PATCH to fix initialisation issue for GC3 (linux-2.5.64 +).

PATCH to fix initialisation issue for GC3 (linux-2.5.64 +).

[Mark Fortescue]

Hi All,

This patch fixes an error in the blanking code for the GCThree SBUS
video card. Now I get a logo and black screen, not just a blank (no
video) screen. It is a trivual fix that has taken too long to identify as
it is such a small typing error during the rewriting of the code for
the new frame buffer system introduced in the 2.5 series kernels.

Given that it still exists in the 2.6.8.1 kernel, I assume that not many
people have tried using the latest kernel on systems with a CGThree.
  
Regards
	Mark Fortescue.
------------------------------------------------------------------------
#######################################################################
#
# Mark Fortescue <mark@mtfhpc.demon.co.uk>
#
# Patch to fix the blanking on the CG3 (SBUS) video card.
# The file (formally cgthreefb.c, now cg3.c) has a typing error that
# messes up the card initialisation when blanking is enabled/disabled.
# Given that the blanking functions are called in the initialisation
# routine, this makes the card un-usable. This patch fixes the error
# and needs to be applied to all kernels from 2.5.64 to 2.8.6.1 and
# possibly later if the CG3 is to be used sucessfully. 
#
#######################################################################
diff -rupd ref-2.6.8.1/drivers/video/cg3.c linux-2.6.8.1/drivers/video/cg3.c
--- ref-2.6.8.1/drivers/video/cg3.c	Sat Aug 14 11:55:35 2004
+++ linux-2.6.8.1/drivers/video/cg3.c	Fri Oct 29 01:03:03 2004
@@ -198,9 +204,9 @@ cg3_blank(int blank, struct fb_info *inf
 
 	switch (blank) {
 	case 0: /* Unblanking */
-		val = sbus_readl(&regs->control);
+		val = sbus_readb(&regs->control);
 		val |= CG3_CR_ENABLE_VIDEO;
-		sbus_writel(val, &regs->control);
+		sbus_writeb(val, &regs->control);
 		par->flags &= ~CG3_FLAG_BLANKED;
 		break;
 
@@ -208,11 +214,16 @@ cg3_blank(int blank, struct fb_info *inf
 	case 2: /* VESA blank (vsync off) */
 	case 3: /* VESA blank (hsync off) */
 	case 4: /* Poweroff */
-		val = sbus_readl(&regs->control);
-		val |= CG3_CR_ENABLE_VIDEO;
-		sbus_writel(val, &regs->control);
+		val = sbus_readb(&regs->control);
+		val &= ~CG3_CR_ENABLE_VIDEO;
+		sbus_writeb(val, &regs->control);
 		par->flags |= CG3_FLAG_BLANKED;
 		break;
+
+	default:
+		printk ("Invalid Blanking mode (%d), unblanking\n", blank);
+		spin_unlock_irqrestore(&par->lock, flags);
+		return 1;
 	}
 
 	spin_unlock_irqrestore(&par->lock, flags);
--------------------------------------------------------------------------

PATCH to fix initialisation issue for GC3 (linux-2.5.64 +).

[Mark Fortescue]

Hi All,

This patch fixes an error in the blanking code for the GCThree SBUS
video card. Now I get a logo and black screen, not just a blank (no
video) screen. It is a trivual fix that has taken too long to identify as
it is such a small typing error during the rewriting of the code for
the new frame buffer system introduced in the 2.5 series kernels.

Given that it still exists in the 2.6.8.1 kernel, I assume that not many
people have tried using the latest kernel on systems with a CGThree.
  
Regards
	Mark Fortescue.
------------------------------------------------------------------------
#######################################################################
#
# Mark Fortescue <mark@mtfhpc.demon.co.uk>
#
# Patch to fix the blanking on the CG3 (SBUS) video card.
# The file (formally cgthreefb.c, now cg3.c) has a typing error that
# messes up the card initialisation when blanking is enabled/disabled.
# Given that the blanking functions are called in the initialisation
# routine, this makes the card un-usable. This patch fixes the error
# and needs to be applied to all kernels from 2.5.64 to 2.8.6.1 and
# possibly later if the CG3 is to be used sucessfully. 
#
#######################################################################
diff -rupd ref-2.6.8.1/drivers/video/cg3.c linux-2.6.8.1/drivers/video/cg3.c
--- ref-2.6.8.1/drivers/video/cg3.c	Sat Aug 14 11:55:35 2004
+++ linux-2.6.8.1/drivers/video/cg3.c	Fri Oct 29 01:03:03 2004
@@ -198,9 +204,9 @@ cg3_blank(int blank, struct fb_info *inf
 
 	switch (blank) {
 	case 0: /* Unblanking */
-		val = sbus_readl(&regs->control);
+		val = sbus_readb(&regs->control);
 		val |= CG3_CR_ENABLE_VIDEO;
-		sbus_writel(val, &regs->control);
+		sbus_writeb(val, &regs->control);
 		par->flags &= ~CG3_FLAG_BLANKED;
 		break;
 
@@ -208,11 +214,16 @@ cg3_blank(int blank, struct fb_info *inf
 	case 2: /* VESA blank (vsync off) */
 	case 3: /* VESA blank (hsync off) */
 	case 4: /* Poweroff */
-		val = sbus_readl(&regs->control);
-		val |= CG3_CR_ENABLE_VIDEO;
-		sbus_writel(val, &regs->control);
+		val = sbus_readb(&regs->control);
+		val &= ~CG3_CR_ENABLE_VIDEO;
+		sbus_writeb(val, &regs->control);
 		par->flags |= CG3_FLAG_BLANKED;
 		break;
+
+	default:
+		printk ("Invalid Blanking mode (%d), unblanking\n", blank);
+		spin_unlock_irqrestore(&par->lock, flags);
+		return 1;
 	}
 
 	spin_unlock_irqrestore(&par->lock, flags);
--------------------------------------------------------------------------

Re: PATCH to fix initialisation issue for GC3 (linux-2.5.64 +).

