Re: [edk2-devel] [PATCH 1/2] ArmPkg: implement EFI_MP_SERVICES_PROTOCOL based on PSCI calls

["Rebecca Cran" <quic_rcran@quicinc.com>]

On 11/28/22 15:59, Kun Qin wrote:
> Hi Rebecca/Ard,
> 
> I was trying to reach out regarding the original patches earlier (see 
> below) but it might fell off your stacks
> due to high traffic on the mailing list. Could you please kindly review 
> the questions when you have a chance?
> 
> In addition, I found another edge case of the MP service: when the AP 
> routine hits a timeout, the metadata
> will be left in the unfinished states. If the AP routine eventually 
> completes and return, this AP will stay in "finished"
> but never become "ready" during this boot. I tried to add below change, 
> which seems to work. But please let me
> know if you have other concerns:
> https://github.com/kuqin12/mu_silicon_arm_tiano/commit/c76072b37018276f2fec2582d0c540be5b40d0f2

Thanks, I'll take a look and integrate the fix into the next revision of 
the patch series.

> 
> Lastly, do you plan to merge these patches in the near future? This will 
> be a great add-on for ARM platforms.

The issue that's currently preventing them from being merged is a 
failure I noticed on the Neoverse N2 FVP: for some reason despite 
enabling the MMU and caches, manual cache flushes are still required for 
the data to be seen between CPUs. I don't know if that's a bug in the 
code or in the FVP model.

-- 
Rebecca Cran