[PATCH v2 0/5] m68k SuperVidel patches

[PATCH v2 0/5] m68k SuperVidel patches

[Michael Schmitz]

Version 2 of Miro Kropacek's patches for Atari SuperVidel, as
posted to debian-68k.

Patch 1 is a bugfix for the Atari mouse input device driver
(miscompiled since v6.2 with use of -funsigned_char). Geert's
suggestion to change the mouse interrupt function signature
adopted (generates equivalent code as seen by inspection and
testing on my 030 Falcon).

Patches 2-4 are fbdev hardware support patches for the SuperVidel
blitter hardware.

Patch 5 is adding support to make use of the ethoc hardware also
in the SuperVidel. Platform device register error handling (reported
by Geert) fixed.

All patches tested by Miro on his CT60 with SuperVidel.

Cheers,

   Michael

[PATCH v2 1/5] m68k: Fix atari mouse movement

[Michael Schmitz]

From: Miro Kropacek <miro.kropacek@gmail.com>

Kernels are built with -funsigned-char since version 6.2, resulting
in signed mouse position deltas getting misinterpreted as unsigned.

Cast deltas passed in (unsigned) scancode buffer to correctly
interpret sign.

Fixes: 3bc753c06dd02a35 ("kbuild: treat char as always unsigned")
Link: https://lists.debian.org/debian-68k/2026/08/msg00000.html
Signed-off-by: Miro Kropacek <miro.kropacek@gmail.com>
Reviewed-by: Michael Schmitz <schmitzmic@gmail.com>
Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
Cc: linux-input <linux-input@vger.kernel.org>

---
Changes in  v2:

- change atamouse_interrupt function signature to signed char *,
  drop signed casts (Geert)
---
 arch/m68k/atari/atakeyb.c        | 2 +-
 arch/m68k/include/asm/atarikb.h  | 2 +-
 drivers/input/mouse/atarimouse.c | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/arch/m68k/atari/atakeyb.c b/arch/m68k/atari/atakeyb.c
index 49a9a459bdf4..86ab73a79177 100644
--- a/arch/m68k/atari/atakeyb.c
+++ b/arch/m68k/atari/atakeyb.c
@@ -39,7 +39,7 @@ void (*atari_MIDI_interrupt_hook) (void);
 /* Hook for keyboard inputdev  driver */
 void (*atari_input_keyboard_interrupt_hook) (unsigned char, char);
 /* Hook for mouse inputdev  driver */
-void (*atari_input_mouse_interrupt_hook) (char *);
+void (*atari_input_mouse_interrupt_hook)(signed char *buf);
 EXPORT_SYMBOL(atari_input_keyboard_interrupt_hook);
 EXPORT_SYMBOL(atari_input_mouse_interrupt_hook);
 
diff --git a/arch/m68k/include/asm/atarikb.h b/arch/m68k/include/asm/atarikb.h
index 68f3622bf591..c5a84c7ee85d 100644
--- a/arch/m68k/include/asm/atarikb.h
+++ b/arch/m68k/include/asm/atarikb.h
@@ -37,7 +37,7 @@ extern void (*atari_MIDI_interrupt_hook) (void);
 /* Hook for keyboard inputdev  driver */
 extern void (*atari_input_keyboard_interrupt_hook) (unsigned char, char);
 /* Hook for mouse inputdev  driver */
-extern void (*atari_input_mouse_interrupt_hook) (char *);
+extern void (*atari_input_mouse_interrupt_hook)(signed char *buf);
 
 int atari_keyb_init(void);
 
diff --git a/drivers/input/mouse/atarimouse.c b/drivers/input/mouse/atarimouse.c
index b1219cc4d9a2..6bac2288c494 100644
--- a/drivers/input/mouse/atarimouse.c
+++ b/drivers/input/mouse/atarimouse.c
@@ -60,7 +60,7 @@ extern int atari_mouse_buttons;
 
 static struct input_dev *atamouse_dev;
 
-static void atamouse_interrupt(char *buf)
+static void atamouse_interrupt(signed char *buf)
 {
 	int buttons, dx, dy;
 
-- 
2.17.1

Re: [PATCH v2 1/5] m68k: Fix atari mouse movement

[Geert Uytterhoeven]

On Sun, 16 Aug 2026 at 09:38, Michael Schmitz <schmitzmic@gmail.com> wrote:
> From: Miro Kropacek <miro.kropacek@gmail.com>
>
> Kernels are built with -funsigned-char since version 6.2, resulting
> in signed mouse position deltas getting misinterpreted as unsigned.
>
> Cast deltas passed in (unsigned) scancode buffer to correctly
> interpret sign.
>
> Fixes: 3bc753c06dd02a35 ("kbuild: treat char as always unsigned")
> Link: https://lists.debian.org/debian-68k/2026/08/msg00000.html
> Signed-off-by: Miro Kropacek <miro.kropacek@gmail.com>
> Reviewed-by: Michael Schmitz <schmitzmic@gmail.com>
> Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
> Cc: linux-input <linux-input@vger.kernel.org>
>
> ---
> Changes in  v2:
>
> - change atamouse_interrupt function signature to signed char *,
>   drop signed casts (Geert)

Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

[PATCH v2 2/5] fbdev: Give atafb proper parent

[Michael Schmitz]

From: Miro Kropacek <miro.kropacek@gmail.com>

The atafb fb device registers no parent, causing a missing symlink
(/sys/class/graphics/fb0/device). Xorg fbdevhw driver looks for that
symlink when scanning for devices, so add a parent node for atafb.

