Hi Julien, Thank you for your observations. You're absolutely right about this.
Currently, the sci_relinquish_resources call doesn't perform any operations because the single-agent doesn't implement a callback. I'll move the sci implementation to be positioned above the tee implementation and prepare a patch for this change. -- Oleksii On 09/09/2025 12:55, Julien Grall wrote: > Hi Oleksii, > > While going through the list of recently committed patches, I noticed > some changes in the relinquish code. > > > On 04/09/2025 15:21, Oleksii Moisieiev wrote: >> @@ -1103,6 +1109,10 @@ int domain_relinquish_resources(struct domain *d) >> ret = relinquish_p2m_mapping(d); >> if ( ret ) >> return ret; > > Style: There is a missing newline. > >> + PROGRESS(sci): > > I don't quite understand why the sci relinquish was added right in the > middle of the P2M relinquish logic. At least to me, it makes more > sense to be closer to TEE (because this is firmware subsystem) and > possibly even before releasing any devices. > > Can you clarify why you chose this placement? > > Cheers, >
