[PATCH 2/2] spec/aarch64: Only apply SUBALIGN(4) to ILP32

[Kinsey Moore]

-----Original Message-----
From: Sebastian Huber <sebastian.huber at embedded-brains.de> 
Sent: Monday, November 16, 2020 07:44
To: Kinsey Moore <kinsey.moore at oarcorp.com>; devel at rtems.org
Subject: Re: [PATCH 2/2] spec/aarch64: Only apply SUBALIGN(4) to ILP32

> On 16/11/2020 14:40, Kinsey Moore wrote:
>
>> -----Original Message-----
>> From: Sebastian Huber<sebastian.huber at embedded-brains.de>
>> Sent: Monday, November 16, 2020 07:35
>> To: Kinsey Moore<kinsey.moore at oarcorp.com>;devel at rtems.org
>> Subject: Re: [PATCH 2/2] spec/aarch64: Only apply SUBALIGN(4) to ILP32
>>
>>> On 16/11/2020 14:15, Kinsey Moore wrote:
>>>
>>>> The SUBALIGN(4) required on rtemsroset and rtemsrwset for ILP32 
>>>> builds was previously present on LP64 builds and causes no issues 
>>>> within RTEMS, but causes relocation/alignment issues when building libbsd.
>>>> This restricts those alignment changes to ILP32 builds.
>>> Please check it in if you think it is necessary.
>>>
>>> What troubles me a bit is that these SUBALIGN() stuff is present at all.
>>>
>>> It also troubles me that the splinkersets01 test case didn't catch this problem.
>> That test was what originally caught the problem during development of the A53 BSP. This patch isn't the addition of a fix for the alignment problems, it's an adjustment of the original fix to be compatible with libbsd's use of rtemsroset/rtemsrwset linker sections.
> So, this SUBALIGN() is just a workaround for some other problem? It would be good to document this known issue somewhere, for example a ticket. Is this an upstream problem in GCC or the GNU linker?

Yes, SUBALIGN() is the workaround for ILP32 defaulting to 8 byte alignment in the rtemsroset and rtemsrwset linker sections because RTEMS expects them to default to 4 byte alignment when using 4 byte pointers. I suspect it's an issue in the GNU linker, but I haven't had a chance to track down the root cause. This patch is necessary because while RTEMS works fine with 4 byte alignment in those sections in the context of 8 byte pointers, libbsd puts things other than pointers in those sections and requires the alignment of those sections to exactly match the pointer size.

I'll open a ticket for documentation purposes.

Kinsey