Stephen Hemminger wrote: >> My kernel has SMP and Message-Signaled Interrupts (MSI) enabled. When I >> boot with MSI enabled, I can see in /proc/interrupts that each of the >> NICs got a unique interrupt assigned. But all traffic generated to the >> 1st NIC goes to the CPU#0, other CPUs show 0 processed interrupts in the >> /proc/interrupts. > > This is actually a designed beneficial state. If IRQ's happen on different > CPU's then there is more cache memory thrashing and performance is lower. > OK, that's a good explanation. Now assuming the designed CPU processes the incoming IRQ and puts (e.g.) incoming packet/skb into a queue. I read once in the past, that these queues are per-CPU (to avoid locking), so the same CPU has to process the packet doing iptables stuff etc. Is that true ? -Svata - 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
