[okl4-developer] Is there any double buffering? S3C2410 lcd - FB

[Geoffrey Lee]

On Tue, Aug 26, 2008 at 05:45:34PM +0200, Dennis Gessner wrote:
> Hi,
> 

Hi Dennis

> first of all, thank you Geoffrey, your hint with
> "enable_fastpaths=False" solved the "clz" problems inside traps.spp. If
> I find some time, I'll rewrite (patch) the traps.spp without all the
> clz-instructions (which are not supported inside the ARMv4 architecture
> of the S3C2410).

That's great to hear.  :)

> 
> Now the kernel (okl4 2.1.1 patch 9) compiles and runs again on my real
> hardware (gta01). What surprises me, it's not working inside qemu
> (segmentation fault) - maybe there is something wrong with qemu...
> 
> Now my problem:
> 
> I am wondering if there is anything like double buffering inside the
> lcd-output (fb) of the device. I initialize the lcd correctly, I get an
> LCD-Mode returned (480x640 at 10).. I write
> values to the fb, but the display is still black.
> 

Does your code work on hardware?  It is possible that the emulation
in qemu is incomplete or incorrect and may produce strange
results.  A crash of qemu however definitely appears to be a qemu
bug.

	-gl

> I am missing some command like "fb_refresh()" or anything that tells
> okl4 that it should repaint the framebuffer with the new values.
> Otherwise I do not understand why my display still stays black.
> 
> Here is what I did:
> 
> ...
> static int init_fb(struct ig_fb * fb, struct ig_lcd * lcd) {
> 
>     struct lcd_control_block *ctrl = lcd->control;
>     fb->size = ctrl->xres * ctrl->yres * ctrl->bpp / 8;
> 
>     fb->ms = memsection_create_dma
> 	     (
> 	        fb->size,
> 	    	&fb->vbase,
> 	    	&fb->pm,
> 	    	L4_WriteThroughMemory
> 	      );
> 
>     virtual_lcd_set_fb(lcd->server, lcd->obj, fb->vbase, NULL);
> 
>     return 0;
> }
> 
> static int write_to_fb(struct ig_fb * fb, struct ig_lcd * lcd) {
> 
>     int x, y, i = 0;
>     char r, g, b;
> 
>     uintptr_t	fb_base;
>     struct lcd_control_block *ctrl = lcd->control;
> 
>     uint16_t col;
> 
>     virtual_lcd_get_fb(lcd->server, lcd->obj, &fb_base, NULL);
>     virtual_lcd_get_mode(lcd->server, lcd->obj, 				
> &ctrl->xres,&ctrl->yres,&ctrl->bpp, NULL);
> 
>     uint16_t *fbp = (uint16_t*) fb_base;
> 
>     printf("xx  LCD Mode is %ldx%ld@%lx\n", lcd->control->xres,
>            lcd->control->yres,lcd->control->bpp);
> 
>     printf("\nLeading 15 values of the fb: \n\n");
>     for (i=0; i < 15; i++) {
> 	printf("%d--",fbp[i]);
>     }
> 
>     i = 0;
> 	
>     printf("\n...fill the fb...\n");
>     for (x=0; x < ctrl->yres; x++) {
>         for (y = 0; y < (ctrl->xres); y++) {
> 
> 	    r = (x<y)?0:255;
> 	    g = (x<(ctrl->xres/2))?0:255;
> 	    b = (y<(ctrl->yres/2))?0:255;
> 
> 	    col = (r&0x1f)      |
> 	          ((g&0x3f)<<5) |
> 		  ((b&0x1f)<<11);
> 
>             /* I also tried to manually set all values to 1024 */
> // 	    fbp[i++] = 1024;
> 
> 	    fbp[i++] = col;
>         }
>     }
> 
>     printf("\nLeading 15 values of the fb: \n\n");
>     for (i=0; i < 15; i++) {
> 	printf("%d--",fbp[i]);
>     }	
>     printf("\n");
> 
>     /* I also tried the following, but this leads to a crash of 		
>  okl4
>     */
> //    fb_base = (uintptr_t) fbp;
> //    virtual_lcd_set_fb(lcd->server, lcd->obj, &fb_base, NULL);
> 
>     return 0;
> }
> 
> What am I doing wrong?
> 
> Best regards,
> Dennis
> 

--