[PATCH v2] microblaze: nommu: Don't clobber R11 on syscalls

[PATCH v2] microblaze: nommu: Don't clobber R11 on syscalls

[steve]

The noMMU syscall trap has a bug that causes R11 to be zero on return to 
userland. Remove the extra "save" of R11 responsible for the bug, as well
as an unnecessary extra "load".

Signed-off-by: Steven J. Magnani <steve@digidescorp.com>
---
diff -uprN a/arch/microblaze/kernel/entry-nommu.S b/arch/microblaze/kernel/entry-nommu.S
--- a/arch/microblaze/kernel/entry-nommu.S	2009-12-09 17:05:11.000000000 -0600
+++ b/arch/microblaze/kernel/entry-nommu.S	2009-12-14 11:29:37.000000000 -0600
@@ -208,8 +208,6 @@ ENTRY(_user_exception)
 	lwi	r1, r1, TS_THREAD_INFO		/* get the thread info */
 	/* calculate kernel stack pointer */
 	addik	r1, r1, THREAD_SIZE - PT_SIZE
-	swi	r11, r0, PER_CPU(R11_SAVE)	/* temporarily save r11 */
-	lwi	r11, r0, PER_CPU(KM)		/* load mode indicator */
 2:
 	swi	r11, r1, PT_MODE		/* store the mode */
 	lwi	r11, r0, PER_CPU(R11_SAVE)	/* reload r11 */

RE: [microblaze-uclinux] [PATCH v2] microblaze: nommu: Don't clobber R11 on syscalls

[Steven J. Magnani]

On Mon, 2009-12-14 at 14:50 -0800, Jim Van Vorst wrote:
> Isn't R11 a temporary?  If it doesn't need to be preserved across function
> calls, why can't it be clobbered during a syscall?

Because the Microblaze syscall macros don't include r11 in the clobber
list. I suppose an alternative solution would just change the userspace
macros and not the kernel, even though there is a bug in the kernel. 

Michal, is the Microblaze ABI considered "locked down" now, or is a
kernel solution still permitted?

------------------------------------------------------------------------
 Steven J. Magnani               "I claim this network for MARS!
 www.digidescorp.com              Earthling, return my space modulator!"

 #include <standard.disclaimer>

Re: [PATCH v2] microblaze: nommu: Don't clobber R11 on syscalls

[Michal Simek]

steve@digidescorp.com wrote:
> The noMMU syscall trap has a bug that causes R11 to be zero on return to 
> userland. Remove the extra "save" of R11 responsible for the bug, as well
> as an unnecessary extra "load".

It seems to me that this patch means that you tested it.
But anyway I was pretty sure that I am right that's why I did this 
change myself. This patch is already in Linus tree.

Thanks for test,
Michal

> 
> Signed-off-by: Steven J. Magnani <steve@digidescorp.com>
> ---
> diff -uprN a/arch/microblaze/kernel/entry-nommu.S b/arch/microblaze/kernel/entry-nommu.S
> --- a/arch/microblaze/kernel/entry-nommu.S	2009-12-09 17:05:11.000000000 -0600
> +++ b/arch/microblaze/kernel/entry-nommu.S	2009-12-14 11:29:37.000000000 -0600
> @@ -208,8 +208,6 @@ ENTRY(_user_exception)
>  	lwi	r1, r1, TS_THREAD_INFO		/* get the thread info */
>  	/* calculate kernel stack pointer */
>  	addik	r1, r1, THREAD_SIZE - PT_SIZE
> -	swi	r11, r0, PER_CPU(R11_SAVE)	/* temporarily save r11 */
> -	lwi	r11, r0, PER_CPU(KM)		/* load mode indicator */
>  2:
>  	swi	r11, r1, PT_MODE		/* store the mode */
>  	lwi	r11, r0, PER_CPU(R11_SAVE)	/* reload r11 */
> 


-- 
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian