On 2007/01/18 11:46, Vijay Sankar wrote: > apps1# vmstat -i > interrupt total rate > irq7/ohci0 3 0 > irq5/ehci0 128 0 > irq10/pciide1 12694 6 > irq5/azalia0 576 0 > irq11/nfe0 3782 1 > irq0/clock 195337 100 > irq8/rtc 250015 128 > Total 462535 237
nothing stands out there. if you leave it doing something which should use a particular device (e.g. ls -lR /, or ping -f something) you should see a bunch of interrupts against that device. if you see a bunch of interrupts against an unrelated device that's a sure sign there's a problem. 'systat vmstat' gives a full-screen display which may be easier to watch interactively if you're testing like that. anything wierd there, try a new bios if there is one before you spend a long time chasing it. some vendors release new bioses that fix up interrupt tables and don't bother listing it in the changelog... (hello supermicro). > I did not see anything unusual with top -- most of the time is being taken by > cc1. you should look at the cpu states line. When the system is idle all should be low. When compiling a kernel you'd expect most cpu% should be in user, some system, maybe a bit in interrupt. Also see what state any blocked processes which you expect to be running are in. (WAIT column)
