Intel: Backport 3 core PMU bugfixes to kernel 6.6 · Pull Request !10411 · openEuler/kernel

Content:
In mainline kernel before v6.11-rc1, PMU name for Intel next Xeon Granite Rapids(GNR) is the same as 4th and 5th gen Xeon(SPR and EMR). A new bugfix makes GNR using a different name for PMU.

The fix patch is among a patch set of three which just landed mainline kernel v6.11-rc1:
fa0c1c9d283b perf/x86/intel: Add a distinct name for Granite Rapids
e5f32ad56b22 perf/x86/intel/ds: Fix non 0 retire latency on Raptorlake
556a7c039a52 perf/x86/intel: Hide Topdown metrics events if the feature is not enumerated

This PR is to backport above fixes to kernel 6.6 for better GNR PMU support.

The GNR distinct name patch(fa0c1c9d283b) depends on 6 more commits from mainline kernel v6.7-rc1:
97588df87b56 perf/x86/intel: Add common intel_pmu_init_hybrid()
b0560bfd4b70 perf/x86/intel: Clean up the hybrid CPU type handling code
299a5fc8e783 perf/x86/intel: Apply the common initialization code for ADL
d87d221f854b perf/x86/intel: Factor out the initialization code for ADL e-core
0ba0c03528e9 perf/x86/intel: Factor out the initialization code for SPR
d4b5694c75d4 perf/x86/intel: Use the common uarch name for the shared functions
which introduces some code refactoring in x86 intel PMU driver.

Two more dependent following bugfixes are added for commit 97588df87b56:
(v6.7-rc3) e8df9d9f4209 perf/x86/intel: Correct incorrect 'or' operation for PMU capabilities
(v6.7-rc1) 950ecdc672ae perf/x86/intel: Fix broken fixed event constraints extension

Intel-kernel issue:
https://gitee.com/openeuler/intel-kernel/issues/IAGLFT

Test:
Before backport on GNR, "dmesg | grep -i 'performance events'" shows "Sapphire Rapids events".
After backport on GNR, "dmesg | grep -i 'performance events'" shows "Granite Rapids events".

Known issue:
N/A

Configs:
N/A

GVPopenEuler/kernel

搜索帮助

GVP openEuler/kernel