linux-kernel - Re: [PATCH v5] arm64/module: Use text-poke API for late relocations.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CADBMgpx==FnFj4okXs1n3NPngh7Os1YpnGrDDe8z_t2X7bzOOQ@mail.gmail.com>
Date: Fri, 30 May 2025 17:11:00 -0700
From: Dylan Hatch <dylanbhatch@...gle.com>
To: Will Deacon <will@...nel.org>
Cc: Catalin Marinas <catalin.marinas@....com>, Ard Biesheuvel <ardb@...nel.org>, 
	Sami Tolvanen <samitolvanen@...gle.com>, Geert Uytterhoeven <geert@...ux-m68k.org>, 
	Song Liu <song@...nel.org>, linux-arm-kernel@...ts.infradead.org, 
	linux-kernel@...r.kernel.org, Roman Gushchin <roman.gushchin@...ux.dev>, 
	Toshiyuki Sato <fj6611ie@...jp.fujitsu.com>
Subject: Re: [PATCH v5] arm64/module: Use text-poke API for late relocations.

Hi Will,

On Fri, May 30, 2025 at 7:13 AM Will Deacon <will@...nel.org> wrote:
>
> and this would be:
>
>         WRITE_PLACE(place, cpu_to_le32(insn), me);
>

I'm seeing this part give a build error:

arch/arm64/kernel/module.c:158:2: error: cannot take the address of an
rvalue of type '__le32' (aka 'unsigned int')
  158 |         WRITE_PLACE(place, cpu_to_le32(insn), me);
      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
arch/arm64/kernel/module.c:56:28: note: expanded from macro 'WRITE_PLACE'
   56 |                 aarch64_insn_copy(place, &(val),
sizeof(*place));       \
      |                                          ^ ~~~

I can't think of a clean way to get around this and still keep a
combined write helper. Setting an intermediate __le32 in the
reloc_insn_* functions would work but we were trying to avoid that.
Setting an intermediate value inside WRITE_PLACE would also work but
then (I think) it won't work for the data relocations because we'd be
converting a signed into unsigned value. Making WRITE_PLACE a function
instead of a macro also fixes the rvalue problem but then the args
'place' and 'val' have to be of a fixed type so we can't do the
typecasting on 'place' and 'val' has the same signed/unsigned value
problem.

Do you have a suggestion here? In the meantime I can send a v6 that
uses an intermediate __le32 for the instruction relocations.

Thanks,
Dylan