From: Theo Buehler <tb@theobuehler.org>
Subject: Re: Fault upgrade counters
To: tech@openbsd.org
Date: Mon, 24 Feb 2025 13:27:29 +0100

On Mon, Feb 24, 2025 at 01:23:51PM +0100, Martin Pieuchot wrote:
> On 24/02/25(Mon) 13:02, Theo Buehler wrote:
> > On Mon, Feb 24, 2025 at 12:36:34PM +0100, Martin Pieuchot wrote:
> > > Diff below adds a new counter to measure lock upgrades, documents it and
> > > unifies the re-lock counters to do only one atomic operation instead of
> > > two.
> > > 
> > > ok?
> > 
> > Is there a reason that you didn't add flt_{,no}up to vmstat.c?
> 
> It isn't clear to me which counters should be present or not in such
> view.  I can add them if this is a request.

It is not, it just wasn't clear to me if it was an oversight. I'm ok
with the diff as it is.

> 
> > other than that this reads ok
> > 
> > > Index: usr.bin/systat/vmstat.c
> > > ===================================================================
> > > RCS file: /cvs/src/usr.bin/systat/vmstat.c,v
> > > diff -u -p -r1.96 vmstat.c
> > > --- usr.bin/systat/vmstat.c	28 Dec 2022 20:49:05 -0000	1.96
> > > +++ usr.bin/systat/vmstat.c	24 Feb 2025 11:22:08 -0000
> > > @@ -407,7 +407,7 @@ showkre(void)
> > >  	PUTRATE(uvmexp.forks_sharevm, VMSTATROW + 2, VMSTATCOL + 3, 6);
> > >  	PUTRATE(uvmexp.fltpgwait, VMSTATROW + 3, VMSTATCOL + 4, 5);
> > >  	PUTRATE(uvmexp.fltrelck, VMSTATROW + 4, VMSTATCOL + 3, 6);
> > > -	PUTRATE(uvmexp.fltrelckok, VMSTATROW + 5, VMSTATCOL + 3, 6);
> > > +	PUTRATE(uvmexp.fltnorelck, VMSTATROW + 5, VMSTATCOL + 3, 6);
> > >  	PUTRATE(uvmexp.fltnoram, VMSTATROW + 6, VMSTATCOL + 3, 6);
> > >  	PUTRATE(uvmexp.fltamcopy, VMSTATROW + 7, VMSTATCOL + 3, 6);
> > >  	PUTRATE(uvmexp.flt_prcopy, VMSTATROW + 8, VMSTATCOL + 3, 6);
> > > 
> > > 
> > 
> 
>