[PATCH] ARM: vexpress: ignore MIDR[0] for uart detection

[PATCH] ARM: vexpress: ignore MIDR[0] for uart detection

[Mian Yousaf Kaukab]

Qemu vexpress-a9 target reports id of ARM Cortex-A9 core as r0p0
and uses legacy map. That is UART0 at address 0x10009000. Ignore
MIDR[0] during uart detection so that r0p0 as well as r0p1 use
legacy map.

Signed-off-by: Mian Yousaf Kaukab <yrehan@gmail.com>
---
 arch/arm/include/debug/vexpress.S | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/arch/arm/include/debug/vexpress.S b/arch/arm/include/debug/vexpress.S
index 524acd5..932db8b 100644
--- a/arch/arm/include/debug/vexpress.S
+++ b/arch/arm/include/debug/vexpress.S
@@ -27,11 +27,16 @@
 
 		@ Make an educated guess regarding the memory map:
 		@ - the original A9 core tile (based on ARM Cortex-A9 r0p1)
-		@   should use UART at 0x10009000
+		@   should use UART at 0x10009000. Moreover, qemu vexpress-a9
+		@   target reports id of ARM Cortex-A9 core as r0p0 and uses
+		@   legacy map.
 		@ - all other (RS1 complaint) tiles use UART mapped
 		@   at 0x1c090000
 		mrc	p15, 0, \rp, c0, c0, 0
-		movw	\rv, #0xc091
+		mov	\rv, #1
+		@ Ignore MIDR[0] during comparison
+		bic	\rp, \rp, \rv
+		movw	\rv, #0xc090
 		movt	\rv, #0x410f
 		cmp	\rp, \rv
 
-- 
1.8.1.2

[PATCH] ARM: vexpress: ignore MIDR[0] for uart detection

[Mian Yousaf Kaukab]

Hi Pawel,

On Sun, Sep 29, 2013 at 12:17 PM, Mian Yousaf Kaukab <yrehan@gmail.com> wrote:
> Qemu vexpress-a9 target reports id of ARM Cortex-A9 core as r0p0
> and uses legacy map. That is UART0 at address 0x10009000. Ignore
> MIDR[0] during uart detection so that r0p0 as well as r0p1 use
> legacy map.
>
> Signed-off-by: Mian Yousaf Kaukab <yrehan@gmail.com>
> ---
>  arch/arm/include/debug/vexpress.S | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/arch/arm/include/debug/vexpress.S b/arch/arm/include/debug/vexpress.S
> index 524acd5..932db8b 100644
> --- a/arch/arm/include/debug/vexpress.S
> +++ b/arch/arm/include/debug/vexpress.S
> @@ -27,11 +27,16 @@
>
>                 @ Make an educated guess regarding the memory map:
>                 @ - the original A9 core tile (based on ARM Cortex-A9 r0p1)
> -               @   should use UART at 0x10009000
> +               @   should use UART at 0x10009000. Moreover, qemu vexpress-a9
> +               @   target reports id of ARM Cortex-A9 core as r0p0 and uses
> +               @   legacy map.
>                 @ - all other (RS1 complaint) tiles use UART mapped
>                 @   at 0x1c090000
>                 mrc     p15, 0, \rp, c0, c0, 0
> -               movw    \rv, #0xc091
> +               mov     \rv, #1
> +               @ Ignore MIDR[0] during comparison
> +               bic     \rp, \rp, \rv
> +               movw    \rv, #0xc090
>                 movt    \rv, #0x410f
>                 cmp     \rp, \rv

Can you please take a look at this patch? Default vexpress_defconfig
does not boot on qemu vexpress-a9 target without it.
I apologize if you get this message twice. I have to resend because
the previous message didn't reach the mailing list.


BR,
Yousaf

[PATCH] ARM: vexpress: ignore MIDR[0] for uart detection

[Pawel Moll]

On Tue, 2013-10-01 at 09:11 +0100, Mian Yousaf Kaukab wrote:
> Can you please take a look at this patch? Default vexpress_defconfig
> does not boot on qemu vexpress-a9 target without it.

Right, I think I've heard about it...

To be honest, I have rather mixed feelings about it. In particular I
could say that if qemu returns r0p0, it doesn't model V2P-CA9, which has
A9 r0p1 :-) Also, the DEBUG_LL and EARLY_PRINTK are pretty low level
features and it seems that they simply shouldn't be enabled in the
defconfig. As I'll be updating defconfig soon, I'll make sure they're
disabled.

Yes, I agree that the "detection" code is a hack. I'm not sure we have
to make it even hackier. Let me have a think about it, as I can't really
spend any time of the matter this week, sorry.

Thanks for your efforts!

Pawel

[PATCH] ARM: vexpress: ignore MIDR[0] for uart detection

[Mian Yousaf Kaukab]

On Tue, Oct 1, 2013 at 6:39 PM, Pawel Moll <pawel.moll@arm.com> wrote:
> To be honest, I have rather mixed feelings about it. In particular I
> could say that if qemu returns r0p0, it doesn't model V2P-CA9, which has
> A9 r0p1 :-) [...]

r0p0 looked legitimate to me :). Ok I will drive this with qemu developers then.

Thanks,

Yousaf