Re: [PATCH] locking/atomic: sh: Use generic_cmpxchg_local for arch_cmpxchg_local()

From: "wuqiang.matt" <wuqiang.matt@bytedance.com>
To: "Masami Hiramatsu (Google)" <mhiramat@kernel.org>,
	Mark Rutland <mark.rutland@arm.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>,
	Rich Felker <dalias@libc.org>,
	John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>,
	Peter Zijlstra <peterz@infradead.org>,
	linux-sh@vger.kernel.org, linux-kernel@vger.kernel.org,
	linux-trace-kernel@vger.kernel.org
Subject: Re: [PATCH] locking/atomic: sh: Use generic_cmpxchg_local for arch_cmpxchg_local()
Date: Wed, 25 Oct 2023 19:06:38 +0800	[thread overview]
Message-ID: <0f48e36b-b8c1-443a-f7c3-b9ddea2c503c@bytedance.com> (raw)
In-Reply-To: <efe7e4ad-6ee8-08c3-d43f-d7c54de04045@bytedance.com>

On 2023/10/25 09:51, wuqiang.matt wrote:
> On 2023/10/25 07:42, Masami Hiramatsu (Google) wrote:
>> On Tue, 24 Oct 2023 16:08:12 +0100
>> Mark Rutland <mark.rutland@arm.com> wrote:
>>
>>> On Tue, Oct 24, 2023 at 11:52:54PM +0900, Masami Hiramatsu (Google) wrote:
>>>> From: Masami Hiramatsu (Google) <mhiramat@kernel.org>
>>>>
>>>> Use generic_cmpxchg_local() for arch_cmpxchg_local() implementation
>>>> in SH architecture because it does not implement arch_cmpxchg_local().
>>>
>>> I do not think this is correct.
>>>
>>> The implementation in <asm-generic/cmpxchg-local.h> is UP-only (and it only
>>> disables interrupts), whereas arch/sh can be built SMP. We should probably add
>>> some guards into <asm-generic/cmpxchg-local.h> for that as we have in
>>> <asm-generic/cmpxchg.h>.
>>
>> Isn't cmpxchg_local for the data which only needs to ensure to do cmpxchg
>> on local CPU?
> 
> asm-generic/cmpxchg.h is only for UP, will throw an error for SMP building:
> 
> #ifdef CONFIG_SMP
> #error "Cannot use generic cmpxchg on SMP"
> #endif

Sorry that I just noticed Masami's patch has asm-generic/cmpxchg-local.h
included, not asm-generic/cmpxchg.h. cmpxchg.h does throw an error for SMP
configs, but cmpxchg-local.h doesn't.

> SH arch seems it does have SMP systems. The arch/sh/include/asm/cmpxchg.h has
> the following codes:
> 
> #if defined(CONFIG_GUSA_RB)
> #include <asm/cmpxchg-grb.h>
> #elif defined(CONFIG_CPU_SH4A)
> #include <asm/cmpxchg-llsc.h>
> #elif defined(CONFIG_CPU_J2) && defined(CONFIG_SMP)
> #include <asm/cmpxchg-cas.h>
> #else
> #include <asm/cmpxchg-irq.h>
> #endif
> 
>> So I think it doesn't care about the other CPUs (IOW, it should not touched by
>> other CPUs), so it only considers UP case. E.g. on x86, arch_cmpxchg_local() is
>> defined as raw "cmpxchg" without lock prefix.
>>
>> #define __cmpxchg_local(ptr, old, new, size)                            \
>>          __raw_cmpxchg((ptr), (old), (new), (size), "")
>>
>>
>> Thank you,
>>
>>
>>>
>>> I think the right thing to do here is to define arch_cmpxchg_local() in terms
>>> of arch_cmpxchg(), i.e. at the bottom of arch/sh's <asm/cmpxchg.h> add:
>>>
>>> #define arch_cmpxchg_local              arch_cmpxchg
> 
> I agree too. Might not be performance optimized but guarantees correctness.
> 
>>> Mark.
>>>
>>>>
>>>> Reported-by: kernel test robot <lkp@intel.com>
>>>> Closes: 
>>>> https://lore.kernel.org/oe-kbuild-all/202310241310.Ir5uukOG-lkp@intel.com/
>>>> Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
>>>> ---
>>>>   arch/sh/include/asm/cmpxchg.h |    2 ++
>>>>   1 file changed, 2 insertions(+)
>>>>
>>>> diff --git a/arch/sh/include/asm/cmpxchg.h b/arch/sh/include/asm/cmpxchg.h
>>>> index 288f6f38d98f..e920e61fb817 100644
>>>> --- a/arch/sh/include/asm/cmpxchg.h
>>>> +++ b/arch/sh/include/asm/cmpxchg.h
>>>> @@ -71,4 +71,6 @@ static inline unsigned long __cmpxchg(volatile void * 
>>>> ptr, unsigned long old,
>>>>                       (unsigned long)_n_, sizeof(*(ptr))); \
>>>>     })
>>>> +#include <asm-generic/cmpxchg-local.h>
>>>> +
>>>>   #endif /* __ASM_SH_CMPXCHG_H */
>>>>
>>
>>
> 