Bug 236423
Summary: | [RHEL5 RT] irq 16: nobody cared (try booting with the "irqpoll" option) | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Product: | Red Hat Enterprise MRG | Reporter: | Jeff Burke <jburke> | ||||||||
Component: | realtime-kernel | Assignee: | Michal Schmidt <mschmidt> | ||||||||
Status: | CLOSED DUPLICATE | QA Contact: | |||||||||
Severity: | medium | Docs Contact: | |||||||||
Priority: | medium | ||||||||||
Version: | 1.0 | CC: | mschmidt | ||||||||
Target Milestone: | --- | ||||||||||
Target Release: | --- | ||||||||||
Hardware: | All | ||||||||||
OS: | Linux | ||||||||||
URL: | http://rhts.lab.boston.redhat.com/cgi-bin/rhts/test_log.cgi?id=1244985 | ||||||||||
Whiteboard: | |||||||||||
Fixed In Version: | Doc Type: | Bug Fix | |||||||||
Doc Text: | Story Points: | --- | |||||||||
Clone Of: | Environment: | ||||||||||
Last Closed: | 2007-04-28 12:50:14 UTC | Type: | --- | ||||||||
Regression: | --- | Mount Type: | --- | ||||||||
Documentation: | --- | CRM: | |||||||||
Verified Versions: | Category: | --- | |||||||||
oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |||||||||
Cloudforms Team: | --- | Target Upstream Version: | |||||||||
Embargoed: | |||||||||||
Attachments: |
|
Description
Jeff Burke
2007-04-13 20:03:20 UTC
Created attachment 152581 [details]
dmesg output
Created attachment 152582 [details]
/proc/interrupts
Created attachment 152583 [details]
lspci -vxxx output
Jeff, I have access to a DL360, but it's an older model than yours (2xPentium III CPUs, no PCI Express,...) and it doesn't have this issue. Can I get access to dl360-01.rhts.boston.redhat.com so I can experiment with it? Does booting with acpi=noirq make the issue go away? Bug 236150 is another "irq:nobody cared" bug and it is mentioned there as a workaround. I have just become aware of the RHTS reservation system. I'll experiment with the machine myself. On dl360-01.rhts.boston.redhat.com only usb_uhci should generate IRQ 16. What we're seeing here is interrupts from IRQ 20 (cciss driver) duplicated on IRQ 16 for unknown reason. One does not need LTP to reproduce it. Any disk activity suffices (eg. cat /dev/cciss/c0d0 > /dev/null). usb_uhci's interrupt handler of course always returns IRQ_NONE, so after every 100000 interrupts "irq 16: nobody cared" message is printed. The interrupts are delivered to IRQ 16 if and only if threaded hardirqs are being used. This can be verified by compiling the kernel with: CONFIG_PREEMPT_NONE=y # CONFIG_PREEMPT_SOFTIRQS is not set CONFIG_PREEMPT_HARDIRQS=y When booted with "hardirq-preempt=off", the interrupt count on IRQ 16 remains zero all the time. After enabling threaded hardirqs: echo 1 > /proc/sys/kernel/hardirq_preemption the problem appears. If hardirq_preemption is then disabled again, IRQ 16 acitivity ceases. I haven't yet found the reason why this happens. This is the same problem as in bug 236150, only the IRQ numbers are different (IRQ 20->16 vs. IRQ 16->9). *** This bug has been marked as a duplicate of 236150 *** |