PF/ALTQ - Stable TSC?
nightrecon at hotmail.com
Fri Dec 16 02:15:00 UTC 2011
> I'm setting up pf with altq support in my kernel on freebsd 9.0-Stable
> (soon to switch to the -RELEASE once it's available).
> The system is a quad-core Xeon E31220, running amd64.
> I've done a bit of googling and found various results. I know the
> freebsd handbook says ALTQ_NOPCC is required on SMP systems. My
> kern.timecounter.smp_tsc=1, which says the TSC is safe to use in SMP
> mode. Is it still required to use ALTQ_NOPCC on _ALL_ smp systems?
> Basically I'm just seeing very different answers with my own research,
> which is why I'm posting to this list.
In the /usr/src/sys/conf/NOTES on an 8.2 box it has this to say:
"ALTQ requires a stable TSC so if yours is broken or changes with CPU
throttling then you must also have the ALTQ_NOPCC option."
I take this to mean that if your TSC is absolutely rock solid and does not
vary or adjust when CPU throttling kicks in you may very well get away with
not having it.
My take on this is it probably doesn't hurt to leave it in, as it would be a
safety net in place for a "just in case" scenario in which case it would
enhance stability. Cheap insurance.
I suspect the Handbook line you were referring to might date back to the 5.x
days, with the quote above being recent. A lot of work in the three
timecounters available went as water under the bridge some time ago.
More information about the freebsd-questions