** Description changed:
[Impact]
The numa mask subset check can often lead to system hang or crash during CPU
hotplug and system suspend operation if NUMA is disabled. This is mostly
observed on HMP systems where the CPU compute capacities are different and ends
up in different scheduler domains. Since cpumask_of_node is returned instead
core_sibling, the scheduler is confused with incorrect cpumasks(e.g. one CPU in
two different sched domains at the same time) on CPU hotplug.
[Fix]
The following patch fixes this issue:
e156ab71a974 arm64: topology: Avoid checking numa mask for scheduler MC
selection
[Test]
+ -- testcase --
+
+ #!/bin/bash
+
+ stress-ng -q --vm 100 --vm-bytes 1G --cpu 210 --timeout 2400s --metrics-brief
&
+ pid=$!
+
+ for i in {1..210}; do
+ num=$((RANDOM%10+$i))
+ echo "Offline CPU $num"
+ echo 0 > /sys/devices/system/cpu/cpu$num/online || exit 1
+ sleep 3s
+ echo "Online CPU $num"
+ echo 1 > /sys/devices/system/cpu/cpu$num/online || exit 1
+ done
+
+ kill -9 $pid
+ ----------------
+ Executed the test case with stock Bionic kernel and with Bionic kernel with
patches applied. No system hang or crash was observed. On Cavium CN99XX Thunder
X2 system.
[Regression Potential]
+ Regression tested on Cavium Thunder X2 CN99XX system, and none were found.
Regression risk is low.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1797202
Title:
[Bionic] arm64: topology: Avoid checking numa mask for scheduler MC
selection
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1797202/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
