[PATCH 1/3] radeonfb: Fix 64 bits resources on 32 bits archs

[PATCH 1/3] radeonfb: Fix 64 bits resources on 32 bits archs

[Benjamin Herrenschmidt]

This fixes radeonfb to not truncate 64 bits resources on 32 bits
platforms. Unfortunately, there are still issues with addresses
returned to userspace via struct fb_fix_screeninfo. This will
have to be dealt with separately.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---

 drivers/video/aty/radeon_base.c |    4 ++--
 drivers/video/aty/radeonfb.h    |    4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)

--- linux-work.orig/drivers/video/aty/radeon_base.c	2008-04-22 11:04:19.000000000 +1000
+++ linux-work/drivers/video/aty/radeon_base.c	2008-04-22 11:05:00.000000000 +1000
@@ -1886,7 +1886,7 @@ static int __devinit radeon_set_fbinfo (
 	info->screen_size = rinfo->mapped_vram;
 	/* Fill fix common fields */
 	strlcpy(info->fix.id, rinfo->name, sizeof(info->fix.id));
-        info->fix.smem_start = rinfo->fb_base_phys;
+        info->fix.smem_start = (unsigned long)rinfo->fb_base_phys;
         info->fix.smem_len = rinfo->video_ram;
         info->fix.type = FB_TYPE_PACKED_PIXELS;
         info->fix.visual = FB_VISUAL_PSEUDOCOLOR;
@@ -1894,7 +1894,7 @@ static int __devinit radeon_set_fbinfo (
         info->fix.ypanstep = 1;
         info->fix.ywrapstep = 0;
         info->fix.type_aux = 0;
-        info->fix.mmio_start = rinfo->mmio_base_phys;
+        info->fix.mmio_start = (unsigned long)rinfo->mmio_base_phys;
         info->fix.mmio_len = RADEON_REGSIZE;
 	info->fix.accel = FB_ACCEL_ATI_RADEON;
 
Index: linux-work/drivers/video/aty/radeonfb.h
===================================================================
--- linux-work.orig/drivers/video/aty/radeonfb.h	2008-04-22 11:03:17.000000000 +1000
+++ linux-work/drivers/video/aty/radeonfb.h	2008-04-22 11:03:27.000000000 +1000
@@ -287,8 +287,8 @@ struct radeonfb_info {
 
 	char			name[DEVICE_NAME_SIZE];
 
-	unsigned long		mmio_base_phys;
-	unsigned long		fb_base_phys;
+	resource_size_t		mmio_base_phys;
+	resource_size_t		fb_base_phys;
 
 	void __iomem		*mmio_base;
 	void __iomem		*fb_base;

Re: [PATCH 1/3] radeonfb: Fix 64 bits resources on 32 bits archs

[Christian Ehrhardt]

Benjamin Herrenschmidt wrote:
> This fixes radeonfb to not truncate 64 bits resources on 32 bits
> platforms. Unfortunately, there are still issues with addresses
> returned to userspace via struct fb_fix_screeninfo. This will
> have to be dealt with separately.

Thanks for this patch Benjamin, I use it together with what we have discussed in the "pci issue - wrong detection of pci ressources" thread.
Unfortunately I now hit exactly that issue with fb_fix_screeninfo you describe.

For everyone the fb_fix_screeninfo has two "unsigned long" vars that need to strore a IO address.
This fails in my case with a 32bit powerpc system (=sizeof(long)=4) which has paddr >4Gb and actually it should affect any 32bit platform with paddr>4Gb.
You see it e.g. when you try to initialize X11, the x11 radeon driver issues a FBIOGET_FSCREENINFO ioctl and because our address is >4Gb it get's clobbered by that unsigned long in the fb_fix_screeninfo structure (the value comes from a resource_size_t variable which has the correct 64bit).

struct fb_fix_screeninfo {
        char id[16];                    /* identification string eg "TT Builtin" */
        unsigned long smem_start;     /* Start of frame buffer mem */
[...]
        unsigned long mmio_start;     /* Start of Memory Mapped I/O   */
[...]

I tried the stupid solution to just change the fb_fix_screeninfo structure to resource_size_t, but that changes the size ioctl transports and would require awareness in the userspace applications using that ioctl.
The X11 & Framebuffer driver work on 64bit systems, so I think it's just an issue of not cutting that data down to 32bit when transporting it (has anyone already checked the x11 drivers, I hope they don't use unsigned long too).

I wanted to ask if there are any known workarounds atm that would allow me to use my X11 for now?

-- 

Grüsse / regards, 
Christian Ehrhardt
IBM Linux Technology Center, Open Virtualization

Re: [PATCH 1/3] radeonfb: Fix 64 bits resources on 32 bits archs

[Michel Dänzer]

On Tue, 2008-04-22 at 17:55 +0200, Christian Ehrhardt wrote:
> 
> You see it e.g. when you try to initialize X11, the x11 radeon driver
> issues a FBIOGET_FSCREENINFO ioctl [...]

FWIW, current versions of the xf86-video-ati radeon driver no longer use
radeonfb for anything.


