Subject: Re: correctly counting user/sys/interrupt time
To: Frank Kardel <Frank.Kardel@Acrys.COM>
From: David Laight <firstname.lastname@example.org>
Date: 04/10/2006 21:29:02
On Mon, Apr 10, 2006 at 09:25:41AM +0200, Frank Kardel wrote:
> while you are at it, have you (or anyone else) considered
> instrumenting locks (lockmanager & maybe even simple) for collecting
> their locking time (min, max, accumulated, maybe even
> a historam histogram)? During debugging interrupts and timecounter
> implementation I got the impression that there may be
> some bigger locks in the code as clock interrupts (100Hz rate)
> vary by more 10% quite a few times. This has been observed on
> MP systems.
However if NetBSD does soft interrupt masking, it should be possible
to measure the time between an IRQ being asserted and the real
interrupt handler being called.
In the past I used a logic analiser (in timing mode) to trigger when
a device IRQ line was asserted for more than a few ms, wired the
'trigger out' to the system NMI line, and thus got a system panic dump
when the IRQ was masked for too long. Turned out to be code that
scrolled the VGA (text) console screen....
David Laight: email@example.com