On 11.10.2022 18:52, Demi Marie Obenour wrote: > On Tue, Oct 11, 2022 at 11:59:01AM +0200, Jan Beulich wrote: >> On 11.10.2022 05:42, Demi Marie Obenour wrote: >>> A previous patch tried to get Linux to use the ESRT under Xen if it is >>> in memory of type EfiRuntimeServicesData. However, this turns out to be >>> a bad idea. Ard Biesheuvel pointed out that EfiRuntimeServices* memory >>> winds up fragmenting both the EFI page tables and the direct map, >> >> Can this statement please be made describe Xen, not Linux? Aiui at least >> the directmap aspect doesn't apply to Xen. > > Should it apply to Xen? My understanding is that Ard’s statements > regarding mismatched attributes refer to any kernel, not just Linux. > You would be in a better position to judge that, though.
We run EFI runtime services functions on their own page tables (with certain areas copied from the directmap). With EfiACPIReclaimMemory converted to E820_ACPI we do not insert those ranges into the directmap (i.e. no difference to EfiRuntimeServices*). At least this latter fact means fragmentation effects - if they exist - are the same for both types. >>> and >>> that EfiACPIReclaimMemory is a much better choice for this purpose. >> >> I think the "better" wants explaining here, without requiring people to >> follow ... > > Something like, “EfiACPIReclaimMemory is the correct type for > configuration tables that are only used by the OS.”? Preferably with "supposedly" inserted, unless you (or Ard) can point out a place in the spec where this is actually written down. Jan
