Re: [PATCH v2 2/2] ARM: support PHYS_OFFSET minimum aligned at 64KiB boundary

[Leizhen (ThunderTown)]

On 2020/9/16 15:57, Russell King - ARM Linux admin wrote:
> On Wed, Sep 16, 2020 at 09:57:15AM +0800, Leizhen (ThunderTown) wrote:
>> On 2020/9/16 3:01, Russell King - ARM Linux admin wrote:
>>> On Tue, Sep 15, 2020 at 09:16:15PM +0800, Zhen Lei wrote:
>>>> Currently, only support the kernels where the base of physical memory is
>>>> at a 16MiB boundary. Because the add/sub instructions only contains 8bits
>>>> unrotated value. But we can use one more "add/sub" instructions to handle
>>>> bits 23-16. The performance will be slightly affected.
>>>>
>>>> Since most boards meet 16 MiB alignment, so add a new configuration
>>>> option ARM_PATCH_PHYS_VIRT_RADICAL (default n) to control it. Say Y if
>>>> anyone really needs it.
>>>>
>>>> All r0-r7 (r1 = machine no, r2 = atags or dtb, in the start-up phase) are
>>>> used in __fixup_a_pv_table() now, but the callee saved r11 is not used in
>>>> the whole head.S file. So choose it.
>>>>
>>>> Because the calculation of "y = x + __pv_offset[63:24]" have been done,
>>>> so we only need to calculate "y = y + __pv_offset[23:16]", that's why
>>>> the parameters "to" and "from" of __pv_stub() and __pv_add_carry_stub()
>>>> in the scope of CONFIG_ARM_PATCH_PHYS_VIRT_RADICAL are all passed "t"
>>>> (above y).
>>>>
>>>> Signed-off-by: Zhen Lei <thunder.leizhen@xxxxxxxxxx>
>>>> ---
>>>>  arch/arm/Kconfig              | 18 +++++++++++++++++-
>>>>  arch/arm/include/asm/memory.h | 16 +++++++++++++---
>>>>  arch/arm/kernel/head.S        | 25 +++++++++++++++++++------
>>>>  3 files changed, 49 insertions(+), 10 deletions(-)
>>>>
>>>> diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
>>>> index e00d94b16658765..19fc2c746e2ce29 100644
>>>> --- a/arch/arm/Kconfig
>>>> +++ b/arch/arm/Kconfig
>>>> @@ -240,12 +240,28 @@ config ARM_PATCH_PHYS_VIRT
>>>>  	  kernel in system memory.
>>>>  
>>>>  	  This can only be used with non-XIP MMU kernels where the base
>>>> -	  of physical memory is at a 16MB boundary.
>>>> +	  of physical memory is at a 16MiB boundary.
>>>>  
>>>>  	  Only disable this option if you know that you do not require
>>>>  	  this feature (eg, building a kernel for a single machine) and
>>>>  	  you need to shrink the kernel to the minimal size.
>>>>  
>>>> +config ARM_PATCH_PHYS_VIRT_RADICAL
>>>> +	bool "Support PHYS_OFFSET minimum aligned at 64KiB boundary"
>>>> +	default n
>>>
>>> Please drop the "default n" - this is the default anyway.
>>
>> OK, I will remove it.
>>
>>>
>>>> @@ -236,6 +243,9 @@ static inline unsigned long __phys_to_virt(phys_addr_t x)
>>>>  	 * in place where 'r' 32 bit operand is expected.
>>>>  	 */
>>>>  	__pv_stub((unsigned long) x, t, "sub", __PV_BITS_31_24);
>>>> +#ifdef CONFIG_ARM_PATCH_PHYS_VIRT_RADICAL
>>>> +	__pv_stub((unsigned long) t, t, "sub", __PV_BITS_23_16);
>>>
>>> t is already unsigned long, so this cast is not necessary.
>>
>> Oh, yes, yes, I copied from the above statement, but forgot to remove it.
>>
>>>
>>> I've been debating whether it would be better to use "movw" for this
>>> for ARMv7.  In other words:
>>>
>>> 	movw	tmp, #16-bit
>>> 	adds	%Q0, %1, tmp, lsl #16
>>> 	adc	%R0, %R0, #0
>>>
>>> It would certainly be less instructions, but at the cost of an
>>> additional register - and we'd have to change the fixup code to
>>> know about movw.
>>
>> It's one less instruction for 64KiB boundary && (sizeof(phys_addr_t) == 8),
>> and no increase or decrease for 64KiB boundary && (sizeof(phys_addr_t) == 4),
>> but one more instruction for 16MiB boundary.
>>
>> And maybe: 16MiB is widely used, but 64KiB is rarely used.
>>
>> So I'm inclined to the current revision.
> 
> Multiplatform kernels (which will be what distros build) will have to
> enable this option if they wish to support this platform. So, in that
> case it doesn't just impacting a single platform, but all platforms.

I will try movw. But it may take a few days, because I feel that the changes
will be a little big.

>