Re: [PATCH -fixes] riscv: cpufeature: Do not drop Linux-internal extensions

[Clément Léger]

On 18/07/2024 23:57, Andrew Jones wrote:
> On Thu, Jul 18, 2024 at 02:29:59PM GMT, Samuel Holland wrote:
>> The Linux-internal Xlinuxenvcfg ISA extension is omitted from the
>> riscv_isa_ext array because it has no DT binding and should not appear
>> in /proc/cpuinfo. The logic added in commit 625034abd52a ("riscv: add
>> ISA extensions validation callback") assumes all extensions are included
>> in riscv_isa_ext, and so riscv_resolve_isa() wrongly drops Xlinuxenvcfg
>> from the final ISA string. Instead, accept such Linux-internal ISA
>> extensions as if they have no validation callback.
> 
> This assumes we'll never need a validation callback for a Linux-internal
> ISA extension. We can make that assumption now and change our mind
> later, but we could also add Xlinuxenvcfg to riscv_isa_ext now and
> modify the places where it matters (just print_isa?). If we add
> Xlinuxenvcfg to the array with a NULL name then we could do something
> like
> 
>  print_isa()
>  {
>     for (...) {
>         ...
>         if (!riscv_isa_ext[i].name)
> 	   continue;
>     }
>  }

I would rather add it to the riscv_isa_ext[] array and avoid handling it
differently. There is already the xandespmu extension in this array so
xlinuxenvcfg can be added as well.

Thanks,

Clément

> 
>>
>> Fixes: 625034abd52a ("riscv: add ISA extensions validation callback")
>> Signed-off-by: Samuel Holland <samuel.holland@xxxxxxxxxx>
>> ---
>>
>>  arch/riscv/kernel/cpufeature.c | 14 ++++++--------
>>  1 file changed, 6 insertions(+), 8 deletions(-)
>>
>> diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
>> index 0366dc3baf33..dd25677d60de 100644
>> --- a/arch/riscv/kernel/cpufeature.c
>> +++ b/arch/riscv/kernel/cpufeature.c
>> @@ -457,28 +457,26 @@ static void __init riscv_resolve_isa(unsigned long *source_isa,
>>  		bitmap_copy(prev_resolved_isa, resolved_isa, RISCV_ISA_EXT_MAX);
>>  		for_each_set_bit(bit, source_isa, RISCV_ISA_EXT_MAX) {
>>  			ext = riscv_get_isa_ext_data(bit);
>> -			if (!ext)
>> -				continue;
>>  
>> -			if (ext->validate) {
>> +			if (ext && ext->validate) {
>>  				ret = ext->validate(ext, resolved_isa);
>>  				if (ret == -EPROBE_DEFER) {
>>  					loop = true;
>>  					continue;
>>  				} else if (ret) {
>>  					/* Disable the extension entirely */
>> -					clear_bit(ext->id, source_isa);
>> +					clear_bit(bit, source_isa);
>>  					continue;
>>  				}
>>  			}
>>  
>> -			set_bit(ext->id, resolved_isa);
>> +			set_bit(bit, resolved_isa);
>>  			/* No need to keep it in source isa now that it is enabled */
>> -			clear_bit(ext->id, source_isa);
>> +			clear_bit(bit, source_isa);
>>  
>>  			/* Single letter extensions get set in hwcap */
>> -			if (ext->id < RISCV_ISA_EXT_BASE)
>> -				*this_hwcap |= isa2hwcap[ext->id];
>> +			if (bit < RISCV_ISA_EXT_BASE)
>> +				*this_hwcap |= isa2hwcap[bit];
>>  		}
>>  	} while (loop && memcmp(prev_resolved_isa, resolved_isa, sizeof(prev_resolved_isa)));
>>  }
>> -- 
>> 2.45.1
>>
> 
> If we'd rather leave Xlinuxenvcfg out of the array (and generally support
> extensions not in the array), then LGTM
> 
> Reviewed-by: Andrew Jones <ajones@xxxxxxxxxxxxxxxx>
> 
> Thanks,
> drew