Hi!
On 13:00 Tue 16 Sep , jasjit singh wrote:
> Thanks Himanshu, Michael and Mitul for you replies. Most of the things are
> now clear to me. I think work queue is best option for me to use.
>
> As my driver is a network driver, I am expecting heavy interrupt load on the
> system. This means new interrupt may (or will) come before the bottom half
> corresponding to previous one has been processed. This can result in queuing
> up of bottom halves. My question is - Like ksoftirqd handles the situation
> when too much work is pending in softirq context, is there a similar entity
> which does the same when too much work is pending with the work queues.
Many network drivers have some mechanisms in case of high load:
- interrupt mitigation: Do periodic polling and process many packets at once
if the number of packets/sec exceeds some threshold.
- Tell the network card to stop receiving, if the system cannot cope with the
load.
-Michi
--
programing a layer 3+4 network protocol for mesh networks
see http://michaelblizek.twilightparadox.com
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to [EMAIL PROTECTED]
Please read the FAQ at http://kernelnewbies.org/FAQ
