Re: [RFC PATCH 1/1] spl: introduce SPL_XIP to config

[Sean Anderson <seanga2@gmail.com>]

On 8/15/22 3:27 AM, Nikita Shubin wrote:
> Hello Sean!
> 
> On Sat, 13 Aug 2022 00:35:59 -0400
> Sean Anderson <seanga2@gmail.com> wrote:
> 
>> Hi Nikita,
>>
>> On 8/11/22 5:37 AM, Nikita Shubin wrote:
>>> From: Nikita Shubin <n.shubin@yadro.com>
>>>
>>> U-Boot and SPL don't necessary share the same location, so we might
>>> end with XIP SPL and U-Boot in "normal" memory.
>>>
>>> This adds an option special for SPL to behave it in XIP manner and
>>> don't use hart_lottery and available_harts_lock.
>>>
>>> Signed-off-by: Nikita Shubin <n.shubin@yadro.com>
>>> ---
>>
>> On a stylistic note, typically cover letters are not used for single
>> patches (but feel free to use them for larger series).
>>
>>>    arch/riscv/cpu/cpu.c                 | 2 +-
>>>    arch/riscv/cpu/start.S               | 4 ++--
>>>    arch/riscv/include/asm/global_data.h | 2 +-
>>>    arch/riscv/lib/asm-offsets.c         | 2 +-
>>>    arch/riscv/lib/smp.c                 | 2 +-
>>>    common/spl/Kconfig                   | 5 +++++
>>>    include/configs/scr7_vcu118.h        | 2 ++
>>>    7 files changed, 13 insertions(+), 6 deletions(-)
>>>
>>> diff --git a/arch/riscv/cpu/cpu.c b/arch/riscv/cpu/cpu.c
>>> index 9f5fa0bcb3..89528748d7 100644
>>> --- a/arch/riscv/cpu/cpu.c
>>> +++ b/arch/riscv/cpu/cpu.c
>>> @@ -19,7 +19,7 @@
>>>     * The variables here must be stored in the data section since
>>> they are used
>>>     * before the bss section is available.
>>>     */
>>> -#ifndef CONFIG_XIP
>>> +#if !defined(CONFIG_XIP) && !defined(CONFIG_SPL_XIP)
>> Shouldn't this be
>>
>> #if CONFIG_IS_ENABLED(XIP)
>>
>> ?
>>
>> ditto for the rest of these
> 
> I think you are right, i am a bit confused with CONFIG vs
> CONFIG_IS_ENABLED.

If you have something like

CONFIG_XIP
CONFIG_SPL_XIP
CONFIG_TPL_XIP

then if(CONFIG_IS_ENABLED(XIP)) works out to roughly

if ((defined(TPL_BUILD) && IS_ENABLED(CONFIG_TPL_XIP)) ||
     (defined(SPL_BUILD) && IS_ENABLED(CONFIG_SPL_XIP)) ||
     (!defined(TPL_BUILD) && !define(SPL_BUILD) && IS_ENABLED(CONFIG_XIP)))

That is, the appropriate config is enabled only when building that stage.
There's some other magic as well to make it work in #if as well, but that's
the gist.

