Re: Console unregistration questions

["Antonino A. Daplas" <adaplas@gmail.com>]

On Thu, 2007-04-26 at 18:30 -0700, Jesse Barnes wrote:
> On Tuesday, April 24, 2007, Antonino A. Daplas wrote:
> > Here's a test patch.  Basically, it checks if the driver is still mapped
> > to the console, and if it is, unregister_framebuffer() will exit
> > immediately.  As mentioned, it's untested, so let me know if there are
> > problems.
> >
> > Note, the driver must also do its part, by checking if the driver is
> > opened by something in userspace.
> 
> Here's a hacked up version of an unregistration patch.  It exports 
> unbind_con_driver from vt.c and makes fbcon and fbmem use it when a driver 
> goes away...
> 

I have no problems with the purpose of this patch, but I would like what
other kernel developers, both fbdev and non-fbdev, think about this.

The main problem with this version is it's possible to have more than 1
fb driver mapped to each console (and people do that).  So if one of
them unregisters, it will also unbind the other one. Code-wise, it's not
a problem.  But users will not expect that behavior.

It's fixable though, as unbind_con_driver() also accepts the first and
last parameter.  fbcon_fb_unbind() will have to walk con2fbmap[], so
something like this untested code instead:

static int fbcon_fb_unbind(int idx) {
	for (i = 0; i < MAX_NR_CONSOLES; i++) {
		/* Assure we do not unbind other drivers */
		if (idx == con2fbmap[i])
			/* can be optimize to minimize multiple calls to
			 	unbind_con_driver() */
			unbind_con_driver(&fb_con, i, i, 0); 
	}
]	

Besides the above, we still have to make adjustments to fbcon so it can
handle one driver going away and while still managing other loaded
drivers.

Tony


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/