stress-ng --hrtimers 1 as root causes system stall on arm64

[dsx724]

This does not seem to affect armhf.

[dsx724]

On Allwinner H5:

[954] dispatching hogs: 1 hrtimers [ 332.053991] rcu: INFO: rcu_preempt detected stalls on CPUs/tasks: [ 332.066385] rcu: Tasks blocked on level-0 rcu_node (CPUs 0-3): P955/2:b..l [ 332.076804] rcu: rcu_preempt kthread starved for 5257 jiffies! g10713 f0x0 RCU_GP_WAIT_FQS(5) ->state=0x0 ->cpu=2 [ 332.090264] rcu: Unless rcu_preempt kthread gets sufficient CPU time, OOM is now expected behavior. [ 332.102635] rcu: RCU grace-period kthread stack dump: [ 332.111015] rcu: Stack dump where RCU GP kthread last ran:

[dsx724]

On Amlogic S905X

sudo stress-ng --hrtimers 1
stress-ng: info: [1911] defaulting to a 86400 second (1 day, 0.00 secs) run per stressor stress-ng: info: [1911] dispatching hogs: 1 hrtimers [ 289.327328] rcu: INFO: rcu_preempt detected stalls on CPUs/tasks: [ 289.327793] rcu: Tasks blocked on level-0 rcu_node (CPUs 0-3): P1846/1:b..l [ 352.347167] rcu: INFO: rcu_preempt detected stalls on CPUs/tasks: [ 352.347632] rcu: Tasks blocked on level-0 rcu_node (CPUs 0-3): P1846/1:b..l

Improves somewhat by setting scaling governor to performance.

[dsx724]

On Rockchip RK3328

root@ubuntu:/home/ubuntu# stress-ng --hrtimers 1 stress-ng: info: [1947] defaulting to a 86400 second (1 day, 0.00 secs) run per stressor stress-ng: info: [1947] dispatching hogs: 1 hrtimers