Re: [patch] voluntary-preempt-2.6.8.1-P0

[Lee Revell]

On Sun, 2004-08-15 at 07:56, Ingo Molnar wrote:
> i've uploaded the -P0 patch:
> 
>  http://redhat.com/~mingo/voluntary-preempt/voluntary-preempt-2.6.8.1-P0
> 
> those who had APIC (and USB, under SMP) problems under previous
> versions, are the problems still present in -P0?
>  

The mlockall issue is still not resolved; however, I did manage to get a
trace, which was probably not possible before because some higher
latency but lower frequency event was overwriting /proc/latency_trace. 
So, maybe mlockall does cause xruns by having many shorter, but long
enough to be problematic, non-preemptible sections.

http://krustophenia.net/testresults.php?dataset=2.6.8.1-P0

Also it seems that extract_entropy still causes high latencies, even
though a call to preempt_schedule was added.  I looked at the code in
random.c a bit and this strikes me as an area where the algorithm could
be improved, rather than adding a scheduling point.  Do we really need
*that* much entropy, right then?  And if so, isn't there a zero-copy
solution?

Lee



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/