On Thu, 20 Mar 2008 16:00:28 +0100 (CET), Jan Engelhardt <jengelh@xxxxxxxxxxxxxxx> wrote: > But if the processing of packets takes most of the time, then you > should not worry about balancing NIC IRQs, but enqueueing the > packets onto the right softirq thread. > >> I don't really know how IRQs are handled on SMP (nor I know >> how irq-balance works) > > IRQs are enabled for one CPU only in the general case, and > balancing it should be straightforward. > I have no idea what happens if the affinity mask allows an IRQ > for more than one CPU, but you should not worry about that, because > > (a) you probably do not want a deliver-to-a-random-CPU behavior nor > (b) deliver-to-all > > (since I doubt the hardware does (c) round-robin automatically) I'm not sure that hardware does not automatically balance IRQs (/proc/interrupts/#/smp_affinity is a bitfield used to balance IRQs, and I suppose the APIC does the balancing). Also, I found this thread on this maling list : http://marc.info/?t=119094666500002&r=1&w=4 I'll take a look at this patch, it looks really interesting. regards, strawks -- To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
