After peeking a bit over sch_* code, I have a few basic, possibly weird,
questions. In the past (and conforming to e.g. tbf's man and comment in
sch_tbf.c) the watchdogs used in schedulers like tbf, htb, hfsc, ...
were all limited to 1/HZ resolution. After switching to hrtimers it's
not the case anymore, still:
- what is the current resolution allowed by hrtimer_start ? During
simple tests with tbf I could easily schedule at pretty much any rate
(tested up to 100mbit) with just a packet sized token bucket. And that
regardless of selected HZ and/or NO_HZ setting. Quite awesome.
- in case of some older hardware - I assume the functionality falls back
to the old behavior, so - overall accuracy is still 1.024us (judging
from comments, and contrary to older kernels where jiffies/
gettimeofday/ timestamp could be selected) - but allowed watchdog
resolution remains 1/HZ ? Also - what is NO_HZ's influence (if any) in
such scenario ?
--
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
