Re: [Buildroot] [arc-buildroot] [PATCH] arch: remove support for BR2_ARC_PAGE_SIZE_4K

From: Thomas Petazzoni via buildroot <buildroot@buildroot.org>
To: Alexey Brodkin via buildroot <buildroot@buildroot.org>
Cc: Alexey Brodkin <Alexey.Brodkin@synopsys.com>,
	ARC Maintainers <arc-buildroot@synopsys.com>,
	Claudiu Zissulescu <Claudiu.Zissulescu@synopsys.com>
Subject: Re: [Buildroot] [arc-buildroot] [PATCH] arch: remove support for BR2_ARC_PAGE_SIZE_4K
Date: Tue, 26 Jul 2022 15:21:08 +0200	[thread overview]
Message-ID: <20220726152108.3470eb7d@windsurf> (raw)
In-Reply-To: <IA1PR12MB62107029D82ADA688CF7C210A18F9@IA1PR12MB6210.namprd12.prod.outlook.com>

Hello Alexey,

On Tue, 19 Jul 2022 18:20:25 +0000
Alexey Brodkin via buildroot <buildroot@buildroot.org> wrote:

> So that seems to be a good catch indeed.
> 
> 1. In older GNU LD such a check doesn't exist and so LD never craps out.
>    And here I mean the latest ARC Binutils ("arc-2020.09-release"), which are based
>    on upstream v2.35. Thus if building with 4KiB pages but ARC Binutils,
>    the problem is not seen. I'm sure even built that way binaries work perfectly
>    fine as I do remember trying it not that long ago myself.
>    
>    BTW that pre-defined "common-page-size" comes from GCC actually, see
>    https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=gcc/config/arc/linux.h;h=0f1ff055be757ad1483910fc4918e7fef7fe4d5e;hb=1ea978e3066ac565a1ec28a96a4d61eaf38e2726#l48

OK.

> 2. And in case of building for ARCv3 processors we don't mess with "page-size"
>    stuff in the "LINK_SPEC", see https://github.com/foss-for-synopsys-dwc-arc-processors/gcc/blob/arc64/gcc/config/arc64/linux32.h#L26

OK, but I don't really see the relationship.

> That said I'd like to get our GCC & Binutils maintainer input on what needs
> to be done with these "page-size" options in the "LINK_SPEC".
> 
> Maybe with use of up-to-date Binutils those default values are no longer needed?

Why no longer needed? Of course is you change the page size, you need
to pass some options to the linker so that it can place sections and
define load addresses (and other related stuff) accordingly.

> Or we may keep default "max" value and then not explicitly set outside "common"
> value will be adjusted as needed automatically.

I don't understand what you mean here. Today, we are setting
max-page-size, and that doesn't work because the common-page-size value
is higher. As I mentioned in my initial patch, passing
-Wl,-z,max-page-size=4096 (already done) and
-Wl,-z,common-page-size=4096 would most likely fix the error, but I
wasn't sure it was the correct thing to do.

Do you have more feedback from your GCC and binutils maintainer perhaps?

Thanks a lot!

Thomas
-- 
Thomas Petazzoni, co-owner and CEO, Bootlin
Embedded Linux and Kernel engineering and training
https://bootlin.com
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot