ksoftirqd using 100% CPU
Hello listers, Last night I upgraded one of the Linux guests used by our developers to kernel level 2.6.16.60-0.95.1-default. (Latest SLES10 SP4 kernel) By about 11:30 AM this morning, I was getting complaints from our developers that the system was hanging. Looking around, I saw that ksoftirqd was using close to 100% cpu. Anyone got any ideas on what to check? I have looked in the archives, but they are a little old. The only messages I can find have to do with a hipersockets time out. Feb 9 11:21:39 bus0104 kernel: NETDEV WATCHDOG: hsi0: transmit timed out Feb 9 11:21:39 bus0104 kernel: qeth: Recovery of device 0.0.5100 started ... Feb 9 11:21:39 bus0104 kernel: qeth: Device 0.0.5100/0.0.5101/0.0.5102 is a HiperSockets card (level: HSEC) Feb 9 11:21:39 bus0104 kernel: with link type HiperSockets. Feb 9 11:21:39 bus0104 kernel: qeth: Hardware IP fragmentation not supported on hsi0 Feb 9 11:21:39 bus0104 kernel: qeth: VLAN enabled Feb 9 11:21:39 bus0104 kernel: qeth: Multicast enabled Feb 9 11:21:39 bus0104 kernel: qeth: IPV6 enabled Feb 9 11:21:39 bus0104 kernel: qeth: Broadcast enabled Feb 9 11:21:39 bus0104 kernel: qeth: Using SW checksumming on hsi0. Feb 9 11:21:39 bus0104 kernel: qeth: Outbound TSO not supported on hsi0 Feb 9 11:21:39 bus0104 kernel: qeth: Device 0.0.5100 successfully recovered! q Anyone got any ideas on what to check? I have looked in the archives, but they are a little old. Thanks, Ron -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/
Re: ksoftirqd using 100% CPU
On Thu, Feb 09, 2012 at 04:43:25PM -0600, Ron Foster at Baldor-IS wrote: > The only messages I can find have to do with a hipersockets time out. > Feb 9 11:21:39 bus0104 kernel: NETDEV WATCHDOG: hsi0: transmit timed out > Feb 9 11:21:39 bus0104 kernel: qeth: Recovery of device 0.0.5100 > started ... Hmm, that shouldn't happen. If you have a support agreement, can you open a service request with Novell / SUSE, please? We'll need the files generated by running "supportconfig" and "dbginfo.sh", and if you already have opened a case with IBM, the PMR would be handy. Regards, Joerg -- Joerg Reuterhttp://yaina.de/jreuter And I make my way to where the warm scent of soil fills the evening air. Everything is waiting quietly out there (Anne Clark) -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/
Re: ksoftirqd using 100% CPU
On Fri, Feb 10th, 2012 at 11:00 PM, Joerg Reuter wrote: > On Thu, Feb 09, 2012 at 04:43:25PM -0600, Ron Foster at Baldor-IS wrote: > > > The only messages I can find have to do with a hipersockets time out. > > Feb 9 11:21:39 bus0104 kernel: NETDEV WATCHDOG: hsi0: transmit timed > out > > Feb 9 11:21:39 bus0104 kernel: qeth: Recovery of device 0.0.5100 > > started ... > > Hmm, that shouldn't happen. Lol ... I was wondering if the OP could see anything from z/VM, but if it really is a softirq/tasklet problem that probably means (Linux) dodgy driver. And in almost all occasions that probably means network ... Keeping an eye on /proc/interrupts might be instructive. Shane ... -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/
Re: ksoftirqd using 100% CPU
Shane, As best we can tell, there were no relevant messages that came out on the vm operator console. Ron From: Linux on 390 Port [LINUX-390@VM.MARIST.EDU] On Behalf Of Shane G [ibm-m...@tpg.com.au] Sent: Friday, February 10, 2012 6:46 AM To: LINUX-390@VM.MARIST.EDU Subject: Re: ksoftirqd using 100% CPU On Fri, Feb 10th, 2012 at 11:00 PM, Joerg Reuter wrote: > On Thu, Feb 09, 2012 at 04:43:25PM -0600, Ron Foster at Baldor-IS wrote: > > > The only messages I can find have to do with a hipersockets time out. > > Feb 9 11:21:39 bus0104 kernel: NETDEV WATCHDOG: hsi0: transmit timed > out > > Feb 9 11:21:39 bus0104 kernel: qeth: Recovery of device 0.0.5100 > > started ... > > Hmm, that shouldn't happen. Lol ... I was wondering if the OP could see anything from z/VM, but if it really is a softirq/tasklet problem that probably means (Linux) dodgy driver. And in almost all occasions that probably means network ... Keeping an eye on /proc/interrupts might be instructive. Shane ... -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/ -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/
Re: ksoftirqd using 100% CPU
Jeorg, We have a support agreement with Novell, so I will be opening up a service request. The developers were getting hostile, so I had to revert the development system that was consistently having the problem to a previous level. But I have found another system that had the problem once yesterday that we can use to gather the documentation. Ron I did not know that we could open a PMR with IBM directly. From: Linux on 390 Port [LINUX-390@VM.MARIST.EDU] On Behalf Of Joerg Reuter [jreu...@suse.de] Sent: Friday, February 10, 2012 6:00 AM To: LINUX-390@VM.MARIST.EDU Subject: Re: ksoftirqd using 100% CPU On Thu, Feb 09, 2012 at 04:43:25PM -0600, Ron Foster at Baldor-IS wrote: > The only messages I can find have to do with a hipersockets time out. > Feb 9 11:21:39 bus0104 kernel: NETDEV WATCHDOG: hsi0: transmit timed out > Feb 9 11:21:39 bus0104 kernel: qeth: Recovery of device 0.0.5100 > started ... Hmm, that shouldn't happen. If you have a support agreement, can you open a service request with Novell / SUSE, please? We'll need the files generated by running "supportconfig" and "dbginfo.sh", and if you already have opened a case with IBM, the PMR would be handy. Regards, Joerg -- Joerg Reuterhttp://yaina.de/jreuter And I make my way to where the warm scent of soil fills the evening air. Everything is waiting quietly out there (Anne Clark) -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/ -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/
Re: ksoftirqd using 100% CPU
On Friday, 02/10/2012 at 10:21 EST, Ron Foster at Baldor-IS wrote: > I did not know that we could open a PMR with IBM directly. Only if your Linux support contract is with IBM. Alan Altmark Senior Managing z/VM and Linux Consultant IBM System Lab Services and Training ibm.com/systems/services/labservices office: 607.429.3323 mobile; 607.321.7556 alan_altm...@us.ibm.com IBM Endicott -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/
Re: ksoftirqd using 100% CPU
>>> On 2/10/2012 at 11:33 AM, Alan Altmark wrote: > On Friday, 02/10/2012 at 10:21 EST, Ron Foster at Baldor-IS > wrote: > >> I did not know that we could open a PMR with IBM directly. > > Only if your Linux support contract is with IBM. But, if you've already opened a PMR with IBM for z/VM or for hardware (for whatever reason related to the incident), we'd like to know about it anyway. I know the same is true in the reverse, if a customer comes to us first, and then later needs to work with IBM on the problem because it involves IBM software or hardware. Mark Post -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/
Re: ksoftirqd using 100% CPU
FYI, Known problem with the .95 kernel. IBM has released a PTF and we are testing it. Ron From: Linux on 390 Port [LINUX-390@VM.MARIST.EDU] On Behalf Of Mark Post [mp...@novell.com] Sent: Friday, February 10, 2012 12:11 PM To: LINUX-390@VM.MARIST.EDU Subject: Re: ksoftirqd using 100% CPU >>> On 2/10/2012 at 11:33 AM, Alan Altmark wrote: > On Friday, 02/10/2012 at 10:21 EST, Ron Foster at Baldor-IS > wrote: > >> I did not know that we could open a PMR with IBM directly. > > Only if your Linux support contract is with IBM. But, if you've already opened a PMR with IBM for z/VM or for hardware (for whatever reason related to the incident), we'd like to know about it anyway. I know the same is true in the reverse, if a customer comes to us first, and then later needs to work with IBM on the problem because it involves IBM software or hardware. Mark Post -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/ -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/