On Sat, Aug 29, 2009 at 11:10 AM, LDB<thesource@xxxxxxxxxxx> wrote: > > My 1 cent here ... > > I would read this well written documentation, > > http://irqbalance.org/documentation.php Yes, I have. > In addition, stop your irqbalance daemon, and execute the > following in the foreground: > > irqbalance --debug Except, as I said, I've already manually rerouted all of the interrupts away from CPU0, /proc/interrupts shows nothing going to CPU0 anymore, and yet CPU0 still has high sys+irq load. (By "manually rerouted", I mean that I killed the irqbalance daemon and essentially put 000000e into /proc/irq/*/smp_affinity, except for the various network interrupts, which I forced individually to CPU1, CPU2, and CPU3.) > NAPI is also called "RX polling", because it uses a mixture of > polling and interrupts to process incoming network frames. I take that to mean that it's pretty much irrelevant to my situation, then, since I have negligible RX traffic. Well over 99% of my traffic is outgoing. > Also, for the e1000 driver, I would look at your kernel's documentation: > > /usr/src/linux/Documentation/networking/e1000.txt > > You will notice the e1000 driver has NAPI (RX polling) enabled by > default, as shown by your ethtool commands above. Thanks. > As far as the tg3 driver, going through the source code for > tg3, /usr/src/linux/drivers/net/tg3.c, you will notice the RX > polling related code for NAPI. Thanks. That would probably have been a better way for me to answer my question about RX vs TX too. I'll read through it. Any hints about how to figure out what this mystery CPU0 irq-related CPU usage is? I'll be trying to get further into systemtap to see if I can get something out of it. It looks like there are trace points about where I'd need them. There isn't really a great mystery, I suppose -- the irq handling really seems like it must be related to the high TX traffic rate, but the questions are (1) why is it going to CPU0, and (2) why is it so different across identical hardware? -- 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
