[PATCH] env: Add option to only ever append environment

[Marek Vasut]

On 6/2/20 6:00 PM, Tom Rini wrote:
> On Tue, Jun 02, 2020 at 05:55:25PM +0200, Marek Vasut wrote:
>> On 6/2/20 4:38 PM, Tom Rini wrote:
>>> On Tue, Jun 02, 2020 at 02:47:12PM +0200, Marek Vasut wrote:
>>>> On 6/2/20 2:44 PM, Tom Rini wrote:
>>>>> On Tue, Jun 02, 2020 at 02:05:39PM +0200, Rasmus Villemoes wrote:
>>>>>> On 02/06/2020 13.04, Marek Vasut wrote:
>>>>>>> On 6/2/20 8:42 AM, Rasmus Villemoes wrote:
>>>>>>>> On 29/05/2020 19.54, Marek Vasut wrote:
>>>>>>>>> +config ENV_APPEND
>>>>>>>>> +	bool "Always append the environment with new data"
>>>>>>>>> +	default n
>>>>>>>>> +	help
>>>>>>>>> +	  If defined, the environment hash table is only ever appended with new
>>>>>>>>> +	  data, but the existing hash table can never be dropped and reloaded
>>>>>>>>> +	  with newly imported data. This may be used in combination with static
>>>>>>>>> +	  flags to e.g. to protect variables which must not be modified.
>>>>>>>>> +
>>>>>>>>>  config ENV_ACCESS_IGNORE_FORCE
>>>>>>>>>  	bool "Block forced environment operations"
>>>>>>>>>  	default n
>>>>>>>>> diff --git a/env/env.c b/env/env.c
>>>>>>>>> index 024d36fdbe..967a9d36d7 100644
>>>>>>>>> --- a/env/env.c
>>>>>>>>> +++ b/env/env.c
>>>>>>>>> @@ -204,7 +204,9 @@ int env_load(void)
>>>>>>>>>  		ret = drv->load();
>>>>>>>>>  		if (!ret) {
>>>>>>>>>  			printf("OK\n");
>>>>>>>>> +#if !CONFIG_IS_ENABLED(ENV_APPEND)
>>>>>>>>>  			return 0;
>>>>>>>>> +#endif
>>>>>>>>
>>>>>>>> Don't use CONFIG_IS_ENABLED() unless you actually introduce both
>>>>>>>> CONFIG_FOO and CONFIG_SPL_FOO. Otherwise the above
>>>>>>>> CONFIG_IS_ENABLED(ENV_APPEND) is guaranteed to evaluate to false in SPL.
>>>>>>>> Of course that only matters if environment support is enabled in SPL,
>>>>>>>> but some actually use that.
>>>>>>>
>>>>>>> We actually want to use CONFIG_IS_ENABLED() as much as possible to make
>>>>>>> these options future-proof, so that others won't have to chase down all
>>>>>>> kinds of #ifdef CONFIG stuff and fix it later on for SPL/TPL/etc.
>>>>>>>
>>>>>>
>>>>>> That makes no sense. You're introducing something whose help text
>>>>>> doesn't spell out that the option only applies to U-Boot proper, and is
>>>>>> completely ignored in SPL (since CONFIG_SPL_ENV_APPEND never exists).
>>>>>> The reason it's ignored in SPL is that you use the SPL-or-not-SPL-aware
>>>>>> CONFIG_IS_ENABLED() helper, and you say that's so that somebody in the
>>>>>> future can implement CONFIG_SPL_ENV_APPEND?
>>>>>>
>>>>>> If you intend for ENV_APPEND to be something that's either set or not
>>>>>> set for a given board, then the code needs to use the SPL-agnostic
>>>>>> IS_ENABLED(CONFIG_ENV_APPEND). If you intend it to be something that can
>>>>>> be set independently for the env support in SPL vs U-Boot proper, you
>>>>>> need to add both config options and, as you do, use CONFIG_IS_ENABLED.
>>>>>
>>>>> How will this code behave if there is a mismatch between SPL and full
>>>>> U-Boot (disabled SPL, enabled full, as the patch stands today) ?
>>>>
>>>> One will append the environment, the other will override it (if you have
>>>> multiple envs enabled).
>>>
>>> So it sounds like it wouldn't be valid to have this option differ
>>> between SPL and main U-Boot?
>>
>> Consider the case where you have default env in SPL, and multiple envs
>> in U-Boot proper.
> 
> Yes, today you can end up with cases where you build something that doesn't
> work as intended (likely something around falcon boot and/or boot count
> limit in env).  Which is what I'm getting at here.  Is there some
> cases where it would make any sense to enable this option in full U-Boot
> but disable it in SPL?

Yes, like my current use case, where I want to configure the SPL
differently than U-Boot itself. SPL doesn't even have environment
support enabled, but it might be needed later.

And also, I don't want to end up in the same problem we currently have
e.g. with USB gadget, where I have to manually #ifdef CONFIG_SPL_BUILD
#undef CONFIG_ options in the board config file.