RHEL Engineering is moving the tracking of its product development work on RHEL 6 through RHEL 9 to Red Hat Jira (issues.redhat.com). If you're a Red Hat customer, please continue to file support cases via the Red Hat customer portal. If you're not, please head to the "RHEL project" in Red Hat Jira and file new tickets here. Individual Bugzilla bugs in the statuses "NEW", "ASSIGNED", and "POST" are being migrated throughout September 2023. Bugs of Red Hat partners with an assigned Engineering Partner Manager (EPM) are migrated in late September as per pre-agreed dates. Bugs against components "kernel", "kernel-rt", and "kpatch" are only migrated if still in "NEW" or "ASSIGNED". If you cannot log in to RH Jira, please consult article #7032570. That failing, please send an e-mail to the RH Jira admins at rh-issues@redhat.com to troubleshoot your issue as a user management inquiry. The email creates a ServiceNow ticket with Red Hat. Individual Bugzilla bugs that are migrated will be moved to status "CLOSED", resolution "MIGRATED", and set with "MigratedToJIRA" in "Keywords". The link to the successor Jira issue will be found under "Links", have a little "two-footprint" icon next to it, and direct you to the "RHEL project" in Red Hat Jira (issue links are of type "https://issues.redhat.com/browse/RHEL-XXXX", where "X" is a digit). This same link will be available in a blue banner at the top of the page informing you that that bug has been migrated.

Description of problem:
Timer interrupt handling uses up to 20 usec

Version-Release number of selected component (if applicable):
3.10.0-327.rt56.204.el7.x86_64

How reproducible:


Steps to Reproduce:
1. run cpu bound process on isolcpu
2. trace for task switch
3. record time between task switches, where ksoftirqd is switched in/out

Actual results:
We see that cpu bound process is preempted for ksoftirqd, which does timer interrupt work, the switches back to cpu bound process.  This work can take up to 20 usec

Expected results:
Timer interrupt work should not switch thread context and process timer interrupt work much faster (perhaps 2-3 usec, not sure of exact amount)

Additional info:
This is currently observed with a DPDK program running at fifo:95

What exactly is the bound task doing? ksoftirqd does timer work, and a timer interrupt can run. And things like task usage accounting can take 20 us.

Also, have you added nohz_full and rcu_nocb to that isolated CPU as well? That could help too.

We have no_full, but not rcu_nocb yet, so we will add that.  I don't believe the user thread is doing anything other than some futex calls once in a while.  We'll trace to see if anything else is going on with the user thread.

Andrew, were you able to do a trace to see if anything else was going on?

Thanks,
Beth

Karl, would it be possible to do another trace to see what is happening during the timer interrupt?

Andrew/Karl, any updates?

Andrew/Karl, any updates? If we can't get any additional updates, we will need to close this BZ.

We have not done a trace, and I am not sure when we will have opportunity to do so.  We can close this and repoen when we have a trace.

Thank you, Andrew. As you suggest, we will close this issue. Please reopen the BZ if you do get a trace and we will take a look at it.