Date: Fri, 19 Oct 2007 22:40:08 -0400 From: "Constantine A. Murenin" <cnst@FreeBSD.org> To: Poul-Henning Kamp <phk@phk.freebsd.dk> Cc: Alexander@Leidinger.net, "Constantine A. Murenin" <cnst@FreeBSD.org>, arch@FreeBSD.org Subject: Re: sensors fun.. Message-ID: <47196A88.90104@FreeBSD.org> In-Reply-To: <83490.1192810356@critter.freebsd.dk> References: <83490.1192810356@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On 19/10/2007 12:12, Poul-Henning Kamp wrote:
> In message <20071019.100516.74722974.imp@bsdimp.com>, Warner Losh writes:
>
>
>>>The kernel-userland interface should happen over a filedescriptor
>>>(either device or unix-domain socket) so that whatever daemon we
>>>park on the fd can just use select/poll/kqueue to wait for events.
>>
>>If we're going to have a stream of data from the kernel, is there any
>>reason to invent another daemon for that? We already have devd that
>>deals with a number of disparate events from the kernel in a fairly
>>generic way.
>
>
> The kernel and userland sensors would result in two kinds of data,
> measurements ("32°C") and events ("new sensor", "high temp")
In such case, measurements can be taken through the sensors framework
and events could be sent via the devctl(4) mechanism. So it doesn't
look like we disagree so far, now does it? :-)
> devd should not see the measurements, but it might be a good idea
> if it could see the events from all sensors (userland AND kernel).
Sure, we can add some devctl_notify calls to the functions that are used
to register sensor devices. I don't see any problem with this.
As far as the userland part goes -- well, I totally agree with Alexander
about his email mentioning eierlegendewollmichsau. :-)
Cheers,
Constantine.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?47196A88.90104>