-- 
Earthling Michel Dänzer           |          http://tungstengraphics.com
Libre software enthusiast         |          Debian, X and DRI developer

Re: [PATCH 1/3] radeonfb: Fix 64 bits resources on 32 bits archs

[Benjamin Herrenschmidt]

On Tue, 2008-04-22 at 17:55 +0200, Christian Ehrhardt wrote:
> I wanted to ask if there are any known workarounds atm that would
> allow me to use my X11 for now?

X is doing a mmap of /dev/mem instead of /dev/fb ?

You can normally map the fb mapping /dev/fb and then map the registers
using /dev/fb at an offset beyond the framebuffer (fix->smem_len).

If X is using /dev/mem instead, then it's being stupid and needs to be
fixed...

Cheers,
Ben.

Re: [Linux-fbdev-devel] [PATCH 1/3] radeonfb: Fix 64 bits resources on 32 bits archs

[Michel Dänzer]

On Wed, 2008-04-23 at 08:21 +1000, Benjamin Herrenschmidt wrote:
> On Tue, 2008-04-22 at 17:55 +0200, Christian Ehrhardt wrote:
> > I wanted to ask if there are any known workarounds atm that would
> > allow me to use my X11 for now?
> 
> X is doing a mmap of /dev/mem instead of /dev/fb ?
> 
> You can normally map the fb mapping /dev/fb and then map the registers
> using /dev/fb at an offset beyond the framebuffer (fix->smem_len).
> 
> If X is using /dev/mem instead, then it's being stupid and needs to be
> fixed...

It's up to the driver, and again, the current radeon driver doesn't use
radeonfb at all anymore...


-- 
Earthling Michel Dänzer           |          http://tungstengraphics.com
Libre software enthusiast         |          Debian, X and DRI developer

Re: [Linux-fbdev-devel] [PATCH 1/3] radeonfb: Fix 64 bits resources on 32 bits archs

[Benjamin Herrenschmidt]

On Wed, 2008-04-23 at 11:32 +0200, Michel Dänzer wrote:
> > X is doing a mmap of /dev/mem instead of /dev/fb ?
> > 
> > You can normally map the fb mapping /dev/fb and then map the
> registers
> > using /dev/fb at an offset beyond the framebuffer (fix->smem_len).
> > 
> > If X is using /dev/mem instead, then it's being stupid and needs to
> be
> > fixed...
> 
> It's up to the driver, and again, the current radeon driver doesn't
> use
> radeonfb at all anymore...

Right. And I'm happy to go fix what remains, don't get me wrong here :-)
We all know how stupid legacy code can be !

I'll try to get some patches to expose a new version of the structure
sometimes in the upcoming week or two (we'll see what time permits) and
then see about fixing X drivers that need fixing, though hopefully with
pci-rework being upstream now, that shouldn't be many.

The main issue is directfb. I hope they'll catch up with the new ioctls.

Ben.

Re: [Linux-fbdev-devel] [PATCH 1/3] radeonfb: Fix 64 bits resources on 32 bits archs

[David Miller]

From: Michel D=E4nzer <michel@tungstengraphics.com>
Date: Wed, 23 Apr 2008 11:32:07 +0200

> On Wed, 2008-04-23 at 08:21 +1000, Benjamin Herrenschmidt wrote:
> > On Tue, 2008-04-22 at 17:55 +0200, Christian Ehrhardt wrote:
> > > I wanted to ask if there are any known workarounds atm that would=

> > > allow me to use my X11 for now?
> > =

> > X is doing a mmap of /dev/mem instead of /dev/fb ?
> > =

> > You can normally map the fb mapping /dev/fb and then map the regist=
ers
> > using /dev/fb at an offset beyond the framebuffer (fix->smem_len).
> > =

> > If X is using /dev/mem instead, then it's being stupid and needs to=
 be
> > fixed...
> =

> It's up to the driver, and again, the current radeon driver doesn't u=
se
> radeonfb at all anymore...

The only portable thing is for X to use the PCI sysfs mmap() stuff,
which current Xorg servers using libpciaccess do.

I know because I'm typing this from a sparc64 using a Radeon
with current Xorg :-)

Re: [Linux-fbdev-devel] [PATCH 1/3] radeonfb: Fix 64 bits resources on 32 bits archs

[Benjamin Herrenschmidt]

On Wed, 2008-04-23 at 02:57 -0700, David Miller wrote:
> > It's up to the driver, and again, the current radeon driver doesn't
> use
> > radeonfb at all anymore...
> 
> The only portable thing is for X to use the PCI sysfs mmap() stuff,
> which current Xorg servers using libpciaccess do.
> 
> I know because I'm typing this from a sparc64 using a Radeon
> with current Xorg :-)

True :-) But I'd still like to fix the fb API for the sake of things
like directfb etc...

Ben.

Re: [PATCH 1/3] radeonfb: Fix 64 bits resources on 32 bits archs

[Sergei Shtylyov]

Hello.

Benjamin Herrenschmidt wrote:

> This fixes radeonfb to not truncate 64 bits resources on 32 bits
> platforms. Unfortunately, there are still issues with addresses
> returned to userspace via struct fb_fix_screeninfo. This will
> have to be dealt with separately.
>
> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
>   

   I'm seeing that these 3 patch series (dating back to April) still 
unapplied. Have everybody forgotten about them? I'd like to see them 
finally applied.

