EDAC driver for ARMv8 L1/L2 cache

York Sun york.sun at nxp.com
Fri Jan 12 08:48:05 PST 2018


On 01/12/2018 08:35 AM, Thor Thayer wrote:
> On 01/09/2018 03:51 PM, York Sun wrote:
>> On 01/09/2018 01:43 PM, Borislav Petkov wrote:
>>> Adding some more people to CC.
>>>
>>> On Tue, Jan 09, 2018 at 08:48:43PM +0000, York Sun wrote:
>>>> Borislav,
>>>>
>>>> Are you aware of any existing (or in development) EDAC driver for ARMv8
>>>> L1/L2 cache? I am thinking to write one if not available yet.
>>>
>>> no I'm not but I see two EDAC drivers for ARM64: thunderx and xgene.
>>>
>>> Please synchronize with their authors and ARM people what would be the
>>> best thing to do and try extracting shared functionality from them
>>> into a common compilation unit instead of duplicating it. I don't want
>>> separate drivers per functional unit.
>>>
>>
>> Thanks for the pointer. Thunderx and xgene's drivers have different
>> implementation on the hardware. I found one patch closer to what I
>> expect, https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpatchwork.kernel.org%2Fpatch%2F7513231%2F&data=02%7C01%7Cyork.sun%40nxp.com%7C29cd81c44431418e2a6308d559da7d3b%7C686ea1d3bc2b4c6fa92cd99c5c301635%7C0%7C0%7C636513717317947616&sdata=flIZL8mxGc4VWbTpl5nPjkRXaB8uNjcc382HcXO08qE%3D&reserved=0. I don't see
>> activities after 2015. I will reach out to the author.
>>
>> York
>>
> 
> Hi York,
> 
> I'll be adding EDAC support for our Stratix10 (Quad ARM Cortex A-53) at 
> some point but I'm not sure when it is scheduled - probably in the next 
> few months. I'll be going through the same process of comparing our 
> architecture to others. Below are links to the Stratix10.
>

I see Stratix10 has A53 core. I am concerned on reading the
CPUMERRSR_EL1 and L2MERRSR_EL1. The are IMPLEMENTATION DEFINED
registers. They may not be available on all SoCs, or all time. The
CPUMERRSR_EL1 needs to be read on each core, the L2MERRSR_EL1 needs to
be read on each cluster. I am not sure how to handle them in EDAC driver
yet. I am hoping Mark Rutland can shed some light as he commented on
similar patches [1][2] before.

York

[1] https://patchwork.kernel.org/patch/7460391/
[2] https://patchwork.kernel.org/patch/7513231/



More information about the linux-arm-kernel mailing list