.
>>> Unfortunately, the u-boot provided by CompuLab does not have the
>>> mtdparts commands built-in but the layout can be passed via cmdline as
>>> well. This seems more flexible to me (and I consider this to be good).
>>
>> That means you need a different version of U-Boot if you want to use
>> an upstream kernel, or you need to play the bootargs games, or just leave
>> it a "black box".
>> So currently, from the three options above, the "black box" is chosen, right?
>> That's a pity, because I'd like things to be open...
>>
> 
> Well, It wouldn't be the first ARM device where the newest/upstream
> u-boot is required. Actually, you already have to use the newest version
> if you want to use the igb ethernet.

You mean, igb in U-Boot, right?
Unless I'm missing something, it should work just fine in Linux with
the same U-Boot Compulab provides.

> 
> Currently, IMO the most convenient way (for distributions) would be to
> ship a boot.scr file changing the bootargs (which is not a "black box"
> but a configuration file people are used to). If you don't want to mess
> with bootargs/cmdline it is also possible to add the partition table via
> boot.scr (or on-flash environment) with the fdt command; in this case it
> wouldn't even be hidden in the u-boot binary (and no u-boot upgrade is
> required to implement this).

That is exactly what I said: "to play the bootargs games".

> 
> But yes, if someone wants to use the upstream kernel the boot.scr has to
> be there; but IMO this is something distributions/people providing
> binary files have to take care about. People using their own kernel
> should take into account that they have to provide a bootloader config.

Yep. While if we provide the default layout in DT, all the above would be
unnecessary.

So, if one will neither use the latest U-Boot, nor play the bootargs
games, the flash will stay a kind of "black box" - and that's a pity.

-- 
Regards,
Igor.