[gentoo-user] ATI-drivers 12.3 with Kernel 3.4

[gentoo-user] ATI-drivers 12.3 with Kernel 3.4

[Helmut Jarausch]

Hi,

does anybody know about patches for the ATI-drivers 12.3 to work with  
git-sources 3.4_rc3 ?

Many thanks for a hint,
Helmut.

[gentoo-user] Re: ATI-drivers 12.3 with Kernel 3.4

[walt]

On 04/16/2012 06:09 AM, Helmut Jarausch wrote:
> 
> Hi,
> 
> does anybody know about patches for the ATI-drivers 12.3 to work with git-sources 3.4_rc3 ?

Well, "patch" is too formal for the ugly hack I use :)

After building your new kernel you should patch one kernel header file
before building ati-drivers (taken from lkml):


--- arch/x86/include/asm/compat.h.orig  2012-04-08 11:51:29.569528342 -0700
+++ arch/x86/include/asm/compat.h       2012-04-08 14:33:58.309972502 -0700
@@ -221,6 +221,7 @@
        return (u32)(unsigned long)uptr;
 }
 
+#ifdef CONFIG_x86_64
 static inline void __user *arch_compat_alloc_user_space(long len)
 {
        compat_uptr_t sp;
@@ -234,6 +235,15 @@
 
        return (void __user *)round_down(sp - len, 16);
 }
+#else
+
+static inline void __user *arch_compat_alloc_user_space(long len)
+{
+ struct pt_regs *regs = task_pt_regs(current);
+ return (void __user *)regs->sp -len;
+}
+
+#endif
 
 static inline bool is_x32_task(void)
 {


Here's the ugly part:  some names in compat.h were changed recently, and
I can't be bothered to do a proper fix, so I hacked this together instead:

--- common/lib/modules/fglrx/build_mod/firegl_public.c  2012-03-23 13:38:48.000000000 -0700
+++ /tmp/firegl_public.c        2012-04-16 10:45:41.426582953 -0700
@@ -4181,7 +4181,7 @@
 {
     unsigned int p;
     KCL_DEBUG5(FN_FIREGL_KAS, "%d\n", level_init);
-    for_each_cpu_mask(p, cpu_possible_map)
+    for (p=0; p<4; p++)
     {
         KCL_DEBUG1(FN_FIREGL_KAS,"Setting initial execution level for CPU # %d\n", p);
         preempt_disable();


NOTE:  my new machine has 4 cpus, numbered 0 through 3, and I hardcoded that
number into the ati code, above, instead of deciphering the new kernel headers.
Ugly, ugly, ugly.  But it works perfectly :)

If you have two cpus you should change the p<4 to p<2, etc.  And then wait
for a professional fix from ati :p

Re: [gentoo-user] Re: ATI-drivers 12.3 with Kernel 3.4

[Helmut Jarausch]

Many thanks, Walt,

your patches work just fine for me.

Helmut.


On 04/16/2012 08:45:14 PM, walt wrote:
> On 04/16/2012 06:09 AM, Helmut Jarausch wrote:
> >
> > Hi,
> >
> > does anybody know about patches for the ATI-drivers 12.3 to work  
> with git-sources 3.4_rc3 ?
> 
> Well, "patch" is too formal for the ugly hack I use :)
> 
> After building your new kernel you should patch one kernel header file
> before building ati-drivers (taken from lkml):
> 
> 
> --- arch/x86/include/asm/compat.h.orig  2012-04-08 11:51:29.569528342  
> -0700
> +++ arch/x86/include/asm/compat.h       2012-04-08 14:33:58.309972502  
> -0700
> @@ -221,6 +221,7 @@
>         return (u32)(unsigned long)uptr;
>  }
> 
> +#ifdef CONFIG_x86_64
>  static inline void __user *arch_compat_alloc_user_space(long len)
>  {
>         compat_uptr_t sp;
> @@ -234,6 +235,15 @@
> 
>         return (void __user *)round_down(sp - len, 16);
>  }
> +#else
> +
> +static inline void __user *arch_compat_alloc_user_space(long len)
> +{
> + struct pt_regs *regs = task_pt_regs(current);
> + return (void __user *)regs->sp -len;
> +}
> +
> +#endif
> 
>  static inline bool is_x32_task(void)
>  {
> 
> 
> Here's the ugly part:  some names in compat.h were changed recently,  
> and
> I can't be bothered to do a proper fix, so I hacked this together  
> instead:
> 
> --- common/lib/modules/fglrx/build_mod/firegl_public.c  2012-03-23  
> 13:38:48.000000000 -0700
> +++ /tmp/firegl_public.c        2012-04-16 10:45:41.426582953 -0700
> @@ -4181,7 +4181,7 @@
>  {
>      unsigned int p;
>      KCL_DEBUG5(FN_FIREGL_KAS, "%d\n", level_init);
> -    for_each_cpu_mask(p, cpu_possible_map)
> +    for (p=0; p<4; p++)
>      {
>          KCL_DEBUG1(FN_FIREGL_KAS,"Setting initial execution level  
> for CPU # %d\n", p);
>          preempt_disable();
> 
> 
> NOTE:  my new machine has 4 cpus, numbered 0 through 3, and I  
> hardcoded that
> number into the ati code, above, instead of deciphering the new  
> kernel headers.
> Ugly, ugly, ugly.  But it works perfectly :)
> 
> If you have two cpus you should change the p<4 to p<2, etc.  And then  
> wait
> for a professional fix from ati :p
> 
> 
> 
> 
>