[U-Boot] [PATCH v2 2/4] ARM: armv7: guard memory reserve for PSCI with #ifdef CONFIG_ARMV7_PSCI

[Jon Masters]

On 09/23/2016 01:09 AM, Jon Masters wrote:
> On 08/30/2016 03:22 AM, Masahiro Yamada wrote:
>> If CONFIG_ARMV7_NONSEC is enabled, the linker script requires
>> CONFIG_ARMV7_PSCI_NR_CPUS regardless of CONFIG_ARMV7_PSCI.
>>
>> Signed-off-by: Masahiro Yamada <yamada.masahiro at socionext.com>
>> ---
>>
>> Changes in v2: None
>>
>>  arch/arm/cpu/u-boot.lds | 4 ++--
>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/arch/arm/cpu/u-boot.lds b/arch/arm/cpu/u-boot.lds
>> index 36c9fd0..0a5fae6 100644
>> --- a/arch/arm/cpu/u-boot.lds
>> +++ b/arch/arm/cpu/u-boot.lds
>> @@ -83,10 +83,10 @@ SECTIONS
>>  #endif
>>  	{
>>  		KEEP(*(.__secure_stack_start))
>> -
>> +#ifdef CONFIG_ARMV7_PSCI
>>  		/* Skip addreses for stack */
>>  		. = . + CONFIG_ARMV7_PSCI_NR_CPUS * ARM_PSCI_STACK_SIZE;
>> -
>> +#endif
>>  		/* Align end of stack section to page boundary */
>>  		. = ALIGN(CONSTANT(COMMONPAGESIZE));
>>  
>>
> 
> In the Fedora U-Boot ELF builds being tested for 32-bit U-Boot on KVM
> (AArch32 on ARMv8 hardware, as a means of decommissioning older v7
> builders and replacing them with VMs on v8 hosts) this results in the
> following for a vexpress_ca15_tc2 U-Boot build target:
> 
>   2 .secure_text  00000130  8082d000  8082d000  0003d000  2**5
>                   CONTENTS, ALLOC, LOAD, READONLY, CODE
>   3 .secure_stack 00000000  8082e000  8082e000  0004e000  2**0
>                   ALLOC
>   4 .__secure_end 00000004  8082e000  8082e000  0003e000  2**0
>                   CONTENTS, ALLOC, LOAD, DATA
> 
> Which QEMU then refuses to load because there's no space allocated for
> the secure_stack. What do you think we should do about this? I haven't
> been tracking the PSCI implementation (which is pretty cool btw, woot!)
> work you've been doing and we don't really need it for the VM case
> anyway, just need the resulting build to link correctly. So sure,
> probably I/we need to understand the status of the PSCI work too.

Ok, so the specific problem seems to be:

commit 5a3aae68c74e029ed29d22999ab6f22687382f2d
Author: Masahiro Yamada <yamada.masahiro at socionext.com>
Date:   Tue Aug 30 16:22:21 2016 +0900

    ARM: armv7: guard memory reserve for PSCI with #ifdef CONFIG_ARMV7_PSCI

    If CONFIG_ARMV7_NONSEC is enabled, the linker script requires
    CONFIG_ARMV7_PSCI_NR_CPUS regardless of CONFIG_ARMV7_PSCI.

    Reviewed-by: Alexander Graf <agraf at suse.de>
    Signed-off-by: Masahiro Yamada <yamada.masahiro at socionext.com>

This introduced the change that results in no secure stack space being
allocated in the ELF image. The vexpress_ca15_tc2 target doesn't claim
to support PSCI /however/ the default includes say that we are
ARMV7_NONSEC and CONFIG_ARMV7_PSCI_NR_CPUS will default to 4 in that
case regardless, so this is why we were previously allocating space for
the (presumably unused) secure_stack (prior to the above commit). That
commit changed things by wrapping it in a conditional on ARMV7_PSCI.

Can someone who is actually a U-Boot person tell us what the intention
is in this case, and what should actually happen to the linker script?

Jon.

-- 
Computer Architect | Sent from my Fedora powered laptop