https://bugzilla.novell.com/show_bug.cgi?id=768391 https://bugzilla.novell.com/show_bug.cgi?id=768391#c0 Summary: irqbalance SuSEconfig documentation Classification: openSUSE Product: openSUSE 11.4 Version: Factory Platform: All OS/Version: SLES 11 Status: NEW Severity: Normal Priority: P5 - None Component: Other AssignedTo: bnc-team-screening@forge.provo.novell.com ReportedBy: brian.cain@gmail.com QAContact: qa-bugs@suse.de Found By: --- Blocker: --- User-Agent: Mozilla/5.0 (Windows NT 6.1) AppleWebKit/536.5 (KHTML, like Gecko) Chrome/19.0.1084.56 Safari/536.5 The documentation in /etc/sysconfig/irqbalance says that “'auto' is the default behaviour which means that one socket but multi core CPUs get a oneshot setup.” The subtlety about multi-socket vs single-socket systems seems incorrect. Where does this come from? Also, the point in general seems wrong: "auto" bars the IRQBALANCE_ONESHOT environment variable from being set: if [ "$IRQBALANCE_ONESHOT" != "auto" ] ; then export IRQBALANCE_ONESHOT fi Reproducible: Always Steps to Reproduce: x:~> sudo service irq_balancer status Checking for service irqbalance unused x:~> sudo service irq_balancer start Starting irqbalance done x:~> sudo service irq_balancer status Checking for service irqbalance running x:~> pgrep irqbalance 1060 x:~> cat /etc/sysconfig/irqbalance ## Path: System/Kernel/IRQ Balancing ## Description: IRQ Balancing configuration ## Type: string ## Default: "" ## ServiceRestart: irq_balancer # # Bitmap of CPUs that are banned from IRQ balancing # IRQBALANCE_BANNED_CPUS="" ## Type: list(auto,enabled) ## Default: auto ## ServiceRestart: Iirq_balancer # # Instead of running as daemon, the IRQ balancer just does a setup # once and then exits. "auto" is the default behaviour which means that # one socket but multi core CPUs get a oneshot setup. "enabled" always # enables oneshot mode, regardless of the CPUs. # IRQBALANCE_ONESHOT=auto ## Type: string ## Default: "" ## ServiceRestart: Iirq_balancer # # Interrupt that don't get balanced as list (separation character doesn't # matter). For example broken chipsets don't allow the timer interrupt # to be set to another CPU than the first, and with this option that # policy can be applied. # IRQBALANCE_BANNED_INTERRUPTS="" x:~> ls -l /sys/devices/system/node/ total 0 -r--r--r-- 1 root root 4096 2012-06-21 10:39 has_cpu -r--r--r-- 1 root root 4096 2012-06-21 10:39 has_normal_memory drwxr-xr-x 2 root root 0 2012-06-21 10:39 node0 drwxr-xr-x 2 root root 0 2012-06-21 10:39 node1 -r--r--r-- 1 root root 4096 2012-06-21 10:39 online -r--r--r-- 1 root root 4096 2012-06-21 10:39 possible Actual Results: IRQBALANCE_ONESHOT=auto did not do a single oneshot IRQ balance and the daemon was running afterwards. Expected Results: IRQBALANCE_ONESHOT=auto "should" (according to reference) do a single oneshot IRQ balance and the daemon should not be running afterwards. In all likelyhood the appropriate solution lies in correcting the misleading content in /etc/sysconfig/irqbalance -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.