Signed-off-by: Miro Kropacek <miro.kropacek@gmail.com>
Reviewed-by: Michael Schmitz <schmitzmic@gmail.com>
Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
Cc: <linux-fbdev@vger.kernel.org>
Link: https://lists.debian.org/debian-68k/2026/08/msg00000.html
---
 drivers/video/fbdev/atafb.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/video/fbdev/atafb.c b/drivers/video/fbdev/atafb.c
index b8ed1c537293..3d540803dc2b 100644
--- a/drivers/video/fbdev/atafb.c
+++ b/drivers/video/fbdev/atafb.c
@@ -3101,6 +3101,11 @@ static int __init atafb_probe(struct platform_device *pdev)
 #endif /* ATAFB_EXT */
 
 //	strcpy(fb_info.mode->name, "Atari Builtin ");
+	/* Parent the fb device properly: without this fb0 registers as a
+	 * virtual sysfs device with no /sys/class/graphics/fb0/device link,
+	 * which makes Xorg's fbdevhw reject it ("No devices detected").
+	 */
+	fb_info.device = &pdev->dev;
 	fb_info.fbops = &atafb_ops;
 	// try to set default (detected; requested) var
 	do_fb_set_var(&atafb_predefined[default_par - 1], 1);
-- 
2.17.1

Re: [PATCH v2 2/5] fbdev: Give atafb proper parent

[Helge Deller]

On 8/16/26 09:37, Michael Schmitz wrote:
> From: Miro Kropacek <miro.kropacek@gmail.com>
> 
> The atafb fb device registers no parent, causing a missing symlink
> (/sys/class/graphics/fb0/device). Xorg fbdevhw driver looks for that
> symlink when scanning for devices, so add a parent node for atafb.
> 
> Signed-off-by: Miro Kropacek <miro.kropacek@gmail.com>
> Reviewed-by: Michael Schmitz <schmitzmic@gmail.com>
> Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
> Cc: <linux-fbdev@vger.kernel.org>
> Link: https://lists.debian.org/debian-68k/2026/08/msg00000.html

Did anything on the fbdev patches (#2-#4) changed between v1 and v2?

Helge
   

Re: [PATCH v2 2/5] fbdev: Give atafb proper parent

[Michael Schmitz]

Hi Helge,

no changes to the fbdev patches - just didn't know how else to avoid 
issues with b4.

Cheers,

	Michael


Am 17.08.2026 um 02:38 schrieb Helge Deller:
> On 8/16/26 09:37, Michael Schmitz wrote:
>> From: Miro Kropacek <miro.kropacek@gmail.com>
>>
>> The atafb fb device registers no parent, causing a missing symlink
>> (/sys/class/graphics/fb0/device). Xorg fbdevhw driver looks for that
>> symlink when scanning for devices, so add a parent node for atafb.
>>
>> Signed-off-by: Miro Kropacek <miro.kropacek@gmail.com>
>> Reviewed-by: Michael Schmitz <schmitzmic@gmail.com>
>> Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
>> Cc: <linux-fbdev@vger.kernel.org>
>> Link: https://lists.debian.org/debian-68k/2026/08/msg00000.html
>
> Did anything on the fbdev patches (#2-#4) changed between v1 and v2?
>
> Helge
>

Re: [PATCH v2 2/5] fbdev: Give atafb proper parent

[Thomas Zimmermann]

Hi

Am 16.08.26 um 09:37 schrieb Michael Schmitz:
> From: Miro Kropacek <miro.kropacek@gmail.com>
>
> The atafb fb device registers no parent, causing a missing symlink
> (/sys/class/graphics/fb0/device). Xorg fbdevhw driver looks for that
> symlink when scanning for devices, so add a parent node for atafb.

Do you really have these devices to test with?

If so, could you also replace the hard coded fb_info with a call to 
framebuffer_alloc()?

>
> Signed-off-by: Miro Kropacek <miro.kropacek@gmail.com>
> Reviewed-by: Michael Schmitz <schmitzmic@gmail.com>
> Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
> Cc: <linux-fbdev@vger.kernel.org>
> Link: https://lists.debian.org/debian-68k/2026/08/msg00000.html
> ---
>   drivers/video/fbdev/atafb.c | 5 +++++
>   1 file changed, 5 insertions(+)
>
> diff --git a/drivers/video/fbdev/atafb.c b/drivers/video/fbdev/atafb.c
> index b8ed1c537293..3d540803dc2b 100644
> --- a/drivers/video/fbdev/atafb.c
> +++ b/drivers/video/fbdev/atafb.c
> @@ -3101,6 +3101,11 @@ static int __init atafb_probe(struct platform_device *pdev)
>   #endif /* ATAFB_EXT */
>   
>   //	strcpy(fb_info.mode->name, "Atari Builtin ");
> +	/* Parent the fb device properly: without this fb0 registers as a
> +	 * virtual sysfs device with no /sys/class/graphics/fb0/device link,
> +	 * which makes Xorg's fbdevhw reject it ("No devices detected").
> +	 */
> +	fb_info.device = &pdev->dev;

Because this is supposed to happen in framebuffer_alloc() and keeping it 
there would be preferable.

Best regards
Thomas

>   	fb_info.fbops = &atafb_ops;
>   	// try to set default (detected; requested) var
>   	do_fb_set_var(&atafb_predefined[default_par - 1], 1);

-- 
--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Frankenstr. 146, 90461 Nürnberg, Germany, www.suse.com
GF: Jochen Jaser, Andrew McDonald, (HRB 36809, AG Nürnberg)

Re: [PATCH v2 2/5] fbdev: Give atafb proper parent

[Michael Schmitz]

Hi Thomas,

Am 17.08.2026 um 18:34 schrieb Thomas Zimmermann:
> Hi
>
> Am 16.08.26 um 09:37 schrieb Michael Schmitz:
>> From: Miro Kropacek <miro.kropacek@gmail.com>
>>
>> The atafb fb device registers no parent, causing a missing symlink
>> (/sys/class/graphics/fb0/device). Xorg fbdevhw driver looks for that
>> symlink when scanning for devices, so add a parent node for atafb.
>
> Do you really have these devices to test with?

