Subject: Re: bufq and sysctl question
To: None <juan@xtrarom.org>
From: YAMAMOTO Takashi <yamt@mwd.biglobe.ne.jp>
List: tech-kern
Date: 01/23/2006 05:28:34
> [juan@Nocturno][~]> sudo sysctl -w kern.bufq.readprio.read_burst=512
> kern.bufq.readprio.read_burst: 128 -> 512
> [juan@Nocturno][~]> sysctl kern.bufq.readprio
> kern.bufq.readprio.read_burst = 512
> kern.bufq.readprio.write_req = 64
> [juan@Nocturno][~]>
> 
> yamt, is it ready to commit?

i don't think having these sysctls is a good idea.
if we should have these knobs, they should be per-bufq.
and, IMO, they should be a part of DIOCGSTRATEGY/DIOCSSTRATEGY.
dks_param/dks_paramlen in struct disk_strategy was intended for
this kind of feature.
(ie. pass a string like "read_burst=512, write_req=64".
i think having an option string parser in kernel is not so wasteful,
as far as it can be used for others like mount(2).)

YAMAMOTO Takashi