[External] Re: [PATCH v4 0/5] Add workarounds for Picoheart erratas

[Yicong Yang]

On 10/12/26 3:10 PM, yunhui cui wrote:
> Hi Yicong,
> 
> On Mon, Oct 12, 2026 at 2:42 PM Yicong Yang <yang.yicong at picoheart.com> wrote:
>>
>> a gentle ping for this patchset...
>>
>> thanks.
> 
>   Thanks for the series. Two issues found in v4:
> 
>   1. The CBO erratum uses MMU-only uaccess helpers, so the NOMMU build
>      currently fails. Could ERRATA_PICOHEART_CBO_CLEAN depend on MMU?
>      Using unsigned long for addr would also avoid the RV32 warning.

thanks for catching. will make it depend on MMU and use unsigned long for addr.

> 
>   2. SR_MXR remains enabled across get_user(), whose might_fault() may
>      schedule. Would it be safer to prepare for the fault first, then use
>      __get_user() with MXR enabled and restore the original MXR state?
> 

the MXR is saved/restored during context switch in __switch_to(), along with
the SR_SUM. so the scheduling won't be a problem. do I miss somewhere?

thanks.