Not the particular hardware that patches 3 and 4 add support for, but 
the 'normal' Videl that is found in every Falcon, yes.

> If so, could you also replace the hard coded fb_info with a call to
> framebuffer_alloc()?

Can't see how that would work - fb_info isn't passed to all functions 
that need to use it, especially in the hardware detection phase (hence 
the use of a static fb_info struct)?

Cheers,

	Michael

>>
>> Signed-off-by: Miro Kropacek <miro.kropacek@gmail.com>
>> Reviewed-by: Michael Schmitz <schmitzmic@gmail.com>
>> Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
>> Cc: <linux-fbdev@vger.kernel.org>
>> Link: https://lists.debian.org/debian-68k/2026/08/msg00000.html
>> ---
>>   drivers/video/fbdev/atafb.c | 5 +++++
>>   1 file changed, 5 insertions(+)
>>
>> diff --git a/drivers/video/fbdev/atafb.c b/drivers/video/fbdev/atafb.c
>> index b8ed1c537293..3d540803dc2b 100644
>> --- a/drivers/video/fbdev/atafb.c
>> +++ b/drivers/video/fbdev/atafb.c
>> @@ -3101,6 +3101,11 @@ static int __init atafb_probe(struct
>> platform_device *pdev)
>>   #endif /* ATAFB_EXT */
>>     //    strcpy(fb_info.mode->name, "Atari Builtin ");
>> +    /* Parent the fb device properly: without this fb0 registers as a
>> +     * virtual sysfs device with no /sys/class/graphics/fb0/device link,
>> +     * which makes Xorg's fbdevhw reject it ("No devices detected").
>> +     */
>> +    fb_info.device = &pdev->dev;
>
> Because this is supposed to happen in framebuffer_alloc() and keeping it
> there would be preferable.
>
> Best regards
> Thomas
>
>>       fb_info.fbops = &atafb_ops;
>>       // try to set default (detected; requested) var
>>       do_fb_set_var(&atafb_predefined[default_par - 1], 1);
>

Re: [PATCH v2 2/5] fbdev: Give atafb proper parent

[Thomas Zimmermann]

Am 17.08.26 um 09:49 schrieb Michael Schmitz:
> Hi Thomas,
>
> Am 17.08.2026 um 18:34 schrieb Thomas Zimmermann:
>> Hi
>>
>> Am 16.08.26 um 09:37 schrieb Michael Schmitz:
>>> From: Miro Kropacek <miro.kropacek@gmail.com>
>>>
>>> The atafb fb device registers no parent, causing a missing symlink
>>> (/sys/class/graphics/fb0/device). Xorg fbdevhw driver looks for that
>>> symlink when scanning for devices, so add a parent node for atafb.
>>
>> Do you really have these devices to test with?
>
> Not the particular hardware that patches 3 and 4 add support for, but 
> the 'normal' Videl that is found in every Falcon, yes.
>
>> If so, could you also replace the hard coded fb_info with a call to
>> framebuffer_alloc()?
>
> Can't see how that would work - fb_info isn't passed to all functions 
> that need to use it, especially in the hardware detection phase (hence 
> the use of a static fb_info struct)?

You'd have to modify the driver to pass the allocated instance, of 
course. It's not a trivial oneliner, but if this driver is still 
relevant, it might be worth improving it.

Best regards
Thomas

>
> Cheers,
>
>     Michael
>
>>>
>>> Signed-off-by: Miro Kropacek <miro.kropacek@gmail.com>
>>> Reviewed-by: Michael Schmitz <schmitzmic@gmail.com>
>>> Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
>>> Cc: <linux-fbdev@vger.kernel.org>
>>> Link: https://lists.debian.org/debian-68k/2026/08/msg00000.html
>>> ---
>>>   drivers/video/fbdev/atafb.c | 5 +++++
>>>   1 file changed, 5 insertions(+)
>>>
>>> diff --git a/drivers/video/fbdev/atafb.c b/drivers/video/fbdev/atafb.c
>>> index b8ed1c537293..3d540803dc2b 100644
>>> --- a/drivers/video/fbdev/atafb.c
>>> +++ b/drivers/video/fbdev/atafb.c
>>> @@ -3101,6 +3101,11 @@ static int __init atafb_probe(struct
>>> platform_device *pdev)
>>>   #endif /* ATAFB_EXT */
>>>     //    strcpy(fb_info.mode->name, "Atari Builtin ");
>>> +    /* Parent the fb device properly: without this fb0 registers as a
>>> +     * virtual sysfs device with no /sys/class/graphics/fb0/device 
>>> link,
>>> +     * which makes Xorg's fbdevhw reject it ("No devices detected").
>>> +     */
>>> +    fb_info.device = &pdev->dev;
>>
>> Because this is supposed to happen in framebuffer_alloc() and keeping it
>> there would be preferable.
>>
>> Best regards
>> Thomas
>>
>>>       fb_info.fbops = &atafb_ops;
>>>       // try to set default (detected; requested) var
>>>       do_fb_set_var(&atafb_predefined[default_par - 1], 1);
>>
>

-- 
--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Frankenstr. 146, 90461 Nürnberg, Germany, www.suse.com
GF: Jochen Jaser, Andrew McDonald, (HRB 36809, AG Nürnberg)

[PATCH v2 3/5] fbdev: Add support for further video bit depths on atafb:external

[Michael Schmitz]

From: Miro Kropacek <miro.kropacek@gmail.com>

Supervidel offers additional video bit depths (8-bit chunky, 16-bit
RGB565 (also on the originaln Videl) and ARGB888.

Add code to support these bit depths.