[William Lee Irwin III]

On Fri, Oct 29, 2004 at 01:27:51AM +0100, Mark Fortescue wrote:
> This patch fixes an error in the blanking code for the GCThree SBUS
> video card. Now I get a logo and black screen, not just a blank (no
> video) screen. It is a trivual fix that has taken too long to identify as
> it is such a small typing error during the rewriting of the code for
> the new frame buffer system introduced in the 2.5 series kernels.
> Given that it still exists in the 2.6.8.1 kernel, I assume that not many
> people have tried using the latest kernel on systems with a CGThree.

Yes, all my systems are headless unfortunately.


-- wli

Re: PATCH to fix initialisation issue for GC3 (linux-2.5.64 +).

[William Lee Irwin III]

On Fri, Oct 29, 2004 at 01:27:51AM +0100, Mark Fortescue wrote:
> This patch fixes an error in the blanking code for the GCThree SBUS
> video card. Now I get a logo and black screen, not just a blank (no
> video) screen. It is a trivual fix that has taken too long to identify as
> it is such a small typing error during the rewriting of the code for
> the new frame buffer system introduced in the 2.5 series kernels.
> Given that it still exists in the 2.6.8.1 kernel, I assume that not many
> people have tried using the latest kernel on systems with a CGThree.

Yes, all my systems are headless unfortunately.


-- wli

Re: PATCH to fix initialisation issue for GC3 (linux-2.5.64 +).

[Lee Revell]

On Thu, 2004-10-28 at 17:34 -0700, William Lee Irwin III wrote:
> Yes, all my systems are headless unfortunately.

*All* of them?  I am impressed...

Lee

Re: PATCH to fix initialisation issue for GC3 (linux-2.5.64 +).

[Lee Revell]

On Thu, 2004-10-28 at 17:34 -0700, William Lee Irwin III wrote:
> Yes, all my systems are headless unfortunately.

*All* of them?  I am impressed...

Lee

Re: PATCH to fix initialisation issue for GC3 (linux-2.5.64 +).

[William Lee Irwin III]

On Thu, 2004-10-28 at 17:34 -0700, William Lee Irwin III wrote:
>> Yes, all my systems are headless unfortunately.

On Thu, Oct 28, 2004 at 11:51:32PM -0400, Lee Revell wrote:
> *All* of them?  I am impressed...

All of the UltraSPARC and SPARC systems, yes.


-- wli

Re: PATCH to fix initialisation issue for GC3 (linux-2.5.64 +).

[William Lee Irwin III]

On Thu, 2004-10-28 at 17:34 -0700, William Lee Irwin III wrote:
>> Yes, all my systems are headless unfortunately.

On Thu, Oct 28, 2004 at 11:51:32PM -0400, Lee Revell wrote:
> *All* of them?  I am impressed...

All of the UltraSPARC and SPARC systems, yes.


-- wli

Re: PATCH to fix initialisation issue for GC3 (linux-2.5.64 +).

[Meelis Roos]

MF> This patch fixes an error in the blanking code for the GCThree SBUS
MF> video card. Now I get a logo and black screen, not just a blank (no
MF> video) screen. It is a trivual fix that has taken too long to identify as
MF> it is such a small typing error during the rewriting of the code for
MF> the new frame buffer system introduced in the 2.5 series kernels.

The "val &= ~CG3_CR_ENABLE_VIDEO;" bit is in main kernel now. But what
about the readl/writel vs readb/writeb - is this also needed? Why?

-- 
Meelis Roos

Re: PATCH to fix initialisation issue for GC3 (linux-2.5.64 +).

[David S. Miller]

On Wed, 16 Feb 2005 16:15:47 +0200
Meelis Roos <mroos@linux.ee> wrote:

> MF> This patch fixes an error in the blanking code for the GCThree SBUS
> MF> video card. Now I get a logo and black screen, not just a blank (no
> MF> video) screen. It is a trivual fix that has taken too long to identify as
> MF> it is such a small typing error during the rewriting of the code for
> MF> the new frame buffer system introduced in the 2.5 series kernels.
> 
> The "val &= ~CG3_CR_ENABLE_VIDEO;" bit is in main kernel now. But what
> about the readl/writel vs readb/writeb - is this also needed? Why?

The main kernel uses byte read/write operations in the cg3 driver.

Re: PATCH to fix initialisation issue for GC3 (linux-2.5.64 +).

[Mark Fortescue]

Hi Mellis,

The register masks are set up for an 8bit byte operation. If you use a
32bit read/write (readl/writel) you get very different results on a big
endian system (SPARC) when you apply a byte mask. This is why you need to
use 8bit (readb/writeb) operations.

In addition, using 8bit read/write will prevent corruption of any of the
next three registers if they have different read/write information.

Regards
	Mark Fortescue.

On Wed, 16 Feb 2005, Meelis Roos wrote:

> MF> This patch fixes an error in the blanking code for the GCThree SBUS
> MF> video card. Now I get a logo and black screen, not just a blank (no
> MF> video) screen. It is a trivual fix that has taken too long to identify as
> MF> it is such a small typing error during the rewriting of the code for
> MF> the new frame buffer system introduced in the 2.5 series kernels.
> 
> The "val &= ~CG3_CR_ENABLE_VIDEO;" bit is in main kernel now. But what
> about the readl/writel vs readb/writeb - is this also needed? Why?
> 
> -- 
> Meelis Roos
>