Back to bug 2176010

Who When What Removed Added
Red Hat Bugzilla 2023-03-07 07:03:50 UTC Pool ID sst_virtualization_hwe_rhel_9
Red Hat One Jira (issues.redhat.com) 2023-03-07 07:06:29 UTC Link ID Red Hat Issue Tracker RHELPLAN-150884
Mario Casquero 2023-03-07 07:07:04 UTC Assignee virt-maint imammedo
jingzhao 2023-03-07 07:10:14 UTC QA Contact virt-bugs chayang
Mario Casquero 2023-03-07 07:14:22 UTC CC zhenyzha
Chao Yang 2023-03-07 09:16:31 UTC QA Contact chayang mcasquer
Nitesh Narayan Lal 2023-03-07 14:41:13 UTC Keywords Triaged
Mario Casquero 2023-03-10 10:44:59 UTC Flags needinfo?(imammedo)
CC imammedo
Igor Mammedov 2023-03-10 12:05:48 UTC Flags needinfo?(imammedo)
Igor Mammedov 2023-03-21 14:49:22 UTC CC babu.moger, bandan.das
Chao Yang 2023-03-23 02:26:47 UTC Flags needinfo?(mcasquer)
Mario Casquero 2023-03-24 07:34:04 UTC Flags needinfo?(mcasquer) needinfo?(imammedo)
Igor Mammedov 2023-03-27 12:39:24 UTC Flags needinfo?(imammedo)
Babu Moger 2023-03-29 19:27:05 UTC CC pkodamat, wei.huang2
Bandan Das 2023-04-14 15:20:52 UTC CC bandan.das bdas
Chao Yang 2023-04-17 05:41:26 UTC Flags needinfo?(mcasquer)
Mario Casquero 2023-05-03 07:30:30 UTC Flags needinfo?(mcasquer)
Mario Casquero 2023-05-08 07:15:47 UTC Flags needinfo?(imammedo)
Igor Mammedov 2023-05-17 15:18:03 UTC CC yduan
Igor Mammedov 2023-05-23 12:28:21 UTC Flags needinfo?(imammedo) needinfo?(babu.moger)
Resolution --- WONTFIX
Doc Type If docs needed, set a value Known Issue
Status NEW CLOSED
Doc Text Cause:

EPYC CPU model encodes NUMA nodes mapping within CPU's APIC ID.
However it isn't handled in QEMU correctly and previous attempts to fix it has lead to overly complicated solution that was hard to maintain, and eventually effort to make it right was abandoned.

Consequence:

Guest kernel will print warnings during boot:

sched: CPU #4's llc-sibling CPU #3 is not on the same node! [node: 1 != 0]. Ignoring dependency.
WARNING: CPU: 4 PID: 0 at arch/x86/kernel/smpboot.c:415 topology_sane.isra.0+0x6b/0x80

Workaround (if any):

use another CPU model then EPYC

Result:
Mis-mapping CPUs to NUMA nodes might affect performance.
Last Closed 2023-05-23 12:28:21 UTC
Yanhui Ma 2023-05-25 09:38:00 UTC CC jherrman, yama
Flags needinfo?(jherrman)
Babu Moger 2023-05-31 14:19:36 UTC Flags needinfo?(babu.moger)
Igor Mammedov 2023-05-31 15:34:15 UTC Flags needinfo?(babu.moger)
Babu Moger 2023-05-31 15:47:39 UTC Flags needinfo?(babu.moger)
Jiri Herrmann 2023-06-05 09:21:16 UTC Flags needinfo?(jherrman) needinfo?(imammedo) needinfo?(yama)
Doc Text Cause:

EPYC CPU model encodes NUMA nodes mapping within CPU's APIC ID.
However it isn't handled in QEMU correctly and previous attempts to fix it has lead to overly complicated solution that was hard to maintain, and eventually effort to make it right was abandoned.

Consequence:

Guest kernel will print warnings during boot:

sched: CPU #4's llc-sibling CPU #3 is not on the same node! [node: 1 != 0]. Ignoring dependency.
WARNING: CPU: 4 PID: 0 at arch/x86/kernel/smpboot.c:415 topology_sane.isra.0+0x6b/0x80

Workaround (if any):

use another CPU model then EPYC

Result:
Mis-mapping CPUs to NUMA nodes might affect performance.
.NUMA node mapping not working correctly on AMD EPYC CPUs

QEMU does not handle NUMA node mapping on AMD EPYC CPUs correctly. As a result, the performance of virtual machines (VMs) with these CPUs might be negatively impacted if using a NUMA node configuration. In addition, the VMs display a warning similar to the following during boot.

----
sched: CPU #4's llc-sibling CPU #3 is not on the same node! [node: 1 != 0]. Ignoring dependency.
WARNING: CPU: 4 PID: 0 at arch/x86/kernel/smpboot.c:415 topology_sane.isra.0+0x6b/0x80
----

To work around this issue, do not use AMD EPYC CPUs for NUMA node configurations.
Igor Mammedov 2023-06-05 09:53:14 UTC Flags needinfo?(imammedo)
Yanhui Ma 2023-06-06 11:30:15 UTC Flags needinfo?(yama)
Lucie Vařáková 2023-07-12 12:23:30 UTC Docs Contact jherrman

Back to bug 2176010