linux-kernel - Re: [PATCH v2 0/7] Module relocation fixes and asm/insn.h header

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] [day] [month] [year] [list]

Message-ID: <CANBLGcxeO0uYutsAHP-TKRCjRqLqbUy-rvK1vcX02eDe0445Hg@mail.gmail.com>
Date:   Wed, 23 Feb 2022 16:53:39 +0100
From:   Emil Renner Berthing <kernel@...il.dk>
To:     Palmer Dabbelt <palmer@...belt.com>
Cc:     linux-riscv <linux-riscv@...ts.infradead.org>,
        Paul Walmsley <paul.walmsley@...ive.com>,
        Albert Ou <aou@...s.berkeley.edu>,
        Peter Zijlstra <peterz@...radead.org>,
        Josh Poimboeuf <jpoimboe@...hat.com>,
        Jason Baron <jbaron@...mai.com>,
        Steven Rostedt <rostedt@...dmis.org>,
        Ard Biesheuvel <ardb@...nel.org>,
        Alexandre Ghiti <alex@...ti.fr>,
        Jisheng Zhang <jszhang@...nel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2 0/7] Module relocation fixes and asm/insn.h header

On Wed, 23 Feb 2022 at 00:15, Palmer Dabbelt <palmer@...belt.com> wrote:
> On Mon, 31 Jan 2022 10:27:13 PST (-0800), kernel@...il.dk wrote:
> > Apologies! I messed up v1. Please consider this patch set only.
> >
> > The first patch removes a bunch of code from the asm/module.h which is
> > included in almost all drivers through linux/module.h. Next are two
> > patches to fix unaligned access when doing module relocations and do
> > proper range checks for auipc+jalr offsets.
> >
> > I'm a little less confident about the following patches, so consider
> > this more of an RFC for those. The idea is to consolidate the RISC-V
> > instruction generation and manipulation similar to arm64's asm/insn.h
> > header.
> >
> > /Emil
> >
> > Emil Renner Berthing (7):
> >   riscv: Remove unneeded definitions from asm/module.h
> >   riscv: Avoid unaligned access when relocating modules
> >   riscv: Fix auipc+jalr relocation range checks
> >   riscv: Add asm/insn.h header
> >   riscv: Use asm/insn.h for module relocations
> >   riscv: Use asm/insn.h to generate plt entries
> >   riscv: Use asm/insn.h for jump labels
> >
> >  arch/riscv/include/asm/insn.h       | 121 ++++++++++++++
> >  arch/riscv/include/asm/module.h     |  87 ----------
> >  arch/riscv/kernel/jump_label.c      |  12 +-
> >  arch/riscv/kernel/module-sections.c |  71 +++++++++
> >  arch/riscv/kernel/module.c          | 237 +++++++++++++---------------
> >  5 files changed, 306 insertions(+), 222 deletions(-)
> >  create mode 100644 arch/riscv/include/asm/insn.h
>
> These generally look good to me, though there's a lot of bit-field
> twiddling so I'll take another look before merging it.  There's a
> handful of minor issues:
>
> * There's a fix in here, mixed into the cleanups.  It's generally best
>   to split those out.

There are two fixes. The 32bit range check on rv64 and unaligned 32bit
access. The code has been like that for years so I was unsure if they
were worth splitting out and adding early. Since you only mention one
I guess that's the range check. I'll send that separately.

> * There's another copy of the insn patterns in our BPF JIT, it'd be nice
>   to clean that up too.  That can be a follow-on, though.
> * It's 2022, but there's some 2020 copyrights.  If this really is old
>   stuff that's OK, I just wanted to check.

Nice catch, but the year is actually correct. These patches have been
well aged in my local repo. The reason is exactly that I never got
around to doing the BPF conversion, so now I decided to just send them
and see if it was worth finishing.

> I'm usually OK just re-ordering patches myself, but I figured I'd have
> to ask about the copyright dates anyway.  LMK if you want to send a v2
> with the fix pulled to the front, and what you want me to do about the
> copyright dates (if you're going to send a v2 then just fix them, but if
> you're not then just telling me is OK).

Thank you. I'll send the range check separately and a v2 converting
the "if (IS_ENABLED(CONFIG_32BIT))" to an #ifdef to avoid the warning
the kernel test robot found.