[U-Boot] [u-boot] [PATCH] arm: sleep: Get the entry point of kernel from SPARE4 register

[Scott Wood]

On 04/05/2016 09:16 PM, Huan Wang wrote:
> Hi, York and Scott,
> 
>> On 04/05/2016 05:11 AM, Alison Wang wrote:
>>> For LS1021A Secure Boot, SPARE2 register is used and modified by the
>>> IBR. To avoid the conflict, SPARE4 is used instead of SPARE2 to store
>>> the entry point of kernel. This patch is to get the entry point of
>>> kernel from SPARE4 instead of SPARE2.
>>>
>>> Signed-off-by: Alison Wang <alison.wang at nxp.com>
>>> ---
>>>  board/freescale/common/arm_sleep.c | 2 +-
>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/board/freescale/common/arm_sleep.c
>>> b/board/freescale/common/arm_sleep.c
>>> index 71ed15e..6d967f0 100644
>>> --- a/board/freescale/common/arm_sleep.c
>>> +++ b/board/freescale/common/arm_sleep.c
>>> @@ -88,7 +88,7 @@ int fsl_dp_resume(void)
>>>  	dp_resume_prepare();
>>>
>>>  	/* Get the entry address and jump to kernel */
>>> -	start_addr = in_le32(&scfg->sparecr[1]);
>>> +	start_addr = in_le32(&scfg->sparecr[3]);
>>>  	debug("Entry address is 0x%08x\n", start_addr);
>>>  	kernel_resume = (void (*)(void))start_addr;
>>>  	secure_ram_addr(_do_nonsec_entry)(kernel_resume, 0, 0, 0);
>>>
>> Alison,
>>
>> Does this change need to be in sync with Kernel change?
>>
>> York
>>
>> Where does this get written?
>>
>> -Scott
> [Alison Wang] Thanks for your replies. Your concerns are right.
> SPARE4 register needs to be written in kernel.
> 
> This is an issue about deep sleep in LS1021A Secure Boot. It is found
> in SDK2.0. The corresponding patch for kernel is sent in SDK2.0. 
> 
> Well, deep sleep uses an old way in SDK2.0. For upstream, deep sleep
> patches haven't been sent out as it will use PSCI and there are some
> issues about PSCI. So the corresponding patch for kernel can't be sent
> out now.

It's not about when the patch is sent.  It's about managing
compatibility.  There needs to be some way to communicate what the
expectations are between Linux and U-Boot, or to limit the change to
chips where this feature has never worked before.  We can't introduce
regressions when the kernel is updated but not U-Boot, and regressions
when U-Boot is updated but not the kernel are almost as bad.

-Scott