[U-Boot] [PATCH 5/10] common/env: Point default environment for GD

[Prabhakar Kushwaha <prabhakar@freescale.com>]

On 4/5/2014 5:21 AM, Scott Wood wrote:
> On Wed, 2014-04-02 at 09:12 +0530, Prabhakar Kushwaha wrote:
>> On 4/2/2014 3:34 AM, Scott Wood wrote:
>>> On Mon, 2014-03-31 at 15:34 +0530, Prabhakar Kushwaha wrote:
>>>> GD(Global Data) structure has pointer to environment variable array.
>>>> but, it always point to default_environment assuming it is running from
>>>> final location.
>>>>
>>>> So update GD pointer with env variable array during SPL boot.
>>>>
>>>> Signed-off-by: Prabhakar Kushwaha <prabhakar@freescale.com>
>>>> ---
>>>>    common/env_common.c |    3 +++
>>>>    1 file changed, 3 insertions(+)
>>>>
>>>> diff --git a/common/env_common.c b/common/env_common.c
>>>> index c0bfc2f..043150a 100644
>>>> --- a/common/env_common.c
>>>> +++ b/common/env_common.c
>>>> @@ -162,6 +162,9 @@ int env_import(const char *buf, int check)
>>>>    	if (himport_r(&env_htab, (char *)ep->data, ENV_SIZE, '\0', 0,
>>>>    			0, NULL)) {
>>>>    		gd->flags |= GD_FLG_ENV_READY;
>>>> +#ifdef CONFIG_SPL_BUILD
>>>> +		gd->env_addr = (unsigned long)ep->data;
>>>> +#endif
>>>>    		return 1;
>>>>    	}
>>>>    
>>> Could you explain a bit more about how the environment is being loaded
>>> during SPL, and how gd->env_addr gets set for non-SPL?
>>>
>>>
>> during SPL or NON- XIP boot (SD, SPI, NAND) in order to access env
>> variables following 2 functions are called sequentially.
>> 1. env_init() : Initialize gd->env_addr with default_environment array
>> 2. env_relocate(). It copied data from flash to heap memory area and
>> call env_import().
>>               here,  This env pointer *buf is used for importing.  but
>> gd->env_addr never updated with new address of env variables.  It always
>> remain pointed to default_environment array set in env_init().
>> It looks to be a issue for non-XIP boot.  As I was not confident about
>> it. So limited to only SPL framework only.
>>
>> for non-SPL i.e. XIP boot (NOR)
>>    env_init (): gd->env_addr always point to NOR address containing
>> env_variables.  Because of XIP No need of copy from flash to buffer.
>> Hence no problem.
> That doesn't really answer my question of how it gets initialized for
> NOR flash.  What line in what file does the initialization?  It looks
> like the answer is env_init() -- so it doesn't "always" hold the
> address.  Why can't env_init() do this on NAND, SPI, etc, provided you
> actually have the environment available before relocation[1]?
>
> I'm also not sure what it has to do with SPL as opposed to non-XIP.
> Doesn't this also apply for non-SPL PBL boot, etc?  Why wouldn't you do
> this in env_relocate_spec()?
>
> BTW, I notice a few Freescale boards initialize this in their spl.c
> file.
>
> Also, you don't mention what the actual bug is you see.  What uses
> gd->env_addr after relocation, other than api/?  I'm not saying that
> api/ isn't important, but I'm wondering if you saw an effect elsewhere.
>
> -Scott
>
> [1] I realize that in many cases we don't, but that's something we need
> to fix, since it's important to how DDR gets set up among other things.
>

I analyzed further with P5020_NAND boot.  i found following observation:
1. env_init happens in board_init_f
2. DDR always initialized with default_environment
3. env_relocate happens in board_init_r

Looks like, during non-XIP & non - SPL boot, ddr can only be initialized 
with fixed settings.
is it a accepted behavior? if Yes, DDR init parameters cannot be changed 
like ECC on/off.

Regards,
Prabhakar