From: "Theo de Raadt" <deraadt@openbsd.org>
Subject: Re: iostat cpu stats
To: "Ted Unangst" <tedu@tedunangst.com>
Cc: tech@openbsd.org
Date: Wed, 07 May 2025 08:00:07 -0600

Please put a comment there "limited to 99 because there is no more printing room"
or something.

Ted Unangst <tedu@tedunangst.com> wrote:

> Today's featured column width is iostat.
> 
> Running iostat -w 1 I see a lot of messy output in the idle column.
> 
>    0   68  0.00    0    0.00   0.00    0    0.00   0  0  1  0  0 99
>    0   70  0.00    0    0.00   0.00    0    0.00   0  0  0  0  0100
>    0   69  2.00    2    0.00   2.00    2    0.00   0  0  0  0  0100
>    0   69  0.00    0    0.00   0.00    0    0.00   0  0  0  0  1 99
> 
> Looking at the code, there's even a comment that 100% is too rare to
> worry about. But my laptop is faster than the vaxen of yore, and I see
> 100 all the time. I think a passable fix is to simply truncate to 99.
> The columns are never guaranteed to sum up, and 100 in one column is
> always a rounding error, so I think this is sufficiently accurate.
> 
>    0   69  0.00    0    0.00   0.00    0    0.00   0  0  0  0  0 99
>    0   69 16.00    2    0.03  16.00    2    0.03   0  0  0  0  0 99
>    0   69 14.25   16    0.22  14.25   16    0.22   0  0  0  0  0 99
>    0   68  0.00    0    0.00   0.00    0    0.00   0  0  0  0  1 99
> 
> Index: iostat.c
> ===================================================================
> RCS file: /home/cvs/src/usr.sbin/iostat/iostat.c,v
> diff -u -p -r1.47 iostat.c
> --- iostat.c	8 Mar 2023 04:43:13 -0000	1.47
> +++ iostat.c	6 May 2025 21:30:33 -0000
> @@ -354,8 +354,12 @@ cpustats(void)
>  	if (!t)
>  		t = 1.0;
>  	/* States are generally never 100% and can use %3.0f. */
> -	for (state = 0; state < CPUSTATES; ++state)
> -		printf("%3.0f", 100. * cur.cp_time[state] / t);
> +	for (state = 0; state < CPUSTATES; ++state) {
> +		double v = 100. * cur.cp_time[state] / t;
> +		if (v > 99)
> +			v = 99;
> +		printf("%3.0f", v);
> +	}
>  }
>  
>  static void
>