Subject: Re: bin/20316: systat interrupt counters
To: David Laight <firstname.lastname@example.org>
From: Johnny Billquist <email@example.com>
Date: 02/12/2003 23:51:30
On Wed, 12 Feb 2003, David Laight wrote:
> > systat have recently had the "new" interrupt counters added.
> > However, when displaying the names of these counters, systat prefers
> > displaying the counter name to the counter group name.
> > This is wrong, I believe.
> I do not believe so. On i386 the full names are:
> group counter
> cpu0 softclock
> cpu0 softnet
> cpu0 softserial
> pic0 pin 14
> pic0 pin 10
> pic0 pin 9
> pic0 pin 4
> pic0 pin 3
> pic0 pin 1
> pic0 pin 12
> pic0 pin 0
> So just showing the group names would be rather pointless.
Sigh. The same problem but the other way...
> > The counter name is group specific, and is not guaranteed to be unique.
> True, but there are very likely to be many counters in a single grpoup.
To quote the man-page (9 evcnt):
Each event counter also has a group name (ev_group) and an event name
(ev_name) which are used to identify the counter. The group name may be
shared by a set of counters. For example, device interrupt counters
would use the name of the device whose interrupts are being counted as
the group name. The counter name is meant to distinguish the counter
from others in its group (and need not be unique across groups).
Given this, the group name ought to be the most "interesting" name to
display. Take as much of the rest as possible, truncating at the right
> > By prefering to show this string if both strings don't fit, you get a list
> > that is difficult to decipher.
> > In addition systat also truncates group and name if they don't fit, which
> > removed important information.
> Run systat in a wider window. Both are actually allowed to be quite long.
I'm running on a VAX. What do you mean "wider window"? My VT320 is as wide
as it is.
> > The name of the event counter is often just "intr"
> > (atleast on VAXen), which means you just get a lists with "intr".
> The display looks a lot better on i386, I put it to you that there are
> more i386 systems running netbsd than vax.
I can't argue with that one.
> Perhaps the actual problem is the unhelpful naming of the counters.
I really cannot do, or say much more on the subject. Could someone who
designed the evcnt stuff perhaps clarify what is to be expected in the two
On my VAX, it currently looks like this, anyway:
interrupt total rate
crl intr 0 0
uba stray intr 0 0
uba0 intr 164 0
mtc0 intr 6 0
uda0 intr 315053 3
de0 intr 513282 5
dmf0 srintr 0 0
dmf0 stintr 0 0
dmf0 piva 0 0
dmf0 pivb 0 0
dmf0 arintr 15723 0
dmf0 axintr 19283 0
dmf0 lpintr 0 0
uba1 intr 0 0
mtc1 intr 6 0
uda1 intr 536097 5
soft serial 0 0
soft net 178646 1
soft clock 161566 1
clock intr 10252164 100
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: firstname.lastname@example.org || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol