Re: [PATCH v3 18/26] arm64: Introduce asm/vdso/processor.h

[Vincenzo Frascino]

On 3/16/20 10:22 AM, Mark Rutland wrote:
> Hi Vincenzo,
> 
> On Mon, Mar 16, 2020 at 09:42:32AM +0000, Vincenzo Frascino wrote:
>> On 3/15/20 6:30 PM, Catalin Marinas wrote:
>>> On Fri, Mar 13, 2020 at 03:43:37PM +0000, Vincenzo Frascino wrote:
>>>> --- /dev/null
>>>> +++ b/arch/arm64/include/asm/vdso/processor.h
>>>> @@ -0,0 +1,31 @@
>>>> +/* SPDX-License-Identifier: GPL-2.0-only */
>>>> +/*
>>>> + * Copyright (C) 2020 ARM Ltd.
>>>> + */
>>>> +#ifndef __ASM_VDSO_PROCESSOR_H
>>>> +#define __ASM_VDSO_PROCESSOR_H
>>>> +
>>>> +#ifndef __ASSEMBLY__
>>>> +
>>>> +#include <asm/page-def.h>
>>>> +
>>>> +#ifdef CONFIG_COMPAT
>>>> +#if defined(CONFIG_ARM64_64K_PAGES) && defined(CONFIG_KUSER_HELPERS)
>>>> +/*
>>>> + * With CONFIG_ARM64_64K_PAGES enabled, the last page is occupied
>>>> + * by the compat vectors page.
>>>> + */
>>>> +#define TASK_SIZE_32		UL(0x100000000)
>>>> +#else
>>>> +#define TASK_SIZE_32		(UL(0x100000000) - PAGE_SIZE)
>>>> +#endif /* CONFIG_ARM64_64K_PAGES */
>>>> +#endif /* CONFIG_COMPAT */
>>>
>>> Just curious, what's TASK_SIZE_32 used for in the vDSO code? You don't
>>> seem to move TASK_SIZE.
>>>
>>
>> I tried to fine grain the headers as much as I could in order to avoid
>> unneeded/unwanted inclusions:
>>  * TASK_SIZE_32 is used to verify ABI consistency on vdso32 (please refer to
>>    arch/arm64/kernel/vdso32/vgettimeofday.c).
>>  * TASK_SIZE is not required by the vdso library hence it is not present in
>>    these headers.
> 
> It would be worth noting the former point in the commit message, since
> it can be surprising.
>

Sure it is a good point, I will add this to the commit message.

> I also think it's worth keeping the definitions together if that's easy,
> as it makes it easier to navigate the codebase, even if TASK_SIZE isn't
> necessary for the VDSO itself.

This can't be done because TASK_SIZE on arm64 requires test_thread_flag() with
is not suited for vDSO. In other words can cause the same problem we are trying
to solve.

> 
> Thanks,
> Mark.
> 

-- 
Regards,
Vincenzo