> > Two physical packages are on a multi-chip module and share resources on > the module. The module then plugs into the baseboard/motherboard. > > Is it possible to elaborate on the resource being shared ?
In our specific case the problem is related to the PPTT's "Identical Implementation" flag. We need a top level node above the physical package nodes to be able to set the "Identical Implementation" flag to indicate that all of the procs in all of the child packages are the same identical implementation. Without this (ie. forcing each physical package to be its own root node) Linux will fail to load the SPE driver when there are multiple identical packages because it detects that some of the procs have a different root node than other procs, implying that the packages don't have identical implementations. > Does it fall into the subject of this thread ? Some resources might be aswell > described in other ACPI tables. The thread you linked looks like it is about non-processor resources, and therefore wasn't in scope for PPTT. I don't think this is related. > > Note: While investigating this we noticed that another vendor also has a > similar PPTT topology to what is being flagged as invalid, so either that > vendor > isn't using EDK2 or they have done something to avoid this check without > submitting a patch to EDK2. > > This check is only present in the DynamicTablesPkg, so it shouldn't be too > restrictive. Ah, correct. If they're not using DynamicTablesPkg to generate the PPTT/SSDT, then they wouldn't hit this. > If the platform is known to use it, is it possible to share which platform it > is ? We see the topology in question (ie. a root node is a parent node to physical package nodes) in Ampere's two-socket Altra PPTT. -=-=-=-=-=-=-=-=-=-=-=- Groups.io Links: You receive all messages sent to this group. View/Reply Online (#113485): https://edk2.groups.io/g/devel/message/113485 Mute This Topic: https://groups.io/mt/103603398/21656 Group Owner: devel+ow...@edk2.groups.io Unsubscribe: https://edk2.groups.io/g/devel/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