Signed-off-by: Miro Kropacek <miro.kropacek@gmail.com>
Reviewed-by: Michael Schmitz <schmitzmic@gmail.com>
Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
Cc: <linux-fbdev@vger.kernel.org>
Link: https://lists.debian.org/debian-68k/2026/08/msg00000.html
---
 drivers/video/fbdev/atafb.c | 89 ++++++++++++++++++++++++++++++++++---
 1 file changed, 82 insertions(+), 7 deletions(-)

diff --git a/drivers/video/fbdev/atafb.c b/drivers/video/fbdev/atafb.c
index 3d540803dc2b..50853d7a08e3 100644
--- a/drivers/video/fbdev/atafb.c
+++ b/drivers/video/fbdev/atafb.c
@@ -2081,8 +2081,12 @@ static int ext_encode_fix(struct fb_fix_screeninfo *fix, struct atafb_par *par)
 			 external_pmode == FB_TYPE_PACKED_PIXELS) ?
 				FB_VISUAL_MONO10 : FB_VISUAL_MONO01;
 	} else {
-		/* Use STATIC if we don't know how to access color registers */
-		int visual = external_vgaiobase ?
+		/* Use STATIC if we don't know how to access color registers;
+		 * SuperVidel 8bpp chunky (fb in SV RAM) uses the Falcon palette
+		 */
+		int visual = (external_vgaiobase ||
+			      (external_depth == 8 &&
+			       external_addr >= 0xa0000000)) ?
 					 FB_VISUAL_PSEUDOCOLOR :
 					 FB_VISUAL_STATIC_PSEUDOCOLOR;
 		switch (external_pmode) {
@@ -2159,6 +2163,35 @@ static int ext_encode_var(struct fb_var_screeninfo *var, struct atafb_par *par)
 	var->transp.offset = 0;
 	var->transp.length = 0;
 	var->transp.msb_right = 0;
+	if (external_pmode == -1 && external_depth == 16) {
+		/* RGB565 truecolor (e.g. SuperVidel native mode) */
+		var->red.offset = 11;
+		var->red.length = 5;
+		var->green.offset = 5;
+		var->green.length = 6;
+		var->blue.offset = 0;
+		var->blue.length = 5;
+	} else if (external_pmode == -1 && external_depth == 32) {
+		/* ARGB8888 truecolor (e.g. SuperVidel native mode) */
+		var->red.offset = 16;
+		var->red.length = 8;
+		var->green.offset = 8;
+		var->green.length = 8;
+		var->blue.offset = 0;
+		var->blue.length = 8;
+		var->transp.offset = 24;
+		var->transp.length = 8;
+	} else if (external_pmode == FB_TYPE_PACKED_PIXELS &&
+		   external_depth == 8 && external_addr >= 0xa0000000) {
+		/* SuperVidel 8bpp chunky: palette has 8 bits per channel.
+		 * Without this, fb_get_color_depth() sees length 0 and
+		 * fbcon falls back to its 2-color palette — the console
+		 * text (color 7) stays black on black.
+		 */
+		var->red.length = 8;
+		var->green.length = 8;
+		var->blue.length = 8;
+	}
 	var->yres_virtual = var->yres;
 	var->xoffset = 0;
 	var->yoffset = 0;
@@ -2193,6 +2226,38 @@ static int ext_setcolreg(unsigned int regno, unsigned int red,
 {
 	unsigned char colmask = (1 << external_bitspercol) - 1;
 
+	if (external_pmode == -1 && external_depth == 16) {
+		/* truecolor: only the pseudo palette for fbcon is needed */
+		if (regno > 15)
+			return 1;
+		((u32 *)info->pseudo_palette)[regno] = (red & 0xf800) |
+						       ((green & 0xfc00) >> 5) |
+						       ((blue & 0xf800) >> 11);
+		return 0;
+	}
+	if (external_pmode == -1 && external_depth == 32) {
+		/* ARGB8888, alpha forced opaque */
+		if (regno > 15)
+			return 1;
+		((u32 *)info->pseudo_palette)[regno] = 0xff000000 |
+						       ((red & 0xff00) << 8) |
+						       (green & 0xff00) |
+						       ((blue & 0xff00) >> 8);
+		return 0;
+	}
+	if (external_pmode == FB_TYPE_PACKED_PIXELS && external_depth == 8 &&
+	    external_addr >= 0xa0000000) {
+		/* SuperVidel native 8bpp chunky scans out via the Falcon
+		 * palette registers, honoring all 8 bits per channel
+		 */
+		if (regno > 255)
+			return 1;
+		f030_col[regno] = ((red & 0xff00) << 16) |
+				  ((green & 0xff00) << 8) |
+				  ((blue & 0xff00) >> 8);
+		return 0;
+	}
+
 	if (!external_vgaiobase)
 		return 1;
 
@@ -2422,7 +2487,9 @@ static void atafb_fillrect(struct fb_info *info, const struct fb_fillrect *rect)
 		return;
 
 #ifdef ATAFB_FALCON
-	if (info->var.bits_per_pixel == 16) {
+	/* chunky modes (Falcon hicolor, external packed/truecolor) */
+	if (info->fix.type == FB_TYPE_PACKED_PIXELS &&
+	    info->var.bits_per_pixel > 1) {
 		cfb_fillrect(info, rect);
 		return;
 	}
@@ -2463,7 +2530,9 @@ static void atafb_copyarea(struct fb_info *info, const struct fb_copyarea *area)
 	int rev_copy = 0;
 
 #ifdef ATAFB_FALCON
-	if (info->var.bits_per_pixel == 16) {
+	/* chunky modes (Falcon hicolor, external packed/truecolor) */
+	if (info->fix.type == FB_TYPE_PACKED_PIXELS &&
+	    info->var.bits_per_pixel > 1) {
 		cfb_copyarea(info, area);
 		return;
 	}
@@ -2517,7 +2586,9 @@ static void atafb_imageblit(struct fb_info *info, const struct fb_image *image)
 	u32 dx, dy, width, height, pitch;
 
