Just for completeness I have attached another rcu_preempt stall
warning from a kinux-3.17.1 kernel.
So it looks like these stall warning go back that far.
It could be that the kernel instrumentation improved enough to detect
this situation around the 3.17 kernel. I can't find any earlier
records
Well with the release of 4.4-rc8 I have built and
started testing the kernel.
With some luck I managed to get a rcu_preempt stall within a few hours
of testing.
Upon booting, I ran a small series of bench marks to make sure
everything is running as expected. limited regression testing and then
j
On Sun, Jan 03, 2016 at 07:27:17PM +1100, Ross Green wrote:
> I would not describe the load on this test machine as high or real time.
>
> Apart from a number of standard daemons not much more is running at all!
>
> I normally build a release kernel as soon as possible and set it running.
> Typic
I would not describe the load on this test machine as high or real time.
Apart from a number of standard daemons not much more is running at all!
I normally build a release kernel as soon as possible and set it running.
Typically I run a series of benchmarks to confirm most things appear
to be wo
On Sun, Jan 03, 2016 at 04:29:11PM +1100, Ross Green wrote:
> Still seeing these rcu_preempt stalls on kernels through to 4.4-rc7
>
> Still have not found a sure fire method to evoke this stall, but have
> found that it will normally occur within a week of running a kernel -
> usually when it is q
Still seeing these rcu_preempt stalls on kernels through to 4.4-rc7
Still have not found a sure fire method to evoke this stall, but have
found that it will normally occur within a week of running a kernel -
usually when it is quiet with light load.
Have seen similar self detected stalls all the
On Fri, Dec 11, 2015 at 10:17:33PM +1100, Ross Green wrote:
> I have been getting these stalls in kernels going back to 3.17.
>
> This stall occurs usually under light load but often requires several
> days to show itself. I have not found any simple way to trigger the
> stall. Indeed heavy worklo
7 matches
Mail list logo
