Re: [BUG]: agpgart for i810 chipsets broken in 2.5.51

[Dave Jones <davej@codemonkey.org.uk>]

On Tue, Dec 10, 2002 at 05:06:17PM +0000, Alan Cox wrote:
 > Given how fragile the AGP code can be I would much rather we had the AGP
 > continue to initialize late. If the AGP init function is something like
 > 
 > 
 > int agp_required(void)
 > {
 > 	static int agp_inited = 0;
 > 
 > 	if(!agp_inited)
 > 	{
 > 		agp_inited = 1;
 > 		agp_do_real_init();
 > 	}
 > }
 > 
 > module_init(agp_required);
 > 
 > 
 > Then the i810 fb driver can do
 > 
 > 	agp_required();
 > 
 > and force the order change only if necessary.

That works for me. It's not ideal, but it's the cleanest
solution suggested so far. I'll hack a check into
agp_init() to do this, which should allow us to close bug #20 [*]

        Dave

[*] http://bugzilla.kernel.org/show_bug.cgi?id=20