 #ifdef ATAFB_FALCON
-	if (info->var.bits_per_pixel == 16) {
+	/* chunky modes (Falcon hicolor, external packed/truecolor) */
+	if (info->fix.type == FB_TYPE_PACKED_PIXELS &&
+	    info->var.bits_per_pixel > 1) {
 		cfb_imageblit(info, image);
 		return;
 	}
@@ -2753,7 +2824,7 @@ static void __init atafb_setup_ext(char *spec)
 		return;
 	depth = simple_strtoul(p, NULL, 10);
 	if (depth != 1 && depth != 2 && depth != 4 && depth != 8 &&
-	    depth != 16 && depth != 24)
+	    depth != 16 && depth != 24 && depth != 32)
 		return;
 
 	p = strsep(&spec, ";");
@@ -3131,7 +3202,11 @@ static int __init atafb_probe(struct platform_device *pdev)
 
 	atafb_set_disp(&fb_info);
 
-	fb_alloc_cmap(&(fb_info.cmap), 1 << fb_info.var.bits_per_pixel, 0);
+	/* truecolor visuals only need the 16-entry console palette; this
+	 * also avoids 1 << 32 overflowing at 32bpp
+	 */
+	fb_alloc_cmap(&(fb_info.cmap), fb_info.var.bits_per_pixel > 8 ?
+		      16 : 1 << fb_info.var.bits_per_pixel, 0);
 
 
 	dev_info(&pdev->dev, "Determined %dx%d, depth %d\n", fb_info.var.xres,
-- 
2.17.1

[PATCH v2 4/5] fbdev: Add support for SuperVidel's SuperBlitter

[Michael Schmitz]

From: Miro Kropacek <miro.kropacek@gmail.com>

The SuperVidel graphics FPGA includes a hardware blitter (bit block
transfer engine) operating within the SuperVidel DDR2 video RAM.

Two versions of this blitter exist. Later versions (>= 9) of the
SuperVidel firmware support an asynchronous command FIFO, older
versions must be polled for command completion.

Add hardware-accelerated copyarea, fillrect and imageblit fb
operations (falling back to the non-accelerated versions for
anything that exceeds blitter capabilities).

Signed-off-by: Miro Kropacek <miro.kropacek@gmail.com>
Reviewed-by: Michael Schmitz <schmitzmic@gmail.com>
Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
Cc: <linux-fbdev@vger.kernel.org>
Link: https://lists.debian.org/debian-68k/2026/08/msg00000.html
---
 drivers/video/fbdev/atafb.c | 197 ++++++++++++++++++++++++++++++++++++
 1 file changed, 197 insertions(+)

diff --git a/drivers/video/fbdev/atafb.c b/drivers/video/fbdev/atafb.c
index 50853d7a08e3..dc2d4154e1b6 100644
--- a/drivers/video/fbdev/atafb.c
+++ b/drivers/video/fbdev/atafb.c
@@ -2304,6 +2304,185 @@ static int ext_detect(void)
 	return 1;
 }
 
