Messages in this thread | | | Date | Tue, 24 May 2022 11:59:53 +0100 | From | Sudeep Holla <> | Subject | Re: [PATCH] arch_topology: Limit threads to one specific cluster |
| |
On Tue, May 24, 2022 at 05:05:22PM +0800, Gavin Shan wrote: > Hi Sudeep, > > On 5/24/22 4:51 PM, Sudeep Holla wrote: > > On Tue, May 24, 2022 at 04:12:12PM +0800, Gavin Shan wrote: > > > The sibling information for one particular CPU is updated after ACPI > > > PPTT table is parsed. struct cpu_topology::thread_sibling tracks the > > > the CPUs in same core. However, cluster isn't considered when it's > > > populated. In this case, multiple threads belonging to different > > > clusters can be put together through the sibling information. It > > > eventually leads to unexpected warning from sched subsystem. > > > > > > For example, the following warning is observed in a VM where we have > > > 2 sockets, 4 clusters, 8 cores and 16 threads and the CPU topology > > > is populated as below. > > > > > > CPU Socket-ID Cluster-ID Core-ID Thread-ID > > > ---------------------------------------------- > > > 0 0 0 0 0 > > > 1 0 0 0 1 > > > 2 0 0 1 0 > > > 3 0 0 1 1 > > > 4 0 1 0 0 > > > 5 0 1 0 1 > > > 6 0 1 1 0 > > > 7 0 1 1 1 > > > 8 1 0 0 0 > > > 9 1 0 0 1 > > > 10 1 0 1 0 > > > 11 1 0 1 1 > > > 12 1 1 0 0 > > > 13 1 1 0 1 > > > 14 1 1 1 0 > > > 15 1 1 1 1 > > > > > > [ 0.592181] CPU: All CPU(s) started at EL1 > > > [ 0.593766] alternatives: patching kernel code > > > [ 0.595890] BUG: arch topology borken > > > [ 0.597210] the SMT domain not a subset of the CLS domain > > > [ 0.599286] child=0-1,4-5 sd=0-3 > > > > > > # cat /sys/devices/system/cpu/cpu0/topology/cluster_cpus_list > > > 0-3 > > > # cat /sys/devices/system/cpu/cpu0/topology/thread_siblings_list > > > 0-1,4-5 > > > > > > This fixes the issue by limiting threads to one specific cluster. > > > With this applied, the unexpected warning disappears in the VM. > > > > > > > I have similar fix but as part of bigger series[1] to get DT support in > > line with ACPI. > > > > Your patch resolves the issue I have. So please ignore mine. Sorry
Thanks for that.
> for the noise. >
No worries, definitely not noise.
-- Regards, Sudeep
| |