RE: [Linux-streams] LiS 2.17.P Clone driver major number change.

[Dave Grothe]

If you write a loadable STREAMS module you can call the streams driver registration routine and get a major assigned dynamically.  You can then use this number in conjunction with lis_mknod() to make your own /dev nodes from within the init routine of your driver.  You can unregister and remove the nodes in your module unload routine.

The LiS loopback driver is now such a module and can serve as an example.

-- Dave

At 04:06 PM 6/23/2004, Gurol Akman wrote:

I always thought the following 2 low-cost features would make LiS very flexible across multiple HW platforms:
 
1. support for dynamic major number assignment;
 
2. automatic creation/removal of special device nodes within device init/cleanup routines;
 
If there is sufficient interest among the LiS community, I can work on patches for both. If many of you view this as a low-priority item as Dave does, then pls ignore this request. Regards -- GA

-----Original Message-----

From: Dave Grothe [mailto:[EMAIL PROTECTED]]

Sent: Wednesday, June 23, 2004 4:56 PM

To: Gurol Akman; [EMAIL PROTECTED]; linuxstreams

Subject: RE: [Linux-streams] LiS 2.17.P Clone driver major number change.

That goes into the "this would be nice to do sometime when I have unlimited free time" stack.

-- Dave

At 03:46 PM 6/23/2004, Gurol Akman wrote:

Is there a technical reason for using hard-coded major numbers in LiS drivers? Wouldn't it be more convenient if all LiS drivers supported dynamic major number assignment (i.e., based on use of reserved major number 0)? This would satisfy the needs of all, I'd think. Regards -- GA

-----Original Message-----

From: Dave Grothe [mailto:[EMAIL PROTECTED]]

Sent: Wednesday, June 23, 2004 4:40 PM

To: [EMAIL PROTECTED]; linuxstreams

Subject: Re: [Linux-streams] LiS 2.17.P Clone driver major number change.

It was merely an attempt to avoid collisions with kernel major device numbers.  Is this causing people pain?  I could change the default back to 220.

-- Dave

At 03:19 PM 6/23/2004, dan_gora wrote:

Hi Dave...

I noticed that the major number of /dev/clone/drvr has been changed in 2.17.P

from 220 to 231.  What was the reason for that change?  Linux 2.6?

BTW... LiS 2.17.P looks great.  It is already helping the performance of at

least one of our drivers (the one that I just tried).  I'm getting 5-10%

better throughput and 5-10% less CPU utilization without doing anything to my

driver!

Keep up the good work...

thanks-

dan

_______________________________________________

Linux-streams mailing list

[EMAIL PROTECTED]

http://gsyc.escet.urjc.es/mailman/listinfo/linux-streams