Skip to contentSkip to conversation

2.4.10pre7aa1

DISCUSSION Dipankar Sarma 2 participants 2 messages Sep 11, 2001

lore
Share & tools
lore
#0 lore raw
Quoted 5 lines · In article <[email protected]> you wrote:

many thanks. At the moment my biggest concern is about the need of call_rcu not to be starved by RT threads (keventd can be starved so then it won't matter if krcud is RT because we won't start using it).

Andrea

I think we can avoid keventd altogether by using a periodic timer (say 10ms) to check for completion of an RC update. The timer may be active only if only if there is any RCU going on in the system - that way we still don't have any impact on the rest of the kernel.

I am working on such a thing - but it will take me a little bit of time to figure out how to do this in linux.

Thanks
Dipankar

Signature · 2 lines
Dipankar Sarma  <[email protected]> Project: http://lse.sourceforge.net
Linux Technology Center, IBM Software Lab, Bangalore, India.
#1 lore raw
Quoted 11 lines · Dipankar Sarma
Quoted 5 lines · In article <[email protected]> you wrote:

many thanks. At the moment my biggest concern is about the need of call_rcu not to be starved by RT threads (keventd can be starved so then it won't matter if krcud is RT because we won't start using it).

Andrea

I think we can avoid keventd altogether by using a periodic timer (say 10ms) to check for completion of an RC update. The timer may be active only if only if there is any RCU going on in the system - that way we still don't have any impact on the rest of the kernel.

the timer can a have bigger latency than keventd calling wait_for_rcu so it should be a loss in a stright bench with light load, but OTOH we only care about getting those callbacks executed eventually and the advantage I can see is that the timer cannot get starved.

Andrea