Re: diagnosing interrupt storms?

From: John Baldwin <jhb_at_FreeBSD.org>
Date: Wed, 10 Mar 2004 13:01:24 -0500
On Tuesday 09 March 2004 01:19 pm, Alfred Perlstein wrote:
> * Doug White <dwhite_at_gumbysoft.com> [040309 09:57] wrote:
> > On Mon, 8 Mar 2004, Alfred Perlstein wrote:
> > > At a certain point after booting my SMP 5-current box gets all
> > > weird, typically I see 50%+ time spent in interrupt.  If I run "top
> > > -S" I typically see one of the ithreads using 50% cpu.
> > >
> > > I'm trying to figure out what it's doing, what has gone wrong etc.
> > >
> > > Are there any sysctls to look at or things I can do to diagnose
> > > this?
> >
> > vmstat -i is useful for figuring out exactly which interrupt is storming,
> > although the ithread might hint it too.
> >
> > Which one is it?
>
> It's typically been em (intel gigE) or the sound interrupt.

Is it on irq 20 and shared with acpi0?  Alternatively, it could be misrouted.  
Are you using X and drm?  Many BIOSs seem to forget that VGA devices can have 
IRQs and don't provide interrupt routing information for them, so drm0 gets 
the wrong IRQ and the correct IRQ gets a storm when the drm0 device 
interrupts.  I guess window doesn't use the VGA IRQ in its graphics drivers 
since BIOS writers seem to only care about what makes Windows works and tend 
to demonstrate a stricking lack of proficiency in English when it comes to 
written standards such as ACPI, etc.

-- 
John Baldwin <jhb_at_FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org
Received on Wed Mar 10 2004 - 09:16:04 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:46 UTC