Re: [PATCH] [RFC] arm64: export this_cpu_has_cap

[Suzuki K Poulose <suzuki.poulose@arm.com>]

On 01/11/2021 09:40, Will Deacon wrote:
> On Mon, Nov 01, 2021 at 09:34:08AM +0000, Suzuki K Poulose wrote:
>> On 01/11/2021 09:01, Will Deacon wrote:
>>> On Fri, Oct 29, 2021 at 07:06:23PM +0100, Catalin Marinas wrote:
>>>> On Fri, Oct 29, 2021 at 02:31:23PM +0100, Suzuki K Poulose wrote:
>>>>> On 29/10/2021 12:30, Arnd Bergmann wrote:
>>>>>> From: Arnd Bergmann <arnd@arndb.de>
>>>>>>
>>>>>> It's now used in a coresight driver that can be a loadable module:
>>>>>>
>>>>>> ERROR: modpost: "this_cpu_has_cap" [drivers/hwtracing/coresight/coresight-trbe.ko] undefined!
>>>>>>
>>>>>> Fixes: 8a1065127d95 ("coresight: trbe: Add infrastructure for Errata handling")
>>>>>
>>>>> Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
>>>>> Tested-by: Suzuki K Poulose <suzuki.poulose@arm.com>
>>>>>
>>>>> Will, Catalin, Mathieu,
>>>>>
>>>>> Do you have a preference on how this fix can be pulled in ? This may
>>>>> be safe to go via coresight tree if it is not too late. Otherwise,
>>>>> it could go via the arm64 tree.
>>>>
>>>> I think Will already closed/tagged the arm64 tree for the upcoming
>>>> merging window, though he could take it as a fix afterwards.
>>>>
>>>> If it doesn't conflict with the arm64 for-next/core, it's fine by me to
>>>> go through the coresight tree.
>>>>
>>>>>> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
>>>>>> ---
>>>>>> Not sure if we actually want this to be exported, this is my local
>>>>>> workaround for the randconfig build bot.
>>>>>> ---
>>>>>>     arch/arm64/kernel/cpufeature.c | 1 +
>>>>>>     1 file changed, 1 insertion(+)
>>>>>>
>>>>>> diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
>>>>>> index ecbdff795f5e..beccbcfa7391 100644
>>>>>> --- a/arch/arm64/kernel/cpufeature.c
>>>>>> +++ b/arch/arm64/kernel/cpufeature.c
>>>>>> @@ -2864,6 +2864,7 @@ bool this_cpu_has_cap(unsigned int n)
>>>>>>     	return false;
>>>>>>     }
>>>>>> +EXPORT_SYMBOL(this_cpu_has_cap);
>>>>
>>>> EXPORT_SYMBOL_GPL? I think this_cpu_has_cap() is a bit more more
>>>> specialised than cpus_have_const_cap().
>>>>
>>>> With that:
>>>>
>>>> Acked-by: Catalin Marinas <catalin.marinas@arm.com>
>>>
>>> Yes, at this stage I think it's best for this to go via the Coresight tree.
>>> So with the _GPL export:
>>>
>>> Acked-by: Will Deacon <will@kernel.org>
>>>
>>> If that doesn't work for some reason, I can take it next week after the
>>> initial arm64 queue has been merged. Please just let me know.
>>
>> As I understand correctly, this will now need to go via arm64 tree. The
>> CoreSight tree changes are pulled into Greg's tree and the next it will
>> happen is for the next release. Usually the fixes don't end up there
>> during the -rc cycles. So, I believe it is better if this goes via
>> arm64.
> 
> Hmm, are you saying that Coresight drivers don't receive fixes outside of
> the merge window? That sounds sub-optimal...

Unfortunately thats how it works today. We should fix this.

Mathieu, Greg,

Do you have any thoughts on how to address this ?

> 
> But in any case, I'm happy to take this as long as it can wait until the
> second half of the merge window.

Yes, please. Thats the quickest path to merging this patch.

Suzuki