On Jun 7, 2021, at 5:37 PM, Sherry Yang sherry.yang@oracle.com wrote:
Could you also backport these two commits adefe55e7258 ("x86/kernel: Convert to new CPU match macros") 2c88d45edbb8 ("x86, sched: Treat Intel SNC topology as default, COD as exception") to 5.4.y?
Commit adefe55e7258 ("x86/kernel: Convert to new CPU match macros") is a prerequisite of the second commit. There are conflicts while cherry-picking commit adefe55e7258 ("x86/kernel: Convert to new CPU match macros"), which are caused by a later commit c84cb3735fd5 ("x86/apic: Move TSC deadline timer debug printk"). Keep the later code base.
Alison Schofield (1): x86, sched: Treat Intel SNC topology as default, COD as exception
Thomas Gleixner (1): x86/kernel: Convert to new CPU match macros
arch/x86/kernel/apic/apic.c | 32 ++++++------- arch/x86/kernel/smpboot.c | 90 +++++++++++++++++++------------------ arch/x86/kernel/tsc_msr.c | 14 +++--- arch/x86/power/cpu.c | 16 +------ 4 files changed, 68 insertions(+), 84 deletions(-)
-- 2.27.0
Hi,
We have seen that the warning “sched: CPU #20's llc-sibling CPU #0 is not on the same node! [node: 1 != 0]. Ignoring dependency. ” applies to 5.4 but we don’t observe the fix in 5.4. I'm sending this email to apply the fix from upstream 2c88d45edbb8 ("x86, sched: Treat Intel SNC topology as default, COD as exception") to 5.4 and also resolve the dependency conflict caused by prerequisite commit adefe55e7258 ("x86/kernel: Convert to new CPU match macros”) by keeping the later code base, please refer to the previous two patches for the detail.
Thanks, Sherry