Re: [PATCH 2/2] x86/boot/compressed: Remove unnecessary sections from bzImage

From: Fangrui Song <maskray@google.com>
To: Arvind Sankar <nivedita@alum.mit.edu>
Cc: Nick Desaulniers <ndesaulniers@google.com>,
	Michael Matz <matz@suse.de>, Borislav Petkov <bp@alien8.de>,
	Nathan Chancellor <natechancellor@gmail.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Ingo Molnar <mingo@redhat.com>, "H. Peter Anvin" <hpa@zytor.com>,
	"maintainer:X86 ARCHITECTURE (32-BIT AND 64-BIT)"
	<x86@kernel.org>, LKML <linux-kernel@vger.kernel.org>,
	clang-built-linux <clang-built-linux@googlegroups.com>,
	Kees Cook <keescook@chromium.org>
Subject: Re: [PATCH 2/2] x86/boot/compressed: Remove unnecessary sections from bzImage
Date: Mon, 24 Feb 2020 14:50:59 -0800	[thread overview]
Message-ID: <20200224225059.2scjklfi4g7wwdkp@google.com> (raw)
In-Reply-To: <20200224224343.GA572699@rani.riverdale.lan>

On 2020-02-24, Arvind Sankar wrote:
>On Mon, Feb 24, 2020 at 02:17:03PM -0800, Fangrui Song wrote:
>> On 2020-02-24, Arvind Sankar wrote:
>> >On Mon, Feb 24, 2020 at 01:28:28PM -0800, Fangrui Song wrote:
>> >> Hi Michael, please see my other reply on this thread: https://lkml.org/lkml/2020/2/24/47
>> >>
>> >> Synthesized sections can be matched as well. For example, SECTIONS { .pltfoo : { *(.plt) }} can rename the output section .plt to .pltfoo
>> >> It seems that in GNU ld, the synthesized section is associated with the
>> >> original object file, so it can be written as:
>> >>
>> >>    SECTIONS { .pltfoo : { a.o(.plt) }}
>> >>
>> >> In lld, you need a wildcard to match the synthesized section *(.plt)
>> >>
>> >> .rela.dyn is another example.
>> >>
>> >
>> >With the BFD toolchain, file matching doesn't actually seem to work at
>> >least for .rela.dyn. I've tried playing around with it in the past and
>> >if you try to use file-matching to capture relocations from a particular
>> >input file, it just doesn't work sensibly.
>>
>> I think most things are working in GNU ld...
>>
>> /* a.x */
>> SECTIONS {
>>    .rela.pltfoo : { a.o(.rela.plt) }  /* *(.rela.plt) with lld */
>>    .rela.dynfoo : { a.o(.rela.data) } /* *(.rela.dyn) with lld */
>> }
>
>The file matching doesn't do anything sensible. If you split your .data
>section out into b.s, and update the linker script so it filters for
>b.o(.rela.data), .rela.dynfoo doesn't get created, instead the default
>.rela.dyn will contains the .data section relocation. If you keep the
>filter as a.o(.rela.data), you get .rela.dynfoo, even though a.o doesn't
>actually contain any .rela.data section any more.

I raised the examples to support my viewpoint "synthesized sections can
be matched, as well as input sections."

If there is really a need (rare, not recommended) to rename output
sections only consisting of synthesized sections (e.g. .plt .rela.dyn),
for linker portability, it is better using a wildcard for the input
filename pattern.

As another example, SECTIONS { /DISCARD/ : { *(.rela.*) } } discards synthesized .rela.*

>>
>> % cat <<e > a.s
>>   .globl foo
>>   foo:
>>     call bar
>>   .data
>>   .quad quz
>> e
>> % as a.s -o a.o
>> % ld.bfd -T a.x a.o -shared -o a.so