>>
>>>    u32 hart_lottery __section(".data") = 0;
>>>    
>>>    /*
>>> diff --git a/arch/riscv/cpu/start.S b/arch/riscv/cpu/start.S
>>> index 26cb877ed1..d824990778 100644
>>> --- a/arch/riscv/cpu/start.S
>>> +++ b/arch/riscv/cpu/start.S
>>> @@ -122,7 +122,7 @@ call_board_init_f_0:
>>>    call_harts_early_init:
>>>    	jal	harts_early_init
>>>    
>>> -#ifndef CONFIG_XIP
>>> +#if !defined(CONFIG_XIP) && !defined(CONFIG_SPL_XIP)
>>>    	/*
>>>    	 * Pick hart to initialize global data and run U-Boot.
>>> The other harts
>>>    	 * wait for initialization to complete.
>>> @@ -155,7 +155,7 @@ call_harts_early_init:
>>>    	/* save the boot hart id to global_data */
>>>    	SREG	tp, GD_BOOT_HART(gp)
>>>    
>>> -#ifndef CONFIG_XIP
>>> +#if !defined(CONFIG_XIP) && !defined(CONFIG_SPL_XIP)
>>>    	la	t0, available_harts_lock
>>>    	amoswap.w.rl zero, zero, 0(t0)
>>>    
>>> diff --git a/arch/riscv/include/asm/global_data.h
>>> b/arch/riscv/include/asm/global_data.h index 9a146d1d49..d71e09c5ab
>>> 100644 --- a/arch/riscv/include/asm/global_data.h
>>> +++ b/arch/riscv/include/asm/global_data.h
>>> @@ -30,7 +30,7 @@ int iccm[CONFIG_NR_CPUS];
>>>    #if CONFIG_IS_ENABLED(SMP)
>>>    	struct ipi_data ipi[CONFIG_NR_CPUS];
>>>    #endif
>>> -#ifndef CONFIG_XIP
>>> +#if !defined(CONFIG_XIP) && !defined(CONFIG_SPL_XIP)
>>>    	ulong available_harts;
>>>    #endif
>>>    };
>>> diff --git a/arch/riscv/lib/asm-offsets.c
>>> b/arch/riscv/lib/asm-offsets.c index f1fe089b3d..bcb3c78654 100644
>>> --- a/arch/riscv/lib/asm-offsets.c
>>> +++ b/arch/riscv/lib/asm-offsets.c
>>> @@ -16,7 +16,7 @@ int main(void)
>>>    {
>>>    	DEFINE(GD_BOOT_HART, offsetof(gd_t, arch.boot_hart));
>>>    	DEFINE(GD_FIRMWARE_FDT_ADDR, offsetof(gd_t,
>>> arch.firmware_fdt_addr)); -#ifndef CONFIG_XIP
>>> +#if !defined(CONFIG_XIP) && !defined(CONFIG_SPL_XIP)
>>>    	DEFINE(GD_AVAILABLE_HARTS, offsetof(gd_t,
>>> arch.available_harts)); #endif
>>>    
>>> diff --git a/arch/riscv/lib/smp.c b/arch/riscv/lib/smp.c
>>> index ba992100ad..cef324954c 100644
>>> --- a/arch/riscv/lib/smp.c
>>> +++ b/arch/riscv/lib/smp.c
>>> @@ -45,7 +45,7 @@ static int send_ipi_many(struct ipi_data *ipi,
>>> int wait) continue;
>>>    		}
>>>    
>>> -#ifndef CONFIG_XIP
>>> +#if !defined(CONFIG_XIP) && !defined(CONFIG_SPL_XIP)
>>>    		/* skip if hart is not available */
>>>    		if (!(gd->arch.available_harts & (1 << reg)))
>>>    			continue;
>>> diff --git a/common/spl/Kconfig b/common/spl/Kconfig
>>> index 07c03d611d..f24e423fc0 100644
>>> --- a/common/spl/Kconfig
>>> +++ b/common/spl/Kconfig
>>> @@ -27,6 +27,11 @@ config SPL_FRAMEWORK
>>>    	  supports MMC, NAND and YMODEM and other methods loading
>>> of U-Boot and the Linux Kernel.  If unsure, say Y.
>>>    
>>> +config SPL_XIP
>>> +	bool "Support SPL in XIP"
>>> +	help
>>> +	  Enable this if SPL is in XIP memory.
>>
>> Can you add another sentence or two? What is "XIP memory"? How does
>> it differ from the standard boot process?
> 
> In this case i treat "XIP memory" as RO memory, if CONFIG_XIP is enabled
> (at least for riscv start.S) we don't use such variables as
> hart_lottery or available_harts_lock, both are locks. The problem is
> that CONFIG_XIP also propagate to main U-Boot, not only SPL.

Great, can you add that to the description?

> For example arch/riscv/cpu/start.S:
> ```
> #ifndef CONFIG_XIP
> 	la	t0, hart_lottery
> 	li	t1, 1
> 	amoswap.w s2, t1, 0(t0)
> 	bnez	s2, wait_for_gd_init
> #else
> 	mv	gp, s0
> 	bnez	tp, secondary_hart_loop
> #endif
> ```
> 
> where tp contains hartid, so we just skip it and all harts with
> non-zero hartid's goto secondary_hart_loop.
> 
> But if we are loading main U-Boot to RAM - this doesn't make sense for
> us since only SPL is in RO memory.
> 
>>
>> Also, it seems like we already have SPL_XIP_SUPPORT. Maybe we can
>> reuse that?
> 
> AFAIK this is different, it's just used to indicate that coping U-Boot
> or kernel is not required if they are located in execute in place
> capable flash memory.

Makes sense

>>
>>>    config SPL_FRAMEWORK_BOARD_INIT_F
>>>    	bool "Define a generic function board_init_f"
>>>    	depends on SPL_FRAMEWORK
>>> diff --git a/include/configs/scr7_vcu118.h
>>> b/include/configs/scr7_vcu118.h index 34545255d1..8f3ba36ce1 100644
>>> --- a/include/configs/scr7_vcu118.h
>>> +++ b/include/configs/scr7_vcu118.h
>>> @@ -29,6 +29,8 @@
>>>    #define SCR7_OCRAM_REGION_SIZE		0xffff
>>>    
>>>    #define SCR7L2_CACHE                    1
>>> +
>>> +#define CONFIG_SYS_UBOOT_BASE           CONFIG_SYS_TEXT_BASE
>>>    #endif
>>>    
>>>    #define CONFIG_SYS_FLASH_BASE           0x00000fffff030000
>>>    
>>
>> What is this part doing? It's not mentioned in the commit message.
> 
> It doesn't have any relation to main part - my bad - a bit relaxed due
> to this being an RFC, sorry.
> 
>>
>> --Sean
>