On Wed, Nov 03, 2004 at 05:12:15AM +0800, Antonino A. Daplas wrote:

On Tuesday 02 November 2004 19:46, Petr Vandrovec wrote:

quoted

On  2 Nov 04 at 19:46, Antonino A. Daplas wrote:

quoted

diff -Nru a/drivers/video/matrox/matroxfb_base.c
b/drivers/video/matrox/matroxfb_base.c ---
a/drivers/video/matrox/matroxfb_base.c  2004-10-23 02:52:37 +08:00 +++
b/drivers/video/matrox/matroxfb_base.c  2004-11-01 08:32:54 +08:00 @@
-1877,16 +1876,21 @@
    }
    printk("fb%d: %s frame buffer device\n",
           ACCESS_FBINFO(fbcon.node), ACCESS_FBINFO(fbcon.fix.id));
-   if (ACCESS_FBINFO(fbcon.currcon) < 0) {
-       /* there is no console on this fb... but we have to initialize
hardware -        * until someone tells me what is proper thing to do */
-       printk(KERN_INFO "fb%d: initializing hardware\n",
-           ACCESS_FBINFO(fbcon.node));
-       /* We have to use FB_ACTIVATE_FORCE, as we had to put
vesafb_defined to the fbcon.var -        * already before, so
register_framebuffer works correctly. */ -       vesafb_defined.activate
|= FB_ACTIVATE_FORCE;
-       fb_set_var(&ACCESS_FBINFO(fbcon), &vesafb_defined);
-   }
+   /*
+    * Tony: If this driver is to be mapped to the console, then
+    *       fbcon will automatically do a set_par for us.  The code
below +    *       may not be needed.
+    */

Code below (which you made unconditional now) was executed if and only if
there is no fbcon attached to this device.  If there is fbcon attached
to this device, fbcon.currcon would be >= 0 after register_framebuffer()
(and fbcon's startup already did needed set_*).  If there is no fbcon,
this adapter is secondary, and it must be initialized too before leaving
driver.  But it must not be initialized before call to
register_framebuffer, as otherwise taking over vgacon won't work correctly.

You could put some 'ACCESS_FBINFO(flags.hw_initialized) = 1' somewhere
into matroxfb's set_par and then test this flag instead of fbcon.currcon.

But your current code probably does not work very well on primary adapter.

Why not just leave the hardware uninitialized after register_framebuffer? Then,
initialize it at the first call to xxxfb_open()?  We can even check a
flag so we do this only once. This way if people accidentally enabled
matroxfb but forgot to enable fbcon, they will still get a usable vga
console.

Because hardware driver is supposed to drive hardware.   Plus I do not agree
with paradigm that open should do something special unless each open provides
completely separate view (like opening /dev/ptmx).  Harddisk also does not
spin up when you open /dev/hda and does not stop when you close it.
					Petr Vandrovec



-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click