3.11-rc7 big-endian support

[Ben Dooks]

On 29/08/13 10:59, Thomas Petazzoni wrote:
> Dear Ben Dooks,
>
> On Tue, 27 Aug 2013 23:02:01 +0100, Ben Dooks wrote:
>> On 27/08/13 22:38, Ben Dooks wrote:
>>> This is a new series of the core work for getting big-endian working
>>> nicely on ARM. This is mainly a rebase on 3.11-rc7 and has not had
>>> a lot of testing on it.
>>>
>>> It is available on:
>>>
>>> 	git://git.baserock.org/delta/linux.git baserock/311-rc7/be/core-v1
>>>
>>> Since the previous version:
>>>
>>> - Added Victor's atomic64 patch
>>> - Fixed BUG() patch to use<asm/opcodes.h>
>>>
>>> I would like to get as much of this merged as possible, so would
>>> it be possible
>>
>> Is it possible to get all or some of this series pulled before the
>> next merge window? Do I need to remove the machine specific parts
>> of the series?
>
> I believe your patch series would get more attention if the cover
> letter was a bit better. It lacks a version number and a changelog. The
> new posting you made as "re-send patch series due to mta issues" does
> not even have a cover letter.

That was in-reply to as it seems that some of the messages never made
it back to me due to a mta issue at my end that needed fixing.

> Also, none of the patches are Cc'ed to the relevant maintainers, so I'm
> not sure how you expect those maintainers to look at your patches?

I actually left the machine specific ones in this series as a convenient
place to keep them before producing a tree for submission.

> Would it be possible to Cc me on future postings of this patch set?
>
> Considering the Acked-by and Reviewed-by you had, I would suggest that
> you push the following patches in Russell King's patch system:
>
>   [PATCH 02/19] ARM: asm: Add ARM_BE8() assembly helper
>   [PATCH 03/19] ARM: fixup_pv_table bug when CPU_ENDIAN_BE8
>                 (after adding Dave Martin Reviewed-by)
>   [PATCH 04/19] ARM: set BE8 if LE in head code
>                 (after adding Dave Martin Reviewed-by and removing the
>                 merge line conflict)
>   [PATCH 05/19] ARM: pl01x debug code endian fix
>   [PATCH 06/19] ARM: twd: data endian fix
>   [PATCH 07/19] ARM: smp_scu: data endian fixes
>   [PATCH 11/19] ARM: alignment: correctly decode instructions in BE8 mode.
>                 (after adding Dave Martin Reviewed-by)
>   [PATCH 12/19] ARM: traps: use<asm/opcodes.h>  to get correct instruction order
>                 (after fixing the typo pointed by Dave Martin)
>   [PATCH 13/19] ARM: module: correctly relocate instructions in BE8
>                 (after adding Dave Martin Reviewed-by)
>   [PATCH 14/19] ARM: set --be8 when linking modules
>                 (after adding Dave Martin Reviewed-by)
>   [PATCH 15/19] ARM: hardware: fix endian-ness in<hardware/coresight.h>
>   [PATCH 16/19] ARM: net: fix arm instruction endian-ness in bpf_jit_32.c
>                 (after adding Dave Martin Reviewed-by)
>   [PATCH 17/19] ARM: Correct BUG() assembly to ensure it is endian-agnostic
>                 (after adding Dave Martin Reviewed-by)
>   [PATCH 18/19] ARM: kdgb: use<asm/opcodes.h>  for data to be assembled as intruction
>                 (after adding Dave Martin Reviewed-by)
>   [PATCH 19/19] ARM: atomic64: fix endian-ness in atomic.h
>
> Considering the release timing, you should really submit those either
> today or tomorrow into Russell's patch system if you want them to be
> merged for 3.12.
>
> Regarding mvebu, the below patch is needed to get the secondary CPUs to
> boot. Other than that, on Armada XP:
>
> Tested-by: Thomas Petazzoni<thomas.petazzoni at free-electrons.com>

I will fix up the original patch, as it seems to have gotten out of sync
with the change to add the coherency fabric code.

> It doesn't work yet on Armada 370, but I'm not sure it's due to your
> patches and anyway isn't a regression since LE continues to work fine
> on Armada 370.

I wonder if it is due to a bootloader issue. I will push out the atags
branch later today and you can try that.

> diff --git a/arch/arm/mach-mvebu/headsmp.S b/arch/arm/mach-mvebu/headsmp.S
> index 3109df0..be067b7 100644
> --- a/arch/arm/mach-mvebu/headsmp.S
> +++ b/arch/arm/mach-mvebu/headsmp.S
> @@ -29,13 +29,13 @@
>    * startup
>    */
>   ENTRY(armada_xp_secondary_startup)
> +ARM_BE8(setend be )                    @ go BE8 if booted LE
> +
>          /* Get coherency fabric base physical address */
>          adr     r0, 1f
>          ldr     r1, [r0]
>          ldr     r0, [r0, r1]
>
> - ARM_BE8(setend        be )                    @ go BE8 if booted LE
> -
>          /* Read CPU id */
>          mrc     p15, 0, r1, c0, c0, 5
>          and     r1, r1, #0xF
>
> Thanks,
>
> Thomas


-- 
Ben Dooks				http://www.codethink.co.uk/
Senior Engineer				Codethink - Providing Genius