Attention is currently required from: Annie Chen, Arthur Heymans, Chen, Gang C, Christian Walter, David Hendricks, Felix Held, Jincheng Li, Lean Sheng Tan, Nico Huber, Nill Ge, Patrick Rudolph, Paul Menzel, TangYiwei, Tim Chu.
2 comments:
File src/soc/intel/xeon_sp/include/soc/chip_common.h:
Patch Set #17, Line 12: void soc_create_ioat_domains(struct bus * const bus, const STACK_RES * const sr);
No space after the asterisk, please. And the added `const` is not part of the […]
Acknowledged
File src/soc/intel/xeon_sp/spr/soc_acpi.c:
Doesn't `stack_enabled` cover this? Looking at the code below it seems to me […]
stack_enabled decides whether the whole stack is enabled.
Bus comparison decides whether an accelerator is enabled in a stack.
To get together both, it seems confusing. Let me check if we could do some simplification.
As to 'we always have the decide nodes in DSDT even if they are not active'.
Yes, this is how it works. OS will parse ACPI table and check if the specific object exists, or if the resource of specific object effective, if no, OS will ignore the reported object or resource.
SPR has DSDT based on static ASL template (the template defines the max possible resource layout across all SKUs, 1S and 2S), and hence the ineffective object case doesn't only occurs in this code, but also in some defined but not existing host bridge objects. In GNR, we take fully dynamically generated AML for all SSDT/DSDT objects to further improve.
To view, visit change 78327. To unsubscribe, or for help writing mail filters, visit settings.