+/* ------------------- SuperVidel SuperBlitter ---------------------- */
+
+/*
+ * Hardware blitter in the SuperVidel FPGA, operating within SV DDR2 RAM.
+ * FW revision >= 9 provides a command FIFO (async operation); older
+ * revisions are programmed directly with busy-polling.
+ */
+#define SVBLIT_REGS_PHYS	0x80010000
+#define SVBLIT_SRC1		0x58	/* bits 26:0 */
+#define SVBLIT_SRC2		0x5c
+#define SVBLIT_DST		0x60
+#define SVBLIT_COUNT		0x64	/* bytes per line - 1 */
+#define SVBLIT_SRC1_OFFSET	0x68	/* line start to next line start */
+#define SVBLIT_SRC2_OFFSET	0x6c
+#define SVBLIT_DST_OFFSET	0x70
+#define SVBLIT_MASK_AND_LINES	0x74	/* bits 11:0: number of lines */
+#define SVBLIT_CONTROL		0x78	/* bit 0: busy/start, bits 4:1: mode */
+#define SVBLIT_VERSION		0x7c	/* bits 9:0: FW revision */
+#define SVBLIT_FIFO		0x80	/* wr: data; rd: bit 0 empty, bit 1 full */
+
+/*
+ * SuperBlitter bug: Instead of declared 2048 bytes, 2032 is the real maximum.
+ */
+#define SVBLIT_MAX_SPAN	2032
+
+static void __iomem *svblit_regs;
+static int svblit_fw;
+
+static inline u32 svblit_rd(unsigned int reg)
+{
+	return __raw_readl(svblit_regs + reg);
+}
+
+static inline void svblit_wr(unsigned int reg, u32 val)
+{
+	__raw_writel(val, svblit_regs + reg);
+}
+
+/* wait until all queued blits have finished */
+static void svblit_wait(void)
+{
+	if (svblit_fw >= 9)
+		/* FIFO empty flag = fewer than 9 longwords queued */
+		while (!(svblit_rd(SVBLIT_FIFO) & 1))
+			cpu_relax();
+	while (svblit_rd(SVBLIT_CONTROL) & 1)
+		cpu_relax();
+}
+
+/*
+ * FW >= 9 queues commands through the 512-longword FIFO: a command is
+ * 9 longwords (registers 0x58..0x78 in order), executed whenever >= 9
+ * words are queued and the blitter is idle. The full flag rises at
+ * >= 500 queued words, so below it there is always room for a whole
+ * command — one flag check per command prevents overflow (dropped
+ * words would desync the 9-word framing until an SV reinit, which is
+ * exactly what overflowing did before this guard existed). Older FW
+ * is programmed directly with busy-polling.
+ *
+ * The line byte count field is 11 bits but see SVBLIT_MAX_SPAN.
+ */
+static void svblit_copy(u32 src, u32 dst, u32 nbytes, u32 src_offset,
+			u32 dst_offset, u32 lines)
+{
+	while (nbytes) {
+		u32 chunk = min(nbytes, SVBLIT_MAX_SPAN);
+
+		if (svblit_fw >= 9) {
+			while (svblit_rd(SVBLIT_FIFO) & 2)
+				cpu_relax();
+			svblit_wr(SVBLIT_FIFO, src);
+			svblit_wr(SVBLIT_FIFO, 0);
+			svblit_wr(SVBLIT_FIFO, dst);
+			svblit_wr(SVBLIT_FIFO, chunk - 1);
+			svblit_wr(SVBLIT_FIFO, src_offset);
+			svblit_wr(SVBLIT_FIFO, 0);
+			svblit_wr(SVBLIT_FIFO, dst_offset);
+			svblit_wr(SVBLIT_FIFO, lines);
+			svblit_wr(SVBLIT_FIFO, 0x01);
+		} else {
+			while (svblit_rd(SVBLIT_CONTROL) & 1)
+				cpu_relax();
+			svblit_wr(SVBLIT_SRC1, src);
+			svblit_wr(SVBLIT_SRC2, 0);
+			svblit_wr(SVBLIT_DST, dst);
+			svblit_wr(SVBLIT_COUNT, chunk - 1);
+			svblit_wr(SVBLIT_SRC1_OFFSET, src_offset);
+			svblit_wr(SVBLIT_SRC2_OFFSET, 0);
+			svblit_wr(SVBLIT_DST_OFFSET, dst_offset);
+			svblit_wr(SVBLIT_MASK_AND_LINES, lines);
+			svblit_wr(SVBLIT_CONTROL, 0x01);
+		}
+
+		src += chunk;
+		dst += chunk;
+		nbytes -= chunk;
+	}
+}
+
+static int svblit_sync(struct fb_info *info)
+{
+	svblit_wait();
+	return 0;
+}
+
+static void svblit_copyarea(struct fb_info *info,
+			    const struct fb_copyarea *area)
+{
+	u32 bytespp = info->var.bits_per_pixel / 8;
+	u32 pitch = info->fix.line_length;
+
+	/*
+	 * The blitter walks lines in ascending order, so overlapping
+	 * moves down/right would read already overwritten data. Those
+	 * are rare for fbcon (scrolling backwards); leave them and
+	 * oversized areas to the CPU.
+	 */
+	if (area->height > 4095 ||
+	    area->dy > area->sy ||
+	    (area->dy == area->sy && area->dx > area->sx)) {
+		svblit_wait();
+		cfb_copyarea(info, area);
+		return;
+	}
+
+	svblit_copy(external_addr + area->sy * pitch + area->sx * bytespp,
+		    external_addr + area->dy * pitch + area->dx * bytespp,
+		    area->width * bytespp, pitch, pitch, area->height);
+	/* async: every CPU access to the fb goes through svblit_wait() */
+}
+
+static void svblit_fillrect(struct fb_info *info,
+			    const struct fb_fillrect *rect)
+{
+	u32 bytespp = info->var.bits_per_pixel / 8;
+	u32 pitch = info->fix.line_length;
+	u8 *line;
+	u32 pix;
+
+	svblit_wait();		/* the CPU is about to touch the fb */
+
+	if (rect->rop != ROP_COPY || rect->height <= 1 ||
+	    rect->height > 4096) {
+		cfb_fillrect(info, rect);
+		return;
+	}
+
+	pix = (info->fix.visual == FB_VISUAL_TRUECOLOR) ?
+		((u32 *)info->pseudo_palette)[rect->color] : rect->color;
+
+	/* draw the first line with the CPU ... */
+	line = (u8 *)info->screen_base + rect->dy * pitch +
+	       rect->dx * bytespp;
+	switch (bytespp) {
+	case 1:
+		memset(line, pix, rect->width);
+		break;
+	case 2:
+		memset16((u16 *)line, pix, rect->width);
+		break;
+	default:
+		memset32((u32 *)line, pix, rect->width);
+		break;
+	}
+
+	/* ... and let the blitter replicate it into the other lines */
+	svblit_copy(external_addr + rect->dy * pitch + rect->dx * bytespp,
+		    external_addr + (rect->dy + 1) * pitch +
+		    rect->dx * bytespp,
+		    rect->width * bytespp, 0, pitch, rect->height - 1);
+}
+
+static void svblit_imageblit(struct fb_info *info,
+			     const struct fb_image *image)
+{
+	svblit_wait();		/* CPU rendering must not race queued blits */
+	cfb_imageblit(info, image);
+}
+
 #endif /* ATAFB_EXT */
 
 /* ------ This is the same for most hardware types -------- */
@@ -3168,6 +3347,24 @@ static int __init atafb_probe(struct platform_device *pdev)
 		phys_screen_base = external_addr;
 		screen_len = external_len & PAGE_MASK;
 		memset (screen_base, 0, external_len);
+
+		/* framebuffer in SV RAM: enable the SuperBlitter */
+		if (external_addr >= 0xa0000000) {
+			svblit_regs = ioremap(SVBLIT_REGS_PHYS, 0x100);
+			if (svblit_regs) {
+				svblit_fw = svblit_rd(SVBLIT_VERSION) & 0x1ff;
+				atafb_ops.fb_fillrect = svblit_fillrect;
+				atafb_ops.fb_copyarea = svblit_copyarea;
+				atafb_ops.fb_imageblit = svblit_imageblit;
+				atafb_ops.fb_sync = svblit_sync;
+				fb_info.flags |= FBINFO_HWACCEL_COPYAREA |
+						 FBINFO_HWACCEL_FILLRECT;
+				dev_info(&pdev->dev,
+					 "SuperBlitter enabled, FW revision %d (%s)\n",
+					 svblit_fw, svblit_fw >= 9 ?
+					 "async FIFO" : "sync");
+			}
+		}
 	}
 #endif /* ATAFB_EXT */
 
-- 
2.17.1

[PATCH v2 5/5] m68k: Add support for Svethlana

[Michael Schmitz]

From: Miro Kropacek <miro.kropacek@gmail.com>

The SuperVidel FPGA contains an OpenCore ethernet package. Add
platform device data and other necessary definitions to support
this device on Atari.

