We should reconfigure the kernel as follows: * CONFIG_NO_HZ_FULL=y * CONFIG_HZ=1000 * CONFIG_PREEMPT=y This will make IPFire systems more responsive to any network events without any significant downsides. There will be some downsize in "throughput" which refers to memory bandwidth and L1 cache invalidation which won't be a problem for our application at all. There is more about this here: https://www.kernel.org/doc/Documentation/timers/NO_HZ.txt
this is superseeded. we have changed it to CONFIG_HZ=100 and enabled PREEMPT__VOLUNTARY
(In reply to Arne.F from comment #1) > this is superseeded. we have changed it to CONFIG_HZ=100 and enabled > PREEMPT__VOLUNTARY Yes, but this isn't even in testing and definitely not shipped.
What is the status of this bug?
PREEMPT_VOLUNTARY has been shipped, but 1000 HZ hasn't been fully tested, yet: > https://git.ipfire.org/?p=people/ms/ipfire-2.x.git;a=commitdiff;h=ec3f35777c33d51dc49368a08f36e9d8868a8d9c
https://patchwork.ipfire.org/project/ipfire/list/?series=2331 This will be part of upcoming Core Update 161, hence I set this to MODIFIED.
https://blog.ipfire.org/post/ipfire-2-27-core-update-161-available-for-testing
https://blog.ipfire.org/post/ipfire-2-27-core-update-161-released