Re: [PATCH v2 6/7] arm: rockchip: rk3568: refactor common rk3568_start method

[Michael Riesch <michael.riesch@wolfvision.net>]

Hi Sascha,

On 9/21/22 14:21, Sascha Hauer wrote:
> On Wed, Sep 21, 2022 at 01:57:25PM +0200, Michael Riesch wrote:
>> Hi Sascha,
>>
>> On 9/21/22 11:17, Sascha Hauer wrote:
>>> On Mon, Sep 19, 2022 at 01:39:47PM +0200, Michael Riesch wrote:
>>>> After the removal of the IO domain configuration code, the low-level
>>>> initialization is the same for all RK356x boards. Add a common
>>>> method rk3568_start to remove this code duplication.
>>>>
>>>> Signed-off-by: Michael Riesch <michael.riesch@wolfvision.net>
>>>> ---
>>>>  arch/arm/boards/pine64-quartz64/lowlevel.c    | 30 +----------------
>>>>  arch/arm/boards/radxa-rock3/lowlevel.c        | 27 ---------------
>>>>  .../arm/boards/rockchip-rk3568-evb/lowlevel.c | 33 +------------------
>>>>  .../arm/mach-rockchip/include/mach/rockchip.h |  1 +
>>>>  arch/arm/mach-rockchip/rk3568.c               | 24 ++++++++++++++
>>>>  5 files changed, 27 insertions(+), 88 deletions(-)
>>>>
>>>> diff --git a/arch/arm/boards/pine64-quartz64/lowlevel.c b/arch/arm/boards/pine64-quartz64/lowlevel.c
>>>> index b295885522..e1beb3e624 100644
>>>> --- a/arch/arm/boards/pine64-quartz64/lowlevel.c
>>>> +++ b/arch/arm/boards/pine64-quartz64/lowlevel.c
>>>> @@ -1,39 +1,11 @@
>>>>  // SPDX-License-Identifier: GPL-2.0-only
>>>>  #include <common.h>
>>>> -#include <linux/sizes.h>
>>>> -#include <asm/barebox-arm-head.h>
>>>>  #include <asm/barebox-arm.h>
>>>> -#include <mach/hardware.h>
>>>> -#include <mach/atf.h>
>>>> -#include <debug_ll.h>
>>>>  #include <mach/rockchip.h>
>>>>  
>>>>  extern char __dtb_rk3566_quartz64_a_start[];
>>>>  
>>>> -static noinline void start_quartz64(void *fdt)
>>>> -{
>>>> -	/*
>>>> -	 * Image execution starts at 0x0, but this is used for ATF and
>>>> -	 * OP-TEE later, so move away from here.
>>>> -	 */
>>>> -	if (current_el() == 3)
>>>> -		relocate_to_adr_full(RK3568_BAREBOX_LOAD_ADDRESS);
>>>> -	else
>>>> -		relocate_to_current_adr();
>>>> -
>>>> -	setup_c();
>>>> -
>>>> -	if (current_el() == 3) {
>>>> -		rk3568_lowlevel_init();
>>>> -		rk3568_atf_load_bl31(fdt);
>>>> -		/* not reached */
>>>> -	}
>>>> -
>>>> -	barebox_arm_entry(RK3568_DRAM_BOTTOM, 0x80000000 - RK3568_DRAM_BOTTOM,
>>>> -			  fdt);
>>>> -}
>>>> -
>>>>  ENTRY_FUNCTION(start_quartz64a, r0, r1, r2)
>>>>  {
>>>> -	start_quartz64(__dtb_rk3566_quartz64_a_start);
>>>> +	rk3568_start(__dtb_rk3566_quartz64_a_start);
>>>>  }
>>>
>>> Here __dtb_rk3566_quartz64_a_start is accessed before setup_c() has been
>>> called. That is not allowed, see the patch I just sent.
>>
>> Does the refactoring make sense to you in general? Can I change it to
>>
>> ENTRY_FUNCTION(start_my_fancy_board, r0, r1, r2)
>> {
>> 	setup_c();
>> 	rk3568_start(__dtb_my_fancy_board_start);
>> }
> 
> Well it's not only setup_c() but also the relocate_to_adr_full() or
> relocate_to_current_adr() part that has to be called before setup_c().

Ah OK! Thanks for the clarification.

> At that point there is not much left to factor out to a common function.
> 
> Unless you want to turn this into preprocessor macros (and I don't
> recommand doing that) my suggestion is that we just live with this bit
> of code duplication.

OK, patches 6+7 can be dropped, then.

Best regards,
Michael