The ethoc module must be enabled separately by the user.

Signed-off-by: Miro Kropacek <miro.kropacek@gmail.com>
Reviewed-by: Michael Schmitz <schmitzmic@gmail.com>
Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
Link: https://lists.debian.org/debian-68k/2026/08/msg00000.html

---
Changes in v2:

- fix svethlana platform register error handling (Geert)
---
 arch/m68k/Kconfig.devices         | 12 +++++
 arch/m68k/atari/config.c          | 74 +++++++++++++++++++++++++++++++
 arch/m68k/include/asm/atariints.h |  2 +-
 arch/m68k/include/asm/irq.h       |  6 +--
 4 files changed, 90 insertions(+), 4 deletions(-)

diff --git a/arch/m68k/Kconfig.devices b/arch/m68k/Kconfig.devices
index e6e3efac1840..48f964d1615e 100644
--- a/arch/m68k/Kconfig.devices
+++ b/arch/m68k/Kconfig.devices
@@ -67,6 +67,18 @@ config ATARI_ETHERNAT
 	  To compile the actual ethernet driver, choose Y or M for the SMC91X
 	  option in the network device section; the module will be called smc91x.
 
+config ATARI_SVETHLANA
+	bool "Atari SuperVidel SVEthlana Ethernet support"
+	depends on ATARI
+	help
+	  Say Y to include support for the SVEthlana network adapter built
+	  into the SuperVidel FPGA. SuperVidel firmware version 10 or newer
+	  is required (older firmware lacks the Ethernet DMA).
+
+	  To compile the actual ethernet driver, choose Y or M for the ETHOC
+	  option in the network device section; the module will be called
+	  ethoc.
+
 config ATARI_ETHERNEC
 	bool "Atari EtherNEC Ethernet support"
 	depends on ATARI_ROM_ISA
diff --git a/arch/m68k/atari/config.c b/arch/m68k/atari/config.c
index ee2d061efb2a..84317e0f447a 100644
--- a/arch/m68k/atari/config.c
+++ b/arch/m68k/atari/config.c
@@ -34,6 +34,7 @@
 #include <linux/platform_device.h>
 #include <linux/usb/isp116x.h>
 #include <linux/module.h>
+#include <net/ethoc.h>
 
 #include <asm/bootinfo.h>
 #include <asm/bootinfo-atari.h>
@@ -748,6 +749,62 @@ static struct platform_device *atari_ethernat_devices[] __initdata = {
 };
 #endif /* CONFIG_ATARI_ETHERNAT */
 
+#ifdef CONFIG_ATARI_SVETHLANA
+/*
+ * SVEthlana: OpenCores 10/100 Mbps Ethernet MAC in the SuperVidel FPGA,
+ * handled by the ethoc driver. Requires SuperVidel firmware version 10
+ * or newer (Ethernet DMA).
+ *
+ * The MAC can only DMA within the SuperVidel DDR RAM, so the packet
+ * buffers are carved out of the top of the DDR; the framebuffer lives
+ * at the bottom.
+ */
+
+#define ATARI_SVETHLANA_PHYS_ADDR	0x80012000
+#define ATARI_SVETHLANA_IRQ		141	/* vector 0xc5 */
+
+#define ATARI_SV_VERSION_PHYS_ADDR	0x8001007c
+#define ATARI_SV_DDR_END		0xa8000000
+/* 128 packet buffers of 1536 bytes each, the maximum ethoc supports */
+#define SVETHLANA_BUF_SIZE		(128 * 1536)
+
+static struct resource svethlana_resources[] = {
+	[0] = {
+		.name	= "ethoc-regs",
+		.start	= ATARI_SVETHLANA_PHYS_ADDR,
+		.end	= ATARI_SVETHLANA_PHYS_ADDR + 0x7ff,
+		.flags	= IORESOURCE_MEM,
+	},
+	[1] = {
+		.name	= "ethoc-buf",
+		.start	= ATARI_SV_DDR_END - SVETHLANA_BUF_SIZE,
+		.end	= ATARI_SV_DDR_END - 1,
+		.flags	= IORESOURCE_MEM,
+	},
+	[2] = {
+		.name	= "ethoc-irq",
+		.start	= ATARI_SVETHLANA_IRQ,
+		.end	= ATARI_SVETHLANA_IRQ,
+		.flags	= IORESOURCE_IRQ,
+	},
+};
+
+static struct ethoc_platform_data svethlana_platform_data = {
+	.phy_id		= -1,
+	.big_endian	= true,
+};
+
+static struct platform_device svethlana_device = {
+	.name		= "ethoc",
+	.id		= -1,
+	.num_resources	= ARRAY_SIZE(svethlana_resources),
+	.resource	= svethlana_resources,
+	.dev		= {
+		.platform_data	= &svethlana_platform_data,
+	},
+};
+#endif /* CONFIG_ATARI_SVETHLANA */
+
 #ifdef CONFIG_ATARI_ETHERNEC
 /*
  * EtherNEC: RTL8019 (NE2000 compatible) Ethernet chipset,
@@ -892,6 +949,23 @@ static int __init atari_platform_init(void)
 	}
 #endif
 