> --- linux-work.orig/drivers/video/aty/radeon_base.c	2008-04-22 11:04:19.000000000 +1000
> +++ linux-work/drivers/video/aty/radeon_base.c	2008-04-22 11:05:00.000000000 +1000
> @@ -1886,7 +1886,7 @@ static int __devinit radeon_set_fbinfo (
>  	info->screen_size = rinfo->mapped_vram;
>  	/* Fill fix common fields */
>  	strlcpy(info->fix.id, rinfo->name, sizeof(info->fix.id));
> -        info->fix.smem_start = rinfo->fb_base_phys;
> +        info->fix.smem_start = (unsigned long)rinfo->fb_base_phys;
>          info->fix.smem_len = rinfo->video_ram;
>          info->fix.type = FB_TYPE_PACKED_PIXELS;
>          info->fix.visual = FB_VISUAL_PSEUDOCOLOR;
> @@ -1894,7 +1894,7 @@ static int __devinit radeon_set_fbinfo (
>          info->fix.ypanstep = 1;
>          info->fix.ywrapstep = 0;
>          info->fix.type_aux = 0;
> -        info->fix.mmio_start = rinfo->mmio_base_phys;
> +        info->fix.mmio_start = (unsigned long)rinfo->mmio_base_phys;
>          info->fix.mmio_len = RADEON_REGSIZE;
>  	info->fix.accel = FB_ACCEL_ATI_RADEON;
>  
> Index: linux-work/drivers/video/aty/radeonfb.h
> ===================================================================
> --- linux-work.orig/drivers/video/aty/radeonfb.h	2008-04-22 11:03:17.000000000 +1000
> +++ linux-work/drivers/video/aty/radeonfb.h	2008-04-22 11:03:27.000000000 +1000
> @@ -287,8 +287,8 @@ struct radeonfb_info {
>  
>  	char			name[DEVICE_NAME_SIZE];
>  
> -	unsigned long		mmio_base_phys;
> -	unsigned long		fb_base_phys;
> +	resource_size_t		mmio_base_phys;
> +	resource_size_t		fb_base_phys;
>  
>  	void __iomem		*mmio_base;
>  	void __iomem		*fb_base;

WBR, Sergei

Re: [PATCH 1/3] radeonfb: Fix 64 bits resources on 32 bits archs

[Sergei Shtylyov]

Hello.

Benjamin Herrenschmidt wrote:

> This fixes radeonfb to not truncate 64 bits resources on 32 bits
> platforms. Unfortunately, there are still issues with addresses
> returned to userspace via struct fb_fix_screeninfo. This will
> have to be dealt with separately.
>
> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
>   

   I'm seeing that this 3 patch series (dating back to April) still 
unapplied. Have everybody forgotten about them? I'd like to see them 
finally applied.

> --- linux-work.orig/drivers/video/aty/radeon_base.c	2008-04-22 11:04:19.000000000 +1000
> +++ linux-work/drivers/video/aty/radeon_base.c	2008-04-22 11:05:00.000000000 +1000
> @@ -1886,7 +1886,7 @@ static int __devinit radeon_set_fbinfo (
>  	info->screen_size = rinfo->mapped_vram;
>  	/* Fill fix common fields */
>  	strlcpy(info->fix.id, rinfo->name, sizeof(info->fix.id));
> -        info->fix.smem_start = rinfo->fb_base_phys;
> +        info->fix.smem_start = (unsigned long)rinfo->fb_base_phys;
>          info->fix.smem_len = rinfo->video_ram;
>          info->fix.type = FB_TYPE_PACKED_PIXELS;
>          info->fix.visual = FB_VISUAL_PSEUDOCOLOR;
> @@ -1894,7 +1894,7 @@ static int __devinit radeon_set_fbinfo (
>          info->fix.ypanstep = 1;
>          info->fix.ywrapstep = 0;
>          info->fix.type_aux = 0;
> -        info->fix.mmio_start = rinfo->mmio_base_phys;
> +        info->fix.mmio_start = (unsigned long)rinfo->mmio_base_phys;
>          info->fix.mmio_len = RADEON_REGSIZE;
>  	info->fix.accel = FB_ACCEL_ATI_RADEON;
>  
> Index: linux-work/drivers/video/aty/radeonfb.h
> ===================================================================
> --- linux-work.orig/drivers/video/aty/radeonfb.h	2008-04-22 11:03:17.000000000 +1000
> +++ linux-work/drivers/video/aty/radeonfb.h	2008-04-22 11:03:27.000000000 +1000
> @@ -287,8 +287,8 @@ struct radeonfb_info {
>  
>  	char			name[DEVICE_NAME_SIZE];
>  
> -	unsigned long		mmio_base_phys;
> -	unsigned long		fb_base_phys;
> +	resource_size_t		mmio_base_phys;
> +	resource_size_t		fb_base_phys;
>  
>  	void __iomem		*mmio_base;
>  	void __iomem		*fb_base;

WBR, Sergei