Re: [RFC PATCH 1/3] arm64: entry: Remove unnecessary calculation for S_SP in EL1h

[Jungseok Lee <jungseoklee85@gmail.com>]

On Sep 7, 2015, at 11:56 PM, Mark Rutland wrote:

Hi Mark,

> On Fri, Sep 04, 2015 at 03:23:05PM +0100, Jungseok Lee wrote:
>> Under EL1h, S_SP data is not seen in kernel_exit. Thus, x21 calculation
>> is not needed in kernel_entry. Currently, S_SP information is vaild only
>> when sp_el0 is used.
> 
> I don't think this is true. The generic BUG implementation will grab the
> saved SP from the pt_regs, and with this change we'll report whatever
> happened to be in x21 instead.
> 
>> Signed-off-by: Jungseok Lee <jungseoklee85@gmail.com>
>> ---
>> arch/arm64/kernel/entry.S | 2 --
>> 1 file changed, 2 deletions(-)
>> 
>> diff --git a/arch/arm64/kernel/entry.S b/arch/arm64/kernel/entry.S
>> index e163518..d23ca0d 100644
>> --- a/arch/arm64/kernel/entry.S
>> +++ b/arch/arm64/kernel/entry.S
>> @@ -91,8 +91,6 @@
>> 	get_thread_info tsk			// Ensure MDSCR_EL1.SS is clear,
>> 	ldr	x19, [tsk, #TI_FLAGS]		// since we can unmask debug
>> 	disable_step_tsk x19, x20		// exceptions when scheduling.
>> -	.else
>> -	add	x21, sp, #S_FRAME_SIZE
>> 	.endif
>> 	mrs	x22, elr_el1
>> 	mrs	x23, spsr_el1
> 
> Immediately after this we do:
> 
> 	stp     lr, x21, [sp, #S_LR]
> 
> To store the LR and SP to the pt_regs which bug_handler would use.
> 
> Am I missing smoething?

No, You're right. As James mentioned, x21 is used in do_sp_pc_abort.

Thanks for the comment.

Best Regards
Jungseok Lee