[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <mhng-0fb8c6a4-9c20-4cdb-87c6-8b27950c1b45@palmerdabbelt-glaptop1>
Date: Thu, 05 Nov 2020 18:48:41 -0800 (PST)
From: Palmer Dabbelt <palmer@...belt.com>
To: will@...nel.org
CC: Will Deacon <willdeacon@...gle.com>, catalin.marinas@....com,
nickhu@...estech.com, green.hu@...il.com, deanbo422@...il.com,
linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
kernel-team@...roid.com
Subject: Re: [PATCH] arm64: Fix a broken copyright header in gen_vdso_offsets.sh
On Tue, 20 Oct 2020 04:32:08 PDT (-0700), will@...nel.org wrote:
> On Fri, Oct 16, 2020 at 05:26:37PM -0700, Palmer Dabbelt wrote:
>> From: Palmer Dabbelt <palmerdabbelt@...gle.com>
>>
>> I was going to copy this but I didn't want to chase around the build
>> system stuff so I did it a different way.
>
> How did you end up doing it? I remember trying to parse the ELF object at
> one time of day and not getting very far.
Mine really isn't any cleaner: I use nm and sed to produce an assembler file
full of absolute symbols, which I compile to produce the object. That way I
don't need a header file, which means I don't need to mess around with build
ordering or change any of the other code (which was already using symbols).
I don't really like this whole thing we do where we just assume that symbols
with a small offset from absolute address zero. We're sort of depending on
that for undefined weak symbols, but that's only for actually zero not almost
zero. Given that, I figured it'd be easier to just do it this way rather than
trying to mess with build stuff.
I probably wouldn't bother changing unless there's something actively bad about
that build stuff (I didn't really look).
Powered by blists - more mailing lists