+#ifdef CONFIG_ATARI_SVETHLANA
+	{
+		int error;
+		void __iomem *sv_version;
+
+		sv_version = ioremap(ATARI_SV_VERSION_PHYS_ADDR, 4);
+		if (sv_version) {
+			if (hwreg_present(sv_version) &&
+			    (__raw_readl(sv_version) & 0x3ff) >= 10)
+				error = platform_device_register(&svethlana_device);
+				if (error && !rv)
+					rv = error;
+			iounmap(sv_version);
+		}
+	}
+#endif
+
 #ifdef CONFIG_ATARI_ETHERNEC
 	{
 		int error;
diff --git a/arch/m68k/include/asm/atariints.h b/arch/m68k/include/asm/atariints.h
index 6321c4495620..cb15361d14e9 100644
--- a/arch/m68k/include/asm/atariints.h
+++ b/arch/m68k/include/asm/atariints.h
@@ -32,7 +32,7 @@
 #define VME_SOURCE_BASE    56
 #define VME_MAX_SOURCES    16
 
-#define NUM_ATARI_SOURCES  141
+#define NUM_ATARI_SOURCES  142
 
 /* convert vector number to int source number */
 #define IRQ_VECTOR_TO_SOURCE(v)	((v) - ((v) < 0x20 ? 0x18 : (0x40-8)))
diff --git a/arch/m68k/include/asm/irq.h b/arch/m68k/include/asm/irq.h
index 2263e92d418a..8bac43d0f49d 100644
--- a/arch/m68k/include/asm/irq.h
+++ b/arch/m68k/include/asm/irq.h
@@ -10,8 +10,8 @@
  * different m68k hosts compiled into the kernel.
  * Currently the Atari has 72 and the Amiga 24, but if both are
  * supported in the kernel it is better to make room for 72.
- * With EtherNAT add-on card on Atari, the highest interrupt
- * number is 140 so NR_IRQS needs to be 141.
+ * With the SVEthlana adapter in the SuperVidel on Atari, the highest
+ * interrupt number is 141 so NR_IRQS needs to be 142.
  */
 #if defined(CONFIG_COLDFIRE)
 #define NR_IRQS 256
@@ -19,7 +19,7 @@
       defined(CONFIG_SUN3X) || defined(CONFIG_VIRT)
 #define NR_IRQS 200
 #elif defined(CONFIG_ATARI)
-#define NR_IRQS 141
+#define NR_IRQS 142
 #elif defined(CONFIG_MAC)
 #define NR_IRQS 72
 #elif defined(CONFIG_Q40)
-- 
2.17.1

Re: [PATCH v2 5/5] m68k: Add support for Svethlana

[Geert Uytterhoeven]

Hi Miro,

On Sun, 16 Aug 2026 at 09:38, Michael Schmitz <schmitzmic@gmail.com> wrote:
> From: Miro Kropacek <miro.kropacek@gmail.com>
>
> The SuperVidel FPGA contains an OpenCore ethernet package. Add
> platform device data and other necessary definitions to support
> this device on Atari.
>
> The ethoc module must be enabled separately by the user.

ETHOC selects MII and PHYLIB. Do you need to enable any MDIO or PHY
drivers?  I don't have the hardware, just asking for {atari,multi}_defconfig.

Thanks!

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH v2 5/5] m68k: Add support for Svethlana

[Geert Uytterhoeven]

On Mon, 17 Aug 2026 at 09:14, Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> On Sun, 16 Aug 2026 at 09:38, Michael Schmitz <schmitzmic@gmail.com> wrote:
> > From: Miro Kropacek <miro.kropacek@gmail.com>
> >
> > The SuperVidel FPGA contains an OpenCore ethernet package. Add
> > platform device data and other necessary definitions to support
> > this device on Atari.
> >
> > The ethoc module must be enabled separately by the user.
>
> ETHOC selects MII and PHYLIB. Do you need to enable any MDIO or PHY
> drivers?  I don't have the hardware, just asking for {atari,multi}_defconfig.

FTR, this increases kernel size by ca. 52 KiB.

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH v2 5/5] m68k: Add support for Svethlana

[Geert Uytterhoeven]

Hi Michael,

On Sun, 16 Aug 2026 at 09:38, Michael Schmitz <schmitzmic@gmail.com> wrote:
> From: Miro Kropacek <miro.kropacek@gmail.com>
>
> The SuperVidel FPGA contains an OpenCore ethernet package. Add
> platform device data and other necessary definitions to support
> this device on Atari.
>
> The ethoc module must be enabled separately by the user.
>
> Signed-off-by: Miro Kropacek <miro.kropacek@gmail.com>
> Reviewed-by: Michael Schmitz <schmitzmic@gmail.com>
> Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
> Link: https://lists.debian.org/debian-68k/2026/08/msg00000.html
>
> ---
> Changes in v2:
>
> - fix svethlana platform register error handling (Geert)

Thanks for the update!

> --- a/arch/m68k/atari/config.c
> +++ b/arch/m68k/atari/config.c

> @@ -892,6 +949,23 @@ static int __init atari_platform_init(void)
>         }
>  #endif
>
> +#ifdef CONFIG_ATARI_SVETHLANA
> +       {
> +               int error;
> +               void __iomem *sv_version;
> +
> +               sv_version = ioremap(ATARI_SV_VERSION_PHYS_ADDR, 4);
> +               if (sv_version) {
> +                       if (hwreg_present(sv_version) &&
> +                           (__raw_readl(sv_version) & 0x3ff) >= 10)
> +                               error = platform_device_register(&svethlana_device);
> +                               if (error && !rv)
> +                                       rv = error;

Missing curly braces:

    error: this ‘if’ clause does not guard... [-Werror=misleading-indentation]:

> +                       iounmap(sv_version);
> +               }
> +       }
> +#endif
> +
>  #ifdef CONFIG_ATARI_ETHERNEC
>         {
>                 int error;

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds