irqbalance: unbalanced irqs or wrong irq assignment on system with shared level 3 cache

Solution Unverified - Updated -

Issue

  • On a ProLiant DL580 G7 system with Fibre Channel adapter (QLogic Corp. ISP2432-based 4Gb), since RHEL 6.6 there is a problem: all the adapter interrupts are assigned to the logical CPU 0.
# grep qla proc/interrupts 
            CPU0       CPU1       CPU2       CPU3      [.................]   CPU78      CPU79 
  25:        100          0          0          0      [.....all zero....]    0          0  IR-IO-APIC-fasteoi   qla2xxx
  26:     945318          0          0          0      [.....all zero....]    0          0  IR-IO-APIC-fasteoi   qla2xxx
  28:    4273234          0          0          0      [.....all zero....]    0          0  IR-IO-APIC-fasteoi   qla2xxx
  30:     991039          0          0          0      [.....all zero....]    0          0  IR-IO-APIC-fasteoi   qla2xxx
  32:    3991989          0          0          0      [.....all zero....]    0          0  IR-IO-APIC-fasteoi   qla2xxx
  37:        100          0          0          0      [.....all zero....]    0          0  IR-IO-APIC-fasteoi   qla2xxx
  40:    3924089          0          0          0      [.....all zero....]    0          0  IR-IO-APIC-fasteoi   qla2xxx
  42:    3974677          0          0          0      [.....all zero....]    0          0  IR-IO-APIC-fasteoi   qla2xxx
  • After update to RHEL 6.6, irqbalance doesn't distribute interrupts which are bound to cache level balance, on systems with shared level 3 cache.

Environment

  • Red Hat Enterprise Linux 6.6
  • irqbalance 1.0.4-10.el6

Subscriber exclusive content

A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.

Current Customers and Partners

Log in for full access

Log In

New to Red Hat?

Learn more about Red Hat subscriptions

Using a Red Hat product through a public cloud?

How to access this content