netdev
[Top] [All Lists]

Re: Kernel connector - userspace <-> kernelspace "linker".

To: jamal <hadi@xxxxxxxxxx>
Subject: Re: Kernel connector - userspace <-> kernelspace "linker".
From: Evgeniy Polyakov <johnpol@xxxxxxxxxxx>
Date: Mon, 20 Sep 2004 12:01:37 +0400
Cc: netdev@xxxxxxxxxxx
In-reply-to: <1095336548.1064.167.camel@xxxxxxxxxxxxxxxx>
Organization: MIPT
References: <1095331899.18219.58.camel@uganda> <1095336548.1064.167.camel@xxxxxxxxxxxxxxxx>
Reply-to: johnpol@xxxxxxxxxxx
Sender: netdev-bounce@xxxxxxxxxxx
On Thu, 2004-09-16 at 16:09, jamal wrote:
> On Thu, 2004-09-16 at 06:51, Evgeniy Polyakov wrote:
> > Hmm, do not know how to describe... 
> > Kind of mega-picture can be found at 
> > http://tservice.net.ru/~s0mbre/?section=gallery&item=connector_design
> > 
> > This driver adds possibility to connect anything with anything using
> > netlink based network.
> > One must register callback and identificator. When driver receives
> > special netlink message with appropriate identificator, appropriate
> > callback will be called.
> > I think that the code better explains what I'm trying to say.
> 
> I dont have time to evaluate the code right now - but will at the end of
> day. Just trying to understand your concepts:
> 
> Essentially you are building a unified messaging for
> userspace-userpace(i.e IPC), userspace-kernel, kernel-kernel with each
> component residing in whatever spot (kernel or userland)having a unique
> name and Id. Is this correct?
> Clearly there could be name/Id conflicts etc. Are you addressing those? 
> Any event and filtering capability already in or planned? Example event
> I want to be notified when module with name "sean paul" comes online and
> filter will be "it has to have ID in the range of 0x200-0x400".
> 
> I am still trying to wakeup but it does sound like a good idea to have a
> generic messaging subsystem.

Attached patch cleans it up a bit and adds notify removal mechanism.
BTW, patch for cn_test.c test module is quite ugly, but it is only for
test. Real users can use cn_netlink_send().

It looks like noone objects, so I will create a patch and send it to
GregKH :) 
I know at least 2 potential users - w1 and pending superio( actually it
requires only connector to be included). I'm quite sure after I will
write bits of documentation for this cruft it will take broader usage.

Thank you.

> cheers,
> jamal 
-- 
        Evgeniy Polyakov

Crash is better than data corruption. -- Art Grabowski

Attachment: connector.diff
Description: Text Data

Attachment: signature.asc
Description: This is a digitally signed message part

<Prev in Thread] Current Thread [Next in Thread>