Re: [PATCH] softirq: stable v3.1[23] (others?) have screaming tasklet disease - ksoftirqd[random] eats 100% CPU

2014-02-19 Thread Mike Galbraith
On Thu, 2014-02-20 at 01:00 +0100, Thomas Gleixner wrote: > On Wed, 19 Feb 2014, Mike Galbraith wrote: > > I'm seeing ksoftirqd chewing 100% CPU on one or more CPUs in both 3.12 > > and 3.13, as below in a 40 core (+smt) box. It should look very > > familiar to CCs, especially Ingo. > > > >

Re: [PATCH] softirq: stable v3.1[23] (others?) have screaming tasklet disease - ksoftirqd[random] eats 100% CPU

2014-02-19 Thread Dan Williams
On Wed, Feb 19, 2014 at 4:00 PM, Thomas Gleixner wrote: > On Wed, 19 Feb 2014, Mike Galbraith wrote: >> I'm seeing ksoftirqd chewing 100% CPU on one or more CPUs in both 3.12 >> and 3.13, as below in a 40 core (+smt) box. It should look very >> familiar to CCs, especially Ingo. >> >> Below,

Re: [PATCH] softirq: stable v3.1[23] (others?) have screaming tasklet disease - ksoftirqd[random] eats 100% CPU

2014-02-19 Thread Thomas Gleixner
On Wed, 19 Feb 2014, Mike Galbraith wrote: > I'm seeing ksoftirqd chewing 100% CPU on one or more CPUs in both 3.12 > and 3.13, as below in a 40 core (+smt) box. It should look very > familiar to CCs, especially Ingo. > > Below, tasklet is disabled by ioat2_free_chan_resources, and what I >

[PATCH] softirq: stable v3.1[23] (others?) have screaming tasklet disease - ksoftirqd[random] eats 100% CPU

2014-02-19 Thread Mike Galbraith
I'm seeing ksoftirqd chewing 100% CPU on one or more CPUs in both 3.12 and 3.13, as below in a 40 core (+smt) box. It should look very familiar to CCs, especially Ingo. Below, tasklet is disabled by ioat2_free_chan_resources, and what I presume was systemd-udevd-1050 starts screaming when it

[PATCH] softirq: stable v3.1[23] (others?) have screaming tasklet disease - ksoftirqd[random] eats 100% CPU

2014-02-19 Thread Mike Galbraith
I'm seeing ksoftirqd chewing 100% CPU on one or more CPUs in both 3.12 and 3.13, as below in a 40 core (+smt) box. It should look very familiar to CCs, especially Ingo. Below, tasklet is disabled by ioat2_free_chan_resources, and what I presume was systemd-udevd-1050 starts screaming when it

Re: [PATCH] softirq: stable v3.1[23] (others?) have screaming tasklet disease - ksoftirqd[random] eats 100% CPU

2014-02-19 Thread Thomas Gleixner
On Wed, 19 Feb 2014, Mike Galbraith wrote: I'm seeing ksoftirqd chewing 100% CPU on one or more CPUs in both 3.12 and 3.13, as below in a 40 core (+smt) box. It should look very familiar to CCs, especially Ingo. Below, tasklet is disabled by ioat2_free_chan_resources, and what I presume

Re: [PATCH] softirq: stable v3.1[23] (others?) have screaming tasklet disease - ksoftirqd[random] eats 100% CPU

2014-02-19 Thread Dan Williams
On Wed, Feb 19, 2014 at 4:00 PM, Thomas Gleixner t...@linutronix.de wrote: On Wed, 19 Feb 2014, Mike Galbraith wrote: I'm seeing ksoftirqd chewing 100% CPU on one or more CPUs in both 3.12 and 3.13, as below in a 40 core (+smt) box. It should look very familiar to CCs, especially Ingo.

Re: [PATCH] softirq: stable v3.1[23] (others?) have screaming tasklet disease - ksoftirqd[random] eats 100% CPU

2014-02-19 Thread Mike Galbraith
On Thu, 2014-02-20 at 01:00 +0100, Thomas Gleixner wrote: On Wed, 19 Feb 2014, Mike Galbraith wrote: I'm seeing ksoftirqd chewing 100% CPU on one or more CPUs in both 3.12 and 3.13, as below in a 40 core (+smt) box. It should look very familiar to CCs, especially Ingo